This present application is also related to co-pending U. S. Patent Application, Application Number 10/778, 896, filed February 17,2004, entitled"Fluid Quality Control Using Broad Spectrum Impedance Spectroscopy" (Atty Docket No.: ETN-026-PAP ; Eaton Ref. No.: 03- ASD-210 (SR) ). Application No. 10/778, 896 is commonly owned by the assignee hereof, and is also hereby fully incorporated by reference herein, as though set forth in full, for its teachings on the use of Impedance Spectroscopy data to monitor and control fluid properties and quality.

BACKGROUND I 1. Field 002 The present invention relates to methods and apparatus for monitoring and controlling the properties of fluids, and more particularly to a method and apparatus for evaluating and improving the performance properties of fluid formulations, such as lubricating fluids blended with additives.

2. Description of Related ylrt 003 When developing fluid formulations, such as lubricating fluids blended with additives, analytical testing is required to ascertain that the properties of the blended lubricant are consistent with intended design properties. Lubricant formulations typically include

selected additives based on the requirements of the intended application for the lubricant. All of the additives included have particular performance properties that are exploited in the final fluid formulation design.. To improve the performance properties of fluid formulations, it is necessary to observe, measure and understand the combined functional effects of the additives on the physical properties of the formulation. Modifying the additives and/or the blended fluid formulation, based on observation and understanding of the additive effects, can improve both the performance and cost effectiveness of the lubricant product.

004 Systems for in-situ (e. g, performed in an operating system, such as an engine or transmission) monitoring the properties of lubricating fluids are known. One such system is disclosed in U. S. Patent No. 6,278, 281 entitled"Fluid Control Monitor"issued to Bauer, et al., Bauer describes a technique employing AC electro-impedance spectroscopy (referred to hereinafter as impedance spectroscopy or"IS"), and is implemented using probe electrodes that are placed in contact with a fluid under test. The method of operation includes making IS measurements at a first frequency that is less than 1 Hz and at a second frequency that is greater than 1 Hz, comparing the two IS measurements, and declaring a"pass"or"fail" condition based on a previously determined empirical relationship. This prior art lubricating fluid monitoring system disadvantageously effectively analyzes only a single property of the IS spectra based on the difference of two IS measurements. Consequently, the IS measurement technique taught by Bauer is not capable of determining the complex properties of compound fluids, as is required when designing fluid formulations having a plurality of additives.

005 A co-pending and commonly assigned U. S. Patent Application, Application Number 10/723, 624, filed November 26,2003, entitled"FLUID CONDITION MONITORING USING BROAD SPECTRUM IMPEDANCE SPECTROSCOPY,"teaches a broad spectrum IS method for determining IS parameters relating to the bulk and interfacial properties of fluids. When developing a performance-based combination of one or more additives in a base fluid, referred to herein as a"formulation,"interactions between a single additive and the base fluid, or between a plurality of additives themselves, can cause unexpected results in the performance properties of the formulation. Because the performance of the formulation depends on both the properties of both its bulk and interface, a method is needed to accurately evaluate these properties with regard to the effects of the additives, singly and in plurality. Therefore, a need exists for a method and apparatus evaluating and improving fluid formulations and additives.

SUMMARY 006 A method and apparatus evaluating and improving the properties of fluid formulations is disclosed. The inventive concept employs impedance spectroscopy (IS) measurements and data analyses to determine IS parameters for fluids representative of a fluid formulation. Correlations are determined between the IS parameters and the properties of the fluids, and a new fluid formulation may be developed based on these correlations.

007 In one exemplary embodiment, IS measurements at three or more frequencies are made using probe electrodes in contact with fluids representative of a fluid formulation. The IS data are analyzed using statistical techniques, equivalent circuit modeling techniques, or a combination thereof. The data analysis provides at least one IS parameter indicative of at least one fluid property for the fluids. At least one correlation is determined between one or more IS parameters and one or more properties of the fluids. A new fluid formulation is developed, responsive to correlations between the IS parameters and the properties of the fluids.

FLUID CONDITION MONITORING USING BROAD SPECTRUM IMPEDANCE SPECTROSCOPY Background of the Invention II 008 The present invention relates to monitoring the condition of a substance using impedance spectroscopy to indicate in real or near real time, i. e., while the substance is being used, the physio-chemical condition of a substance based on a correlation of measurements from electrical signals, using a statistical technique, to previously determined values.

009 The use of impedance spectroscopy techniques to monitor fluid conditions is previously known. For example, U. S. Patent Nos. 6,433, 560,6, 380,746, 6,377, 052, and 6,278, 281, along with U. S. Published Application 2003/0141882 of Zou et al. , all assigned to the assignee of the present invention, all teach different configurations of electrodes for measuring current and then computing values for impedance. Further, the'281 patent teaches comparing the difference in currents at two frequencies with known bulk ad interfacial impedance measurements to determine fluid conditions. Similarly, the'052 and'746 patents teach comparing the difference in impedance values calculated from exciting electrodes at two frequencies to determine fluid condition.

010 The value of using impedance spectroscopy to monitor fluid condition lies in the fact that it is desirable to be able to determine when a fluid, for example, an engine lubricant, has degraded to the point where it has either exhausted or come close to exhausting its useful life. Similarly, it is desirable to know how many hours of useful life remain with respect to a fluid sample. For an application such as monitoring the condition of an engine lubricant, it is desirable to be able to monitor the fluid condition while the engine is operating, as opposed to performing tests in a laboratory.

Oil Lubricating fluids comprise three basic components: (1) base stock, (2) additives, and (3) contaminants. These components are known to influence the bulk and interfacial properties of the lubricant. Lubricating fluids also possess interfacial properties, such as wear protection and corrosion protection, that are present at the interface between the fluid and the

metal it protects. Different portions of the impedance spectrum correspond to bulk and interfacial properties. Previous applications of impedance spectroscopy have measured bulk properties and interfacial properties separately, but the capability of measuring the two together has not been previously appreciated. Accordingly, it would be desirable if bulk and interfacial properties could be measured together in order to give a more complete picture of the engine lubricant.

012 Existing methods of monitoring fluid conditions using impedance spectroscopy fail to contemporaneously measure a plurality of fluid properties. Rather, prior art methods, including those disclosed by the above-mentioned patents and publication, calculate impedance in different ways in order to determine a value for one fluid property. It can be shown that up to 90% of available information is not utilized when single parameter measurement techniques are employed. Thus, there is a need for systems and methods implementing a multiple parameter function or functions able to make quantitative measures of a broad range of fluid condition metrics.

Brief Summary of the Invention 013 The present invention uses impedance spectroscopy to determine the amounts of additives, contaminants, and other components that are present in a fluid. In some embodiments, the present invention is applied to engine lubricants.

014 The invention comprises using the impedance spectrum to monitor the condition of a fluid in real time by disposing an electrode mechanism in the fluid and exciting the electrode mechanism sequentially with a specified number of alternating voltages, wherein each of the alternating voltages is at a different frequency in a range of frequencies. The invention measures the current in the electrode mechanism at each of said frequencies in the range of frequencies. From the measurement of current at each of the frequencies, real and reactive impedance at each of the frequencies is calculated.

015 In one embodiment, the invention comprises a method of analyzing a substance, comprising the steps of : (1) generating first and second plots of spectra over a range of frequencies; (2) creating a third plot that comprises the first and second plots by sequentially

assigning x-values to selected frequencies in the first plot and selected frequencies the second plot; and (3) repeating steps (1) - (2) at least once to generate a plurality of third plots. Next, the method comprises building a spectral matrix that comprises data taken from the plurality of third plots and a result matrix comprising known quantities of a plurality of components in the substance. A Principal Component Analysis is performed on the spectral matrix to identify at least one principal component having significant influence on the spectral matrix.

A reduced spectral matrix having at least one column, wherein each column in the reduced spectral matrix is associated with a principal component having significant influence on the spectral matrix, is next created. Then, a statistical technique is performed that uses the reduced spectral matrix and the result matrix to create at least one prediction equation. The at least one prediction equation is then used to predict at least one property in a second substance in real time.

016 In some embodiments, at least one datum derived from a Nyquist plot is added to the spectral matrix. The at least one datum derived from a Nyquist plot may include at least one datum from the bulk region of the Nyquist plot and at least one datum from the interfacial region of the Nyquist plot. This could also include parameters derived from equivalent circuit modeling of the impedance spectra.

BRIEF DESCRIPTION OF THE DRAWINGS 017 FIGURE 1 is a block diagram of a simplified system evaluating and improving fluids based on impedance spectroscopy.

018 FIGURE 2 is an illustration of typical impedance spectroscopy data shown as a Bode plot.

019 FIGURE 3 is an illustration of typical impedance spectroscopy data shown as a Nyquist plot.

020 FIGURE 4 is an illustration of an equivalent circuit for modeling impedance spectra data.

021 FIGURE 5 is a plot of typical impedance spectroscopy data obtained for a base oil and for two concentrations of an anti-oxidant in the base oil.

022 FIGURE 6 is an illustration of a first alternative equivalent circuit for modeling impedance spectra data..

023 FIGURE 7 is a plot of typical impedance spectroscopy data obtained for two concentrations of a calcium sulfonate detergent in a base oil.

024 FIGURE 8 is a plot of typical impedance spectroscopy data obtained for two concentrations of a magnesium sulfonate detergent in a base oil.

025 FIGURE 9 is a first time series of impedance spectroscopy data for thermal. decomposition in a formulation containing ZDDP.

026 FIGURE 10 is a second time series of impedance spectroscopy data for thermal decomposition of a formulation including ZDDP.

027 FIGURE 11 is an illustration of a second alternative equivalent circuit for modeling impedance spectra data.

028 FIGURE 12 is a plot of an equivalent circuit parameter Rfluid versus time for thermal decomposition of a formulation including ZDDP.

029 FIGURE 13 is a plot of an equivalent circuit parameter Ri versus time for thermal decomposition of a formulation including ZDDP.

030 FIGURE 14 is a plot of an equivalent circuit parameter Cfilm versus time for thermal decomposition of a formulation including ZDDP.

031 FIGURE 15 is a plot of an equivalent circuit parameter Rfilm versus time for thermal decomposition of a formulation including ZDDP.

032 FIGURE 16 is a plot of an equivalent circuit parameter film versus time for thermal decomposition of a formulation including ZDDP.

033 FIGURE 17 is a plot of impedance spectroscopy data obtained for a first formulation comprising a base oil plus detergent and a second formulation comprising a base oil, ZDDP, a dispersant, and a detergent.

034 FIGURE 18 is a time-dependent plot of FTIR oxidation data at 1833-1640 inverse- cm obtained from an oxidation test for a first formulation comprising a base oil plus detergent and a second formulation comprising a base oil, ZDDP, a dispersant, and a detergent.

035 FIGURE 19 is a plot of Total Acid Number (TAN) data versus FTIR oxidation data at 1833-1640 inverse-cm for a formulation consisting of a base oil, ZDDP, a dispersant, and a detergent.

036 FIGURE 20 is a plot of FTIR Peak Area data at 660 and 970 inverse-cm versus Hot Oil Oxidation Test (HOOT) sample number, obtained for a formulation consisting of a base oil, ZDDP, a dispersant, and a detergent.

037 FIGURE 21 is a plot of C03 Peak Area FTIR data at 860 inverse-cm (cm-1) versus Hot Oil Oxidation Test sample number, obtained for a formulation consisting of a base oil, ZDDP, a dispersant, and a detergent.

038 FIGURE 22 shows a plurality of plots of impedance spectroscopy data, each plot representing a selected Hot Oil Oxidation Test process time, obtained for a formulation consisting of a base oil, ZDDP, a dispersant, and a detergent.

039 FIGURE 23 is a plot of exemplary impedance spectroscopy data illustrating evaluation points for a differential impedance parameter.

040 FIGURE 24 is a plot of a differential impedance parameter LOG dZmaglow versus Hot Oil Oxidation Test sample number, obtained for a first formulation consisting of a base oil plus detergent and a second formulation consisting of a base oil, ZDDP, a dispersant, and a detergent.

041 FIGURE 25 is a plot of a differential impedance parameter LOG dZmaglow versus FTIR data at 645 inverse-cm, obtained for selected samples from a Hot Oil Oxidation Test sequence, for a formulation consisting of a base oil, ZDDP, a dispersant, and a detergent.

042 FIGURE 26 is a plot of a differential impedance parameter LOG dZmaglow versus FTIR data at 970 inverse-cm, obtained for selected samples from a Hot Oil Oxidation Test sequence, for a formulation consisting of a base oil, ZDDP, a dispersant, and a detergent.

043 FIGURE 27 is a plot of a differential impedance parameter LOG dZmaglow versus FTIR data at 860 inverse-cm, obtained for selected samples from a Hot Oil Oxidation Test sequence, for a formulation consisting of a base oil, ZDDP, a dispersant, and a detergent.

044 FIGURE 28 is a plot of a differential impedance parameter LOG dZmaglow versus FTIR oxidation data at 1833-1640 inverse-cm, obtained for selected samples from a Hot Oil Oxidation Test sequence, for a formulation consisting of a base oil, ZDDP, a dispersant, and a detergent.

045 FIGURE 29 is a plot of the correlation between FTIR oxidation data at 1833-1640 inverse-cm and impedance spectroscopy parameter data analyzed by GMDH with respect to oxidation.

046 FIGURE 30 is an illustration of a third alternative equivalent circuit for modeling impedance spectra data.

047 FIGURE 31 is a plot of the correlation between FTIR oxidation data at 1833-1640 inverse-cm and impedance spectroscopy equivalent circuit parameter data analyzed by GMDH with respect to oxidation.

048 FIGURE 32 is a flowchart diagram illustrating an exemplary method for evaluating and improving the properties of fluids in accordance with the present inventive concept.

Brief Description of the Drawings II 049 FIG. 33 depicts a system by which real time predictions of fluid conditions are made.

050 FIG. 34 describes a system that may be used for collecting data used to develop an information library used to predict fluid conditions.

051 FIG. 35 depicts a flowchart providing an overview of the method by which information library 120* is developed and used to predict fluid conditions in real time.

052 FIG. 36 provides a flowchart describing a method for building a matrix of spectral data, i. e. , a spectral matrix.

053 FIG. 37 provides a flowchart describing a method for building a result matrix.

054 FIG. 38 shows the process by which the present invention uses Principal Component analysis (PCA).

055 FIG. 39 shows Bode plots of resistive and reactive impedance spectra superimposed on the same graph.

056 FIG. 40 shows an example of combined Bode plots on a graph.

057 FIG. 41 provides an example of a result matrix 900* populated with empirically achieved data representing conditions of an engine lubricant.

058 FIG. 42 shows regression coefficients plotted on a graph.

059 FIG. 43 shows a reduced spectral matrix.

060 FIG. 44 shows a chart 1200* giving examples of data, derived from a Nyquist plot of resistive impedance versus reactive impedance, that can be added to a combined Bode plot.

061 FIG. 45 shows a graph comprising three Nyquist plots for three different lubricating fluid samples.

062 FIG. 46 describes an alternative embodiment in which data from each of the Bode plots of real and reactive impedance spectra are processed separately, with results then combined for use in predicting fluid properties.

DETAILED DESCRIPTION 063 Throughout this description, embodiments and variations are described for the purpose of illustrating uses and implementations of the inventive concept. The illustrative description should be understood as presenting examples of the inventive concept rather than as limiting the scope of the concept as disclosed herein.

064 FIGURE 1 shows a block diagram of a simplified formulation evaluation and improvement system 100 based on fluid properties determined using IS measurements and analysis. A fluid container 102 may comprise a glass or metal container, blending tank, or an in-line container, such as a pipe. The container 102 contains a fluid 104 representative of a fluid formulation, such as a lubricant with additives. The fluid 104 may be a fluid selected from a plurality of fluids representative of one or more fluid formulations. For use in performing IS measurements and experimental operations, the fluid 104 is typically maintained at a constant temperature using a temperature sensor and controller 118. The temperature sensor and controller 118 is in contact with the container 102 or the fluid 104.

The temperature setting for the controller 118 may be set by a system operator (not shown), or it may be set using a data processing system 110 operatively coupled to the controller 118. A measured temperature from the device 118 may be conveyed to the processing system 110.

065 An IS probe 106 is in contact with the fluid 104. Many suitable IS probes are known to those skilled in the electrochemical arts. U. S. Patent No. 6,278, 281 entitled"Fluid Control Monitor"issued to Bauer, et al. , on August 21,2001, describes a plurality of suitable electrode probes that may be used in conjunction with the present invention. An exemplary IS probe device design utilizing concentric tubular electrodes is disclosed in U. S. Patent Application No. 2003/0141882,10/060107, filed January 31,2002, entitled"Probe Assembly for a Fluid Condition Monitor and Method of Making Same. "Both the issued patent and the published patent application cited above are commonly owned by the assignee hereof, and both are hereby fully incorporated by reference herein as though set forth in full, for their teachings on IS probe devices, and for their teachings on methods and equipment relating to IS measurements of fluids. Other exemplary IS probe designs include parallel plate electrodes, interdigitated electrodes and spiral electrodes.

066 As shown in FIGURE 1, the IS probe 106 is operatively coupled to an impedance spectrometer 108. The construction and operation of impedance spectrometers are well known to persons skilled in the electrochemical arts, and commercial impedance spectrometers are available. IS instrumentation generally comprises an array of impedance and frequency response analyzers, as well as"lock-in"amplifiers. The equipment provides a source of AC signals of varying frequency. The IS equipment also provides circuitry for detecting the magnitude of electric current conducted through the sample. An exemplary combination of IS instrumentation may include an EG&G Potentiostat/Galvanostat Model 283 (EG&G is a Division of URS Corporation, of San Francisco, California), and a Solartron Impedance/Gain-Phase Analyzer Model 1260 (hereinafter, Solartron 1260-Solartron Analytical is a member of the Roxboro Group plc, of Cambridge, United Kingdom). The Solartron 1260 provides an AC signal of varying frequency. Signal levels ranging from 125 mV (in fully formulated lubricating fluids) to 1000 mV (in the base fluid) have been found to produce well-defined IS data. The impedance spectrometer 108 is operatively coupled to a data processing system 110. In one embodiment, the data processing system 110 comprises a personal computer (PC). IS data acquisition may be accomplished using commercial PC- based computer programs such as"Z-PlotTM", and"Z-ViewTM" (see, for example, the operating manual entitled"Zplot for Windows, "Scribner Associates, Inc. , version 2.1, 1998), as well as other software that may be custom-developed by persons skilled in the arts of scientific data acquisition. These PC-based computer programs for IS data acquisition are well known in the art.

067 As shown in FIGURE 1, in one embodiment, the data processing system 110 includes a central processing unit (CPU) 112 and a memory 114, both of which are operatively coupled to receive data from the impedance spectrometer 108, and configured to output data to a formulation design operator 116. The memory 114 stores software instructions used for acquiring data. The memory 114 also may be used to store data from measurements performed sequentially on a plurality of fluids 104, software for data analysis functions, and data analysis results for a plurality of fluids 104. These functions are described more fully hereinbelow.

068 The formulation design operator 116 may comprise one or more persons, an apparatus, or any combination thereof. The results of IS data analyses performed by the data

processor 110 are output to the operator 116, and the operator 116 may develop or produce a modified, improved, or new fluid formulation 120, in accordance with teachings hereinbelow.

Exemplary Data 069 Referring now to FIGURE 2, typical IS data are illustrated in the form of Bode plots 200,202, which are well-known to persons skilled in the electronic arts. The upper plot 200 shows the resistive part of impedance versus the logarithm of frequency. The lower plot 202 shows the reactive part of impedance versus the logarithm of frequency. These exemplary data represent an IS spectrum for a given lubricant fluid.

070 Although the data are shown in FIGURE 2 as continuous curves 200,202, persons skilled in the arts of scientific data acquisition will understand that the curves 200,202 actually represent a plurality of connected point measurements. For example, the curves 200, 202 may comprise ten data points per decade. Alternatively, far fewer points may be employed, as for example are used in U. S. Patent No. 6,278, 281 cited above, wherein only two points are used. In some exemplary embodiments of the present invention, IS spectra data includes at least three points, and typically tens or hundreds of points are used. More than a few hundred points typically are not required. For the practice of the present invention, the IS spectra points will generally (although exceptions may occur) span a frequency range sufficient to represent IS parameters associated with both the bulk fluid and the interface between the fluid and the electrode (the fluid/electrode interface). These IS parameters, and their frequency ranges, are described in more detail below. In general, the IS spectra includes frequencies that are both above and below 1 Hz.

071 FIGURE 3 illustrates the same IS data depicted in FIGURE 2 using a Nyquist plot 300, which is well known to persons skilled in the electronic arts. The data show a minimum in the reactance, known as the"Nyquist minimum". As described below in more detail, data for frequencies lower than the Nyquist minimum can be associated with an interface reactivity caused by electrically-active phenomena occurring at the fluid/electrode interface. Data for frequencies greater than the minimum can be associated with the electrical properties of the fluid bulk, and with a fluid film present on the electrode.

Exemplary Data Analysis Using Equivalent Circuit Modeling 072 FIGURE 4 shows an equivalent circuit model 400 that may be used to assist data analysis in accordance with one embodiment of the present invention.

073 As shown in FIGURE 4, there is an impedance, associated with the bulk fluid electrical properties, comprising Rfluid and CflUidn having a time constant-c fluid as follows: 074'U fluid = R fluid C fluid 075 Persons skilled in the electrochemical arts shall recognize that for many fluids, such as lubricating fluids, the bulk electrical transport occurs primarily via ionic conduction.

For fluids exhibiting ionic conduction there is also impedance associated with the fluid/electrode interface. The electrical phenomena resulting from contact between the fluid and the electrode surface is represented in FIGURE 4 by elements shown between the vertical dashed lines. The interfacial impedance includes a capacitance Cdl created by polarization arising from a double layer formation, as ions orient themselves in response to the presence of the charged metal surface. Because the fluid is an ionic conductor and the electrode an electronic conductor, a charge transfer reaction must be operative for current to flow across the fluid/electrode interface. This current leakage across the fluid/electrode interface is represented in FIGURE 4 by a parallel path.

076 As shown in FIGURE 4, charge transport is accompanied by energy that is required for charge transfer, represented by a resistance R ; t, and possibly adsorption, and possibly diffusion as detectable steps in the overall process. The specific nature of this electro-chemical reaction determines the form of the detected impedance associated with this reaction path. If the reaction is fast, a diffusion-limited character is often evident.

Conversely, reactions involving slow kinetics, i. e., reactions involving rate-limiting adsorption of intermediate species, yield impedance character of a different form. When sufficiently defined from measured data, an interface time constant value T i can be defined in a manner that is analogous to the bulk value. The interface time constant i i reflects either the heterogeneous rate constant of the reaction at the interface, or the magnitude of the diffusion coefficient for the reacting species (reactant or product). This provides an effective value for

the net reactivity of species at the fluid/electrode interface, and is calculated from the following relationship: 077 T i = (R i Q i) 078 where observed values of n range between 0.5-1. 0. The observation of diffusion as rate-limiting implies that the rate of reaction at the surface is fast as the potential is modulated by an AC signal, such that local depletion (or accumulation) of the surface-active species occurs. The presence of this diffusion gradient in concentration is observed in the measured impedance as a Constant Phase Element (CPE), denoted as Q i, where the CPE exponent n is equal to 0.5. The determination of n-values for interfacial phenomena at or close to unity is indicative of adsorption, rather than diffusion, as the rate-limiting step at the electrode. To summarize, n-values approximating 0.5 indicate that diffusion is the rate limiting process, n-values approximating 1.0 indicate that adsorption and surface reaction is the rate limiting process, while values for n between 0.5 and 1.0 indicate that both processes are significant as regards the rate limiting process.

079 Referring again to FIGURE 4, another interface time constant, T film may also be observed. For appropriately configured measurements and fluid samples, this value is determined according to the following equation: 080 T film = R film C film, 081 where the resistance R film, and the capacitance C film, reflect the electrical properties and geometry of the fluid film that may form on the electrode.

082 Impedance data analysis in accordance with the above described equivalent circuit model may be performed by the data processing system 110 of FIGURE 1, using, as an example, the well known Complex Non-Linear Least Squares fitting technique employed by the PC-based computer program,"Equivalent Circuit", written by Boukamp (B. A. Boukamp, "Equivalent Circuit (Equivckt. A46')" User's Manual, Dept. of Chemical Technology, Universiteit Twente, Netherlands, 1988 and 1989. This fitting technique is also described in an article written by B. A. Boukamp, "A Nonlinear Least Squares Fit procedure For Analysis of Immitance Data of Electrochemical Systems"Solid State Ionics, Vol. 20, pp. 31-44,1986).

The above-cited User's Manual and article are incorporated by reference herein for their teachings on data analysis.

083 The results of the equivalent circuit data analysis technique described above include values for the circuit elements shown in FIGURE 4. These values are referred to herein generally as"IS parameters, "and more specifically as"equivalent circuit IS parameters. "These equivalent circuit IS parameters may include, without limitation, the following: a bulk fluid resistance Rfluid, a bulk fluid capacitance Cfluid, a bulk fluid time constant T fluid, an interface capacitance Cdl, a charge transfer resistance RWt, an interface time constant T i, an interface resistance Ri, a Constant Phase Element Q i, a Constant Phase Element exponent n, a film time constant T film, a film resistance R film, and a film capacitance C film. Equivalent circuit IS parameters, and other IS parameters described below, may be related to fluid properties, in accordance with teachings presented in more detail below. The equivalent circuit model and corresponding data analysis given above is by way of example only. The scope of the present invention also encompasses the use of other equivalent circuit models and associated analysis methods, software, and techniques that suitably represent IS measurements and properties of fluids.

Exemplary Data Analysis Using Statistical Techniques 084 In addition to the equivalent circuit data analysis technique described above, the data analysis processes implemented by the data processing system 110 may include one or more well known statistical techniques. For example, the data analysis performed by the data processing system 110 may include, without limitation, the following techniques : Principal Component Analysis (PCA), Multivariate Least Squares Regression (MLR), Principal Component Regression (PCR), Group Method for Data Handling (GMDH), Pattern Recognition analysis, Cluster analysis, and Neural Net analysis. A description of these techniques in reference to IS data analysis is disclosed in the incorporated and co-pending U. S. Patent Application, Application Number 10/723,624, filed November 26,2003, titled "FLUID CONDITION MONITORING USING BROAD SPECTRUM IMPEDANCE SPECTROSCOPY" (Eaton Ref. No. 03-ASD-255 (SR) ). This above-incorporated application is set forth in full in Appendix A of the present application.

085 Exemplary commercially available software that may be used by the data processing system 110 of FIGURE 1 in implementing the processes required for the statistical analysis techniques include the following software applications:"The UnscramblerTM"by Camo Process, AS; Norway; Spectrum Quant+'" by PerkinElmer, Inc., Norwalk, Connecticut; and MatLab by MathworksTM, Inc. , Natick, Massachusetts.

086 Data analysis using statistical techniques also provides IS parameters that may be related to fluid properties. Examples of such IS parameters obtained from statistical techniques are described in reference to exemplary data presented hereinbelow.

Exemplary Implementations 087 This section presents data and analyses that exemplify embodiments of the present inventive concept in conjunction with lubricating oil formulations. IS measurements are performed on the fluids, the IS data obtained are analyzed to determine IS parameters that are indicative of fluid properties, and the fluid properties are correlated to formulation components and properties. The inventive process provides information that may be used to modify formulations and develop new formulations.

Example 1: Base Fluid Plus Antioxidant 088 FIGURE 5 illustrates IS data obtained for a base fluid and formulation comprising two concentration levels of a Phenolic antioxidant (AO) added to the base fluid. LoTreat of the phenolic antioxidant is 0.5% and HI treat is 1.0%. The base fluid, EM 150 SN, is a mildly hydrofinished 150 solvent neutral, hydrocarbon base stock. The data exhibits a semicircular trace at high frequencies, indicative of the electrical properties of the fluid bulk, and a low frequency tail, indicative of an additional interfacial contribution to the measured impedance.

As persons skilled in the art will understand, this is the expected lumped response when the time constant values for the bulk and interface are within two orders of magnitude.

089 The data of FIGURE 5 may be analyzed according to a simplified equivalent circuit model as illustrated in an equivalent circuit model 600 shown in FIGURE 6, and the IS parameters obtained thereby are shown in TABLE 1.

TABLE 1. IS Parameters for Base Fluid and Base Fluid Plus AO FLUID Rfl, id Cfluid'Cfluid Ri Qi Ti n (ohms) (farads) (sec) (ohms) (farads) (sec) Base Fluid 1. 34 x 101° 1. 4 x 101° l8g 7. 83 x 109 9. 46 x10 1° 7. 41 1. 0 Base Fluid 7. 5 x 109 1. 39 x 10-l° 1. 05 3. 85 x 109 9. 86 xlO-t° 3. 79 1. 0 plus AO, low conc. Base Fluid 6. 25 x 109 1. 42 x 10-1° 0. 87 2. 76 x 109 9. 23 xlO 10 2. 54 1. 0 plus AO, high conc.

090 As shown in TABLE 1, adding AO to the base fluid reduces the resistance of the bulk fluid without significantly modifying the capacitance. In the Table, the ion concentration increase is seen from the resistance decrease, since the mobility relates to viscosity, and there was no observable viscosity difference between the formulations. Likewise, the interfacial impedance is reduced by the AO additive. The CPE exponent n (described above) is found to be unity for all three fluids. As noted above, the determination of n-values for interfacial phenomena at or close to unity is indicative of adsorption (interface reaction rate) as the rate- limiting step at the electrode. Thus, the change in interface impedance caused by the AO additive reflects a change in the net rate of reaction (approximately a factor of two increase), as is determined by comparing the calculated interface time constant values to that of the base fluid.

091 In this example, the relative interface reaction rate exemplifies a fluid property, responsive to formulation properties such as AO additive concentration, that may be obtained from the IS parameters shown in TABLE 1. Further, the relative bulk resistivities of the fluids are indicated by the IS parameters Rfluid shown in the table. Bulk resistivities are modified by ion concentrations and ion mobilities. Thus, ion concentrations and ion mobilities are also

exemplary fluid properties, dependent on formulation additives, that can be determined from the IS parameters.

Example 2: Base Fluid Plus Detergents 092 FIGURES 7 and 8 show IS data illustrating the effects of two different detergent additives, each at two different concentration levels, using the same base fluid as used in Example 1 described above. The base fluid data are not shown in these Figures because it would be off scale. Because the base fluid is the same as used in Example 1 above, the data illustrated in FIGURE 5 is representative thereof. For FIGURE 7 the detergent additive is calcium sulfonate, and for FIGURE 8 the detergent additive is HOB magnesium sulfonate.

Lo treat of the detergent is 3% and hi treat is 6%. The detergent in Figure 7 was a commercially available neutral, synthetic calcium alkylbenzene sulfonate. The detergent in Figure 8 was a commercially available highly overbased, synthetic magnesium alkylbenzene sulfonate. The terminology HOB indicates a highly overbased detergent type. The IS parameters obtained from the data of FIGURES 7 and 8 are shown in TABLE 2 below. For these data analyses, the simplified equivalent circuit model illustrated in FIGURE 6 is used.

TABLE 2. Exemplary IS Parameters for Base Fluid and Base Fluid Plus Detergents FLUID Rfluid Cfluid Tnuid Ri Qi li n (ohms) (farads) (sec) (ohms) (farads) (sec) Base Fluid 1. 34 x 101° 1. 4 x 10-1° 1. 88 7. 83 x 109 9. 46 x10 10 7. 41 1. 0 Base Fluid 8. 39 x lO6 1. 28 x 10 10 1. 11 x 3. 47 x 107 3. 85 x10 7 19. 6 0. 87 plus Ca 10-3 sulfonate, low cone. Base Fluid 4. 44 x 106 1. 26 x 10 10 5. 51 x 4. 73 x 107 5. 90 x10-7 69. 9 0. 78 plus Ca 10-4 sulfonate, high conc. Base Fluid 9. 96 x 106 1. 27 x 10 10 1. 23 x 8. 34 x 106 2. 03 x10 7 2. 88 0. 5 plus HOB Mg 10-3 sulfonate, low conc. Base Fluid 5. 03 x 106 1. 30 x 10 10 6. 33 x 4. 78 x 106 5. 18 x10 7 6. 13 0. 5 plus HOB Mg 10-4 sulfonate, high conc.

093 The exemplary IS parameters shown in Table 2 indicate that the detergent additives reduce the bulk resistivity of a formulation, relative to the bulk resistivity of the base fluid, by approximately three orders of magnitude, with only a small effect upon the bulk capacitance. The changes in bulk resistivity can be interpreted as an increase in ion concentration. This ion concentration implies the amount of free neutral salts in the bulk formulation. The 6% concentration of HOB magnesium sulfonate has a neutral salt concentration between the 3% neutral sulfonate and the 6% neutral sulfonate. On the other hand, the interface resistivity implies the amount of salts on the interface. The concentration of interface ions relates to the amount of detergent's carbonate in the formulation. Other data

(not shown in Table 2) indicates that engine oil that has been formulated with a complete set of additives typically exhibits interfacial behavior that is limited by the rate of transport (diffusion) to/from the metal electrode surface (i. e. , as noted above, the CPE exponent n has a value of approximately 0.5). The exemplary IS parameters of TABLE 2 indicate that detergent addition alone causes interfacial behavior indicative of adsorption as the rate limiting factor (and/or a mixture of adsorption and diffusion, i. e. corresponding to CPE exponent n-values in a range approximating 0.75), resulting in longer interfacial time constant values, typically on the order of 10 to 100 seconds. While this fluid property was found for the majority of the detergents investigated, the HOB Mg sulfonate causes interfacial behavior indicative of higher reactivity, where diffusion is rate-limiting, as indicated by CPE exponent n-values of 0.5 in TABLE 2 for the HOB Mg sulfonate formulations. This result suggests that other, undetermined, surface-active species may be present in these formulations, perhaps incorporated via the process oil used to carry the HOB Mg sulfonate additive. This result thus illustrates that the fluid properties determined using the present inventive method provide useful information for screening additive composition and performance.

094 In this example, as in the previous example, the relative interface reaction rate exemplifies a fluid property, responsive to the formulation design and additives, that may be obtained using the exemplary IS parameters shown in TABLE 2. Likewise, the relative bulk resistivities for the fluids can be determined from the IS resistance values Rfluid shown in Table 2. Thus, the bulk resistivity (indicative of ion concentrations and mobilities) is also an exemplary fluid property that can be determined from the IS parameters according to this example. The product of ion concentrations and ion mobilities is another fluid property that can be determined. Ion mobility is related to viscosity of the fluid, which can be observed by other means well known to persons skilled in the arts of petroleum engineering. If it is known that the viscosity and ion mobilities are relatively constant, than the inventive method can be used to determine changes in ion concentrations based on observed changes in Rfluid and bulk resistivity.

Example 3: Thermal Decomposition of ZDDP in a Base Fluid 095 Zinc dithiodialkylphosphates (ZDDPs) comprise multifunctional additives used in lubricating fluids. They act as antioxidants, improving the wear inhibition of the lubricant, and protecting metals against corrosion. In this example, a formulation including a high-level concentration of secondary ZDDP additive is heated to 120 degrees Celsius for various periods to observe changes in fluid properties caused by a thermal decomposition test. Lo treat of the ZDDP is 0.62% and hi treat is 1.24%. The term"secondary"is a compound/product definition of the ZDDP that describes its chemical make-up of the alkyl group attached to the phosphorous. An alkyl group can be primary, secondary, tertiary or aryl. The present example teaches the use of IS parameters in determining changes in fluid properties as a function of time during thermal decomposition.

096 In this example, the base fluid containing the high treat (concentration) level of secondary ZDDP is heated to a temperature of 120 degrees Celsius. Following IS probe immersion, IS impedance measurements are conducted for a period of six hours at temperature. Exemplary IS data measured at 0.3, 1.0, 2.0 and 3.0 hours at temperature are shown in FIGURE 9. Exemplary IS data measured at 3.0, 4.0, 5.0 and 6.0 hours at temperature are shown in FIGURE 10. These data indicate that the bulk impedance of the fluid decreases for approximately 3 hours (FIGURE 9), and then increases for the balance of the testing period (FIGURE 10). In the absence of viscosity changes, this behavior suggests the initial generation of ions in the fluid, followed by depletion of ions, as exposure of the fluid to the metal electrode of the IS probe ensues at temperature. Simultaneously, as described further below, the character of the interfacial contribution to the measured impedance is found to emerge as the bulk resistance drops, changing only slightly after two hours at temperature.

097 Equivalent circuit modeling of these IS data (FIGURES 9 and 10) shows that three IS parameter time constants are present for measurements beginning at and subsequent to one hour at temperature. The corresponding equivalent circuit model 1100 is illustrated in FIGURE 11.

TABLE 3. Base Fluid Plus Secondary ZDDP, IS Parameters as a Function of Time at 120 Degrees Celsius TllW Pfluid Cfluid'Ufluid Cfl. Rfl.'Cfl. Ri cj Ti (hours) ohms (farads sec (hours) (ohms) (farads) (see) (farads) (ohms) (sec) (ohms) (farads) (sec) 0. 3 2. 26 x 1. 23 x 2. 07 x not not not 1. 52 2. 27 x 3. 46 108 10 1° 10 z present present present X10 o7 1 0-7 1. 0 1. 75 x 1. 20 x 2. 06 x 1. 68 x 1. 41 2. 37 x 2. 29 2. 79 x 1. 33 lo8 lo-'° lo los xlo'lo'Xlo'lo8 2. 0 1. 24 x 1. 16 x 1. 44 x 2. 76 x 8. 49 2. 35 x 3. 49 2. 72 x 0. 951 10g 10 1 10 z 10 8 x106 10 1 x10 10 8 3. 0 1. 14 x 1. 17 x 1. 32 x 3. 22 x 6. 74 2. 17 x 4. 03 2. 79 x 1. 12 108 l0-lo 10-2 10-8 x106 10-1 x107 10-8 4. 0 1. 19x 1. 17x 1. 37x 3. 29x 6. 48 2. 13x 3. 99 2. 96x 1. 18 108 10-1° 10-2 10-S x106 1°-1 x107 10-8 5. 0 132 x 1. 17 x 1. 54 x 3. 00 x 6. 80 2. 04 x 4. 24 2. 92 x 1. 24 10"10''° 10' 10'"xl 10''xl0 10'" 6. 0 1. 44 x 1. 23 x 1. 70 x 2. 98 x 6. 53 1. 94 x 4. 25 4. 71 x 2. 0 1o8 l0-lo 10-2 10-8 x106 1°-1 x107 10-8

098 The resistance Rfl,, id (associated with bulk resistivity of the fluid) is shown as a function of time at temperature in the plot 1200 of FIGURE 12, where the decrease from 0.3 to 3 hours, followed by an increase from 3.0 to 6.0 hours, as described above, is evident. The corresponding changes in bulk fluid capacitance Cfluid are within a range of about 5 % or less, as shown in TABLE 3 above. These exemplary results illustrate how the present inventive method may be used in determining changes in fluid properties (e. g. , ion concentration and - 25-

mobility) resulting from changes in formulation composition (e. g., chemical changes due to thermal decomposition).

099 FIGURE 13 illustrates a plot 1300 of the interfacial resistance Ri as a function of time at temperature. These data show a rapid increase for about 2 hours followed by a slower rate of increase beginning at about 3 hours. This behavior appears to reflect reactivity derived from adsorption that is governed by the concentration of surface active portion of the reactants in the fluid, which are decreasing with time due to oxidation or thermal decomposition. These exemplary results illustrate how the present inventive method may be used in determining changes in a fluid property (e. g., interface reactivity changes) resulting from changes in formulation composition (e. g., changes in concentration of reactants due to thermal decomposition).

0100 While the time constants associated with the fluid itself and the reactivity at the metal surface are readily visible at the respective ends of the spectra (FIGURES 8 and 9), a third time constant is also detected after one hour of exposure. Although not prominent visually in the spectra, it is plausible that this parameter is derived from the presence/formation of an adsorbed film. FIGURES 14,15 and 16 illustrate the IS parameters relating to this fluid property as a function of exposure time.

0101 The time dependence of the capacitance Cfilm relating to the third time constant is shown in a plot 1400 of FIGURE 14. In contrast, the resistance Rflm behaves in an inverse fashion, as illustrated in a plot 1500 of FIGURE 15, suggesting that film geometry is undergoing a pronounced change during the first three hours of exposure. This appears to be indicative of film growth over a narrow area of coverage, followed by an increase in total coverage area as the generation of more species occurs. This progression would yield a film capacitance that is low initially (small area), and a resistance that is high also due to the reduced area. Broader area coverage over time would cause both values to increase and decrease, respectively, consistent with the observed behavior. Further, the film time constant Tfilm values shown in a plot 1600 of FIGURE 16 indicate that if this response is derived from a ZDDP-based film, it is not geometry alone that is changing. As the product of R and C yield a value that is geometry independent, the finite slope that is noted suggests that either the film resistivity and/or the polarizability also vary as exposure time increases. These exemplary

results illustrate how the present inventive method may be used in determining changes in another fluid property (ie., surface film formation) resulting from changes in formulation composition due to thermal decomposition.

Example 4: Hot Oil Oxidation Test--Data Analysis Using Differential Impedance Parameter 0102 In yet another embodiment of the present inventive concept, this example teaches novel IS parameters, and alternative methods for using data analyses in determining fluid and formulation properties. In particular, the present example teaches novel methods for determining changes in fluid properties during a Hot Oil Oxidation Test (HOOT).

0103 The HOOT is a high-temperature (e. g., 160 degrees Celsius in this example) oxidation test that does not introduce combustion gasses into the fluids. The HOOT experiments described herein are performed in glass containers, thus the fluids are not exposed to metal during the oxidation process. There is also no mechanical fluid degradation such as may cause wear or mechanical shear. Fluid samples are removed from the oxidation vessel at periodic intervals and examined using both Fourier Transform Infra-Red (FTIR) spectroscopy and IS, to evaluate changes resulting from HOOT processing. HOOT oxidation is limited to the free radical, oxygen-sourced, oxidation pathways and thermal decompositions. The free radical pathways include production of alcohols, ketones, ester/ lactones, and carboxylic acids. These oxidation mechanisms attack the base oil as a major pathway. They also attack the hydrocarbon chains of all of the additives. The presence of antioxidants gives additional decomposition pathways for the free radicals, and thus slows down the hydrocarbon oxidation.

0104 Two formulations are employed in this example. One formulation consists of a base oil plus a 3 % concentration by weight of sulfonate detergent, and a second formulation consists of a base oil plus the following concentrations of additives by weight: 3% detergent, 1% ZDDP, and 6% dispersant. In these formulations, the base oil is a mildly hydrofinished 150 solvent neutral, hydrocarbon base stock. The ZDDP is a commercially available secondary ZDDP. The detergent is a commercially available highly overbased calcium alkylbenzene sulfonate. The dispersant is a commercially available succinimide type dispersant. These formulations are referred to herein as the detergent-only formulation and the ternary formulation, respectively. Both formulations are"partial"formulations in that

they do not include the full range of additives typically found in final product lubricants.

FIGURE 17 illustrates the IS data for the formulations prior to the HOOT. These data indicate that the combined presence of all three additives, and interaction among these species, can be observed using broad spectrum IS measurements.

0105 FIGURE 18 shows a plot of FTIR peak area calculations over the range of 1833- 1640 cm~1 (obtained by methods well known to persons skilled in the arts of FTIR measurements and analysis) versus HOOT process time, for the two formulations. The FTIR peak area calculation for 1833-1640 cm~1 is a well known indication or metric of the oxidation process (also referred to herein by the equivalent terms"FTIR oxidation", "total oxidation", and"total carbonyl") in a lubricant. As noted above, the HOOT subjects the formulations to a temperature of 160 degrees Celsius, and samples are removed from the HOOT container at intervals for FTIR and IS measurements, performed at 80 degrees Celsius. The 15 sample numbers, 0 to 14, shown in the FIGURE 18 span an interval range of 145 hours. The sample numbers 0 through 14 correspond to the following sample times: 0 hours, 1 hour, 2 hours, 3 hours, 6 hours, 8.5 hours, 18 hours, 24 hours, 31.5 hours, 44 hours, 55.5 hours, 72.5 hours, 96 hours, 120.5 hours, and 145 hours, respectively. The data of FIGURE 18 reveal that significantly lower total oxidation results for the ternary formulation than for the detergent- only formulation. Due to the aromatic portion of this additive, and its sulfonate polar head group, the detergent additive is a pro-oxidant. As shown in FIGURE 19, these total oxidation values correlate well with the increase in Total Acid Number per ASTM D664, which is an alternative method, well known to persons skilled in the arts of petroleum engineering, of measuring the degree of lubricant total oxidation. The data points shown in the FIGURE 19 correspond, from left to right, to the sample numbers 1,3, 7,10, and 14, respectively.

0106 In addition to oxidation products, the presence of a ZDDP concentration in the fluid is characterized via FTIR peak area calculations at 645 cm (indicative of P=S bond concentration) and 970 cm~1 (indicative of P-O-C bond concentration), as illustrated in FIGURE 20. During decomposition of this additive, non-ionic intermediates are produced, which further decompose into ionic products. The final decomposition product is zinc phosphate or pyro-phosphate, which has poor solubility in the oil matrix. As the free radicals are formed, the ZDDP quenches them, and prevents the hydrocarbon oxidation. The calcium

carbonate forms salts with the oxidized acids, yielding the carbonate concentration depletion (via FTIR data at 860 cm') as a function of ongoing fluid sampling, as shown in FIGURE 21.

0107 FIGURE 22 shows exemplary IS data at 80 degrees Celsius obtained for the ternary formulation, as measured after cumulative HOOT exposure intervals for the sample numbers 0, 1, 3,4, 6 and 14. These spectra exhibit a bidirectional behavior (i. e. , the magnitude of the impedances first increase, and then subsequently decrease) as a function of HOOT time.

0108 FIGURE 23 illustrates an IS parameter that may be correlated to the chemical changes in the fluid properties described above in reference to FIGURES 18,19, 20 and 21.

As illustrated in FIGURE 23, an interface impedance difference (termed"dZmaglow") may be calculated for two frequencies that are selected to reduce the influence of bulk impedance changes on the data analyses. For the present example, dZmaglow may be defined in accordance with the following equation: 0109 dZmaglow = SQRT [(Zresistance @ 100mHz - Zresistance @ 10Hz)2 + (Zreactance @ 100mHz - Zreactance @ 10Hz)2].

0110 For these exemplary isothermal data, the Nyquist minimum occurs close to 10 Hz.

For a more general case, the definition of dZmaglow may employ selected frequencies wherein the higher frequency is close to the Nyquist minimum, and the lower frequency may be selected to be at least a factor of 10 below the higher frequency. The IS difference parameter dZmaglow is an interfacial impedance parameter that is indicative of the relative interface reactivity of electro-active species at the metal electrode surface (ie., an exemplary fluid property), and provides a novel and advantageous means for observing related changes in formulation properties such as total oxidation products, ZDDP concentration, carbonate depletion, etc. As will be shown in TABLES 4 and 5 below, other exemplary IS difference parameters (e. g., dZmaghigh, dZmagmid, and dZmagverylow) may also be defined and implemented. In FIGURE 24, plots of the dZmaglow parameter versus HOOT sample number for the detergent-only formulation and the ternary formulation are shown. The data exhibit bi-directional impedance changes that differ markedly between the different fluids.

These data are indicative of the different chemical processes that occur in the two formulations when exposed to the HOOT process, and illustrate how the IS difference

parameter dZmaglow can be used to distinguish between the chemical properties of the two formulations.

0111 In the early stage of HOOT processing of the ternary formulation, the IS difference parameter dZmaglow increases as ZDDP decomposition (both P=S and P-O-C bonds) is detected by FTIR. This correlation is shown in FIGURES 25 and 26, and occurs prior to substantial increases in total oxidation as also determined by FTIR. The data points shown in the FIGURES 25 and 26 correspond, from left to right, to sample numbers 0,1, 2,3, 4,5 and 6, respectively. It is observed that the IS difference parameter dZmaglow subsequently decreases as C03 concentration depletion ensues, as shown in FIGURE 27, coincident with the onset of oxidation. After the C03 is depleted at later HOOT process times, the decrease in dZmaglow indicates increased interface reactivity that is related to oxidation products, as seen in FIGURE 28 by the decrease in dZmaglow occurring between samples 9 and 14 indicated by the callouts. The data points shown in the FIGURES 27 and 28 correspond, from left to right, to sample numbers 0 to 14, respectively. For FIGURE 27, the data points for sample numbers 13 and 14 are coincident and indistinguishable.

0112 The teachings according to this example show the use of an IS difference parameter (e. g., dZmaglow) to determine exemplary fluid properties (i. e. , total oxidation, P=S bond concentration, P-O-C bond concentration, ZDDP concentration, C03 concentration, etc.) relating to changes in formulation composition changes (i. e., additive depletion and the generation of oxidation products in a formulation) during HOOT processing.

Example 5: Hot Oil Oxidation Test--Data Analysis Using Statistical Techniques 0113 In the description below, another embodiment of the present inventive concept is set forth, wherein a plurality of IS parameters are analyzed in combination using statistical techniques. As noted above, exemplary statistical tedmiques include, without limitation, Principal Component Analysis (PCA), Multivariate Least Squares Regression (MLR), Principal Component Regression (PCR), Group Method for Data Handling (GMDH), Pattern Recognition analysis, Cluster analysis, and Neural Net analysis. A description of these techniques in reference to IS data analysis is disclosed in the co-pending U. S. Patent Application, Application Number 10/723,624, filed November 26,2003, titled"FLUID CONDITION MONITORING USING BROAD SPECTRUM IMPEDANCE

SPECTROSCOPY" (Eaton Ref. No. 03-ASD-255 (SR) ). This above-incorporated application is set forth in full in Appendix A of the present application.

0114 To apply the GMDH technique, a more extensive list of possible IS parameters is developed to include both raw data extractions and derived calculations. These exemplary IS parameters are summarized in a TABLE 4 shown below.

TABLE 4. Exemplary IS Parameters for Statistical Analysis zreaUOkHz zimaglOkHz zreallOOHz zimaglOOHz zreal 1 OHz zimaglOHz zreallOOmHz zimaglOOmHz zreallOmHz zimaglOmHz invzreal 1 OkHz= 1/zreal 1 OkHz ; invzimaglOkHz=llzimaglOkHz ; invzrea ! 100Hz==l/zreallOOHz ; invzimaglOOHz=l/zimaglOOHz ; invzreal 1 OHz= 1/zreal 1 OHz ; invzimaglOHz=l/zimaglOHz ; invzreallOOmHz=l/zreallOOmHz ; invzimagl OOmHz=l/zimagl OOmHz ; invzreal 1 OmHz= 1/zreal 1 OmHz ; invzimaglOmHz=l/zimaglOmHz ; TANDhigh= (zreallOHz/zimaglOHz) ; TANDlow= (zreal 100mHz/zimag l OOrnHz) ; TANDvlow= (zreallOmHz/zimaglOmHz) ; dZmaghigh= (((zreall OOmHz-zreallOOHz) ^2) + ( OOmHz-zimaglOOHz) ^2)) ^. 5 ; dZmaglow= ( ( (zreaMOOmHz-zreaMOHz)"2) + ( (zimaglOOmHz-zimaglOHz)"2))". 5 ; dZmagmid= ( ( (zrea110mHz-zrea110Hz) ^2.) + ( (zimaglOmHz-zimaglOHz) ^2)) ^. 5 ; dZmagverylow= (((zreal lOmHz-zreall OOmHz) ^2) + ( OmHz-zimagl OOmHz) ^2)) ^. 5 ; Anglel=Re [ArcSin [-l* (zimaglOOmHz-zimaglOOHz)/dZmaghigh] *180/Pi] ; Angle2=Re [ArcSin [-1 * (zimagl 0mHz-zimaglOHz)/dZmaglow] * 180/Pi] ; Angle3=Re [ArcSin [-1 * (zimaglOmHz-zimagl00mHz)/dZmagverylow] * 180/Pi

0115 The selection of the IS parameters shown in the TABLE 4 will be readily understood by persons skilled in the arts of computer science and statistical analysis. For example, zreallOkHz, zreallOOHz, zreallOHz, zreallOOmHz, zreallOmHz, represent the resistive parts of the complex impedance Z evaluated at 10 kHz, 100 Hz, 100 mHz, and 10 mHz, respectively. Similarly, zimaglOkHz, zimaglOOHz, zimaglOHz, zimaglOOmHz, zimaglOmHz, represent the reactive part of the complex impedance Z evaluated at 10 kHz, 100 Hz, 100 mHz, and 10 mHz, respectively. These evaluation frequencies are appropriate for the exemplary data disclosed herein. Other IS parameters are defined by the equations as shown in the TABLE 4.

0116 Other evaluation frequencies may also be used for the IS parameters of TABLE 4, in accordance the present inventive concept, as persons skilled in the arts of data analysis will readily understand from the teachings herein. Expressed in terms of generalized evaluation frequencies fm, fln and 2nr the IS parameters listed in TABLE 4 may be represented in a generalized form as follows: zreal (fm), zimag (fm), invzreal (fm), invzimag (fm), TAND (fm), dZmag (fln, f2n), and ANGLE (fln, fzn) ; where m and n are positive integers indexing the distinct frequencies for which specific instances of each parameter are evaluated; where the generalized IS parameter invzreal (fm) = 1/zreal (fm) ; where the generalized IS parameter invzimag (fm) = 1/zimag (fm) ; where the generalized IS parameter TAND (fm) = zreal (fm)/zimag (fm), represents one or more IS parameters as exemplified by TANDhigh, TANDlow, and TANDvlow ; where the generalized IS parameter dZmag (f1n, f2n) = {[zreal(f1n)-zreal(f2n)]2 + [zimag (fln)- zimag (f2n)]2}1/2, represents one or more IS parameters as exemplified by dZmaghigh, dZmaglow, dZmagmid, and dZmagverylow ; and where the generalized IS parameter ANGLE (fln, f2n) = Re {ArcSin [-l* (zimag (fln)- zimag (f2n))/dZmag (fln, f2n)] *180/#}, represents one or more IS parameters as exemplified by Anglel, Angle2 and Angle3. The generalized frequencies fm, fln and f2n are chosen from within the frequency range of the IS measurements, and fln f2. Typically, at least one value for fln or f2n may be selected to correspond to the Nyquist minimum, as exemplified by dZmaglow. The IS parameters exemplified in TABLE 4, and the generalized forms of these IS parameters defined above, are equivalently referred to herein as"statistical IS parameters." 0117 In addition to analyzing data for the ternary formulation described above, GMDH calculations are also performed for data obtained for two sets of HOOT processed fluid samples containing additional components. These new fluids consists of the ternary

combination described above plus: a) 1% phenate detergent; and, b) 1% phenate detergent and 1% viscosity index improver. The data analyses are performed using the GMDH technique as found in commercially available software (e. g.,"KnowledgeMiner 5. 0" cited hereinabove and in Appendix A). The GMDH correlation equations for selected IS parameters (as shown in TABLE 4) and for FTIR oxidation data (i. e., total oxidation metric, the FTIRpeak area calculations over the range of 1833-1640 cm~l, as described above in reference to FIGURE 18) are shown in the TABLE 5 below. TABLE 5. Exemplary GMDH Equations Relating to Selected Statistical IS Parameters and FTIR Oxidation Data X2-dZmaghigh X33-invzimaglOHz X10-TANDhigh X31-invzimaglOkHz X32-invzimaglOOHz X13-zreaIlOkHz X28-invzreallOHz X18-zimaglOkHz X26-invzreallOkHz X12-TANDvlow X29-invzreal I OOmHz al2=-2. 567e+7X26^2-7. 406e+10X28^2 + 4. 173e+10X26X28 +. 402e+4X26-2. 763e+0 all=-1. 972e+10X31^2 + 8. 492e-5Xi3 + 1. 288e+1X13X31-1. 334e+6X31-9. 714e+0 z22= + 3. 477e-lall^2 + 2. 432e-lal2^2-1. 164e-lal2-5. 698e-lallal2 + 1. 039e+0all bl2= + 1. 219e-9X18^2 + 6. 031e+7X32^2 + 4. 736e+4X32 + 2. 370e-lX18X32 + 3. 374e-4X18 + 2. 268e+1 bll=-7. 586e-4X10^2+ 7. 458e-1lX18^2-1. 957e-6XlOX18-2. 765e-lX10-2. 036e+0 z21=-5. 163e-2bil^2 + 5. 446e-2bl2^2 + 3. 951e-ibii + 6. 158e-lbl2 z32= + 1. 152e-lz21z22-1. 140e-lz21^2 + 5. 493e-lz21 + 4. 538e-lz22 z31=-9. 137e-llX18^2-2. 774e+11X29^2 + 4. 004e+6X29 + 3. 547e+1X18X29 + 1. 063e+0 z42=-8. 974e-2z31^2-3. 637e-2z32^2 + 9. 957e-lz32 + 1. 229e-lz31z32 z41= + 1. 714e-9Xt3^2-1. 781e+11X28^2 + 2. 945e+6X28-2. 373e+1X13X28-1. 817e-4X13 + 4. 274e-1 z52= + 3. 067e-2z41^2-2. 644e-2z42^2-1. 507e-lz41 + 1. 130e+0z42 z51= + 8. 399e-2X10 + 2. 074e+0 z62= + 1. 045e+0z52 + 1. 469e-2z51z52 + 4. 828e-2z51 z61= + 4. 482e-6X12^2 + 9. 482e-lOX18^2 + 2. 282e-4X18-1. 149e-6X12X18-1. 452e-lX12 + 1309e+1 z72=-1. 300e-lz61^2-1. 177e-tz62^2 + 9. 555e-lz62 + 2. 547e-lz61z62 + 4. 900e-2z61 c12=-4. 683e+7X33^2-9. 254e+3X33-6. 113e+8X26X33 + 4. 346e+3X26-1. 437e+0 cll= + 4. 025e-lOX13^2 + 5. 941e-10X18^2 + 2. 238e-4X18-1. 723e-9X13Xi8-2. 438e-4X13 + 1. 845e+1 z71=-1. 462e-lcilcl2 + 1. 322e-lcll^2 + 1. 007e+Ocil z82= + 6. 888e-lz71^2 + 6. 488e-lz72^2 + 1. 366e+0z72-1. 341e+0z71z72-3. 724e-lz71 z81= + 1. 714e-9X13^2-1. 781e+11X28^2 + 2. 945e+6X28-2. 373e+1X13X28-1. 817e-4X13 + 4. 274e-1 z92=-2. 035e-2z82^2-8. 351e-2z81 + 2. 301e-2z81z82 + 1. 072e+0z82 dl2= + 1. 439e+7X32^2 + 1. 842e+4X32-7. 550e+5X31-6. 500e+0 dll= + 8. 399e-2X10 + 2. 074e+0 z91=-5. 261e-2dll^2-4. 385e-2dl2^2 + 9. 558e-ldl2-1. 696e-ldlldl2 z102=-9. 407e-2z92^2-3. 684e-lz91 + 9. 376e-2z91z92 + 1. 366e+0z92 zl01=-3. 695e+7X33^2 + 2. 239e-2X2X33-2. 557e+4X33-4. 591e-1 Oxidation (FTIR) =-9. 297e-2zlOl^2 + 8. 430e-2zlO2zlO2 + 5. 460e-lzlOl + 6. 145e+Ozl02 + 5. 019e+0

0118 A plot of actual vs. predicted or estimated oxidation values determined from the selected IS parameters is shown in FIGURE 29. The figure data includes data for both the ternary formulation and for the two new fluids with additional components. The oxidation values (also referred to equivalently herein as the oxidation metric, total oxidation metric or total carbonyl metric) are determined from the FTIR peak area calculations over the range of 1833-1640 cm~1. In this example, the fluid property that can be determined from the exemplary IS parameters is the total oxidation, which may be related to formulation properties such as the selected additives and additive concentrations, as well as to the HOOT processing time.

Example 6: Hot Oil Oxidation Test--Data Analysis Using Equivalent Circuit Modeling Combined with Statistical Techniques 0119 In yet another embodiment of the inventive concept, the IS data may be analyzed using a combination of equivalent circuit modeling and statistical techniques. In this aspect of the inventive concept, the frequency-dependent IS data are expressed in terms of frequency- independent circuit parameter values as exemplified in FIGURE 30. This is accomplished using a Complex Non-Linear Least Squares (CNLLS) technique as described in the above section entitled"Exemplary Data Analysis Using Equivalent Circuit Modeling". Upon reduction, this model (FIGURE 30) yields values for the electrically-active properties of the fluid, both static (bulk) and dynamic (interfacial reactivity). These IS parameters are derived for the ternary formulation (base oil plus ZDDP plus dispersant plus sulfonate detergent) and for the two new fluids with additional components (the ternary formulation plus: a) 1% phenate detergent ; and, b) 1% phenate detergent and 1% viscosity index improver), and correlated against FTIR oxidation values using the GMDH technique as described above.

0120 For this aspect of the inventive concept, TABLE 6 below shows the GMDH correlation equations using the equivalent circuit IS parameters of FIGURE 30, and FTIR oxidation values. Persons skilled in the arts of data analysis will readily understand that any of the following equivalent circuit IS parameters could likewise be used in accordance with the example illustrated by TABLE 6: a bulk fluid resistance Rfluid, a bulk fluid capacitance Cfluid, a bulk fluid time constant T fluid, an interface capacitance Cdl, a charge transfer resistance Rct, an interface time constant T i, an interface resistance Ri, a Constant Phase Element Q i, a Constant Phase Element exponent n, a film time constant c film, a film resistance R film, and a film capacitance C film. TABLE 6. Exemplary GMDH Equations Relating Selected Equivalent Circuit IS Parameters and FTIR Oxidation Data X7 =QI X9 = R8 Xll R7 X3 =C1 X12 = Q2 X4 = R4 X6 = R6 X2 = R2 X10 = C5 zll= +2. 10922*10 13*X3X12-1. 10256*10 18*X3X3-2. 43898*10 6*X12A2-7. 53010*10-I zl2 = + 4. 24133*10 3*X12 + 1. 92502*10-2*X4X12-3. 27766 10 6X12 2-6. 99263 10-I z21a= +2. 77658*10^3*X12-8. 22530*10A-I z21b= = + 3. 81709*10AS*X7-1. 27437 z22a= +1. 6392*zll-6. 30134*101*zl2+7. 50432*101*zll*zl2-5. 77612*10-l*zll2 za32 = + 9. 32239 10-I z22-3. 62732 10-I z21a*z22 + 2. 13920*10-1*z22 2 zc31=-2. 68192*10^-7*X6+6. 79587*10^5*X7-6. 55073*10^10*X7^2-9. 40777*10^-1 zb32 = + 9. 10268 10-I z22a-1. 74415 10-I z21b 2 + 1. 36023 10-I z22a 2 za42=-1. 25133*10A-l*z21a+ 1. 08294*za32 zb42 +3. 18639*10-l*zc31+S. 45805*10-l*zb32+3. 05694*101*ze31*zb32-8. 15632*102*zb322 za51= +4. 55819*10^-12*Xll+5. 16039*10^3*X12-2. 84055*10^-24*Xll^2-3. $0058*10^6*X12^2-7. 74618*10^-1 za52 = + 3. 51598* 10^-1 *za41 + 8. 59137* 10^-1 *za42 + 3. 57450* 10^-1 *za41 *za42-1. 25593 * 10^-1 *za42^2 zb51 =+ 1. 02t89*zb22-9. 31893*102*zb21*zb22 zb52 = + 3. 70104 10-I za42 + 6. 40380* 10-I *zb42 za61 =-3. 52579*10A-l*zaSI + 1. 29288*za52-1. 37206*10A-l*zaSIA2 zb62=-1. 14779*zb51 +2. 16861*zb52+9. 21526*zb51*zb52-4. 75962*zb512-4. 48006*zb522 z72= +4. 95926*10**l*za61 +5. 13511*10**l*zb62 z71 = + 5. 74734* 10** 1 *zb32 + 432448* 10** 1 *za32 z81 = +3. 79240*10**12X11-2. 22245*10**1 z82=-7. 54167*10**1*z71 + 1. 72484**z72 z91 = + 5. 51207*10**6X9-7. 51330*10**1 z92= +9. 93505* ! 0** ! *z82+ !. 47721*10**2*zS ! *z81 z101 =+3. 81709*10^5*X7-1. 27437 zl02 =-5. 61370 10-2 z91 + 9. 88742*10-I *z92-5. 29448*10-2*z91*z92 Oxidation (FTIR) = +3. 42695*10*zl02+1. 27275*10"2*zl012-1. 08217*102*zl022+3. 12363*10*3

0121 A plot of actual vs. predicted, or estimated, oxidation values determined from the selected equivalent circuit IS parameters is shown in FIGURE 31. The figure data includes data for both the ternary formulation and for the two new fluids with additional components.

The actual oxidation values are determined from the FTIR peak area calculations over the range of 1833-1640 cm~1 as previously described. In this example, the fluid property that can be determined from the exemplary IS parameters is the total oxidation metric, which may be related to formulation properties such as the selected additives and additive concentrations, as well as to the HOOT processing time.

- 37-

Formulation Design Improvement 0122 The examples set forth above demonstrate how the practice of the present invention provides extensive information about fluid properties. Persons skilled in the arts of lubricating fluids shall appreciate that these methods may be implemented in many ways for the purpose of improving the design of fluid formulations and developing new formulations.

0123 In one embodiment of the inventive concept, the methods described above can be used to design fluid formulations having improved lifetimes under oxidative stress. This may be accomplished using HOOT processing on a plurality of fluid formulations having differing additives, additive combinations, and additive concentrations. Using the IS measurements and data analysis methods, the formulations having the highest resistance to oxidative stress may be determined and related to the formulation properties. This information can then be used to design improved formulations.

0124 In yet another embodiment of the present inventive concept, the methods set forth above can be used to improve the design of lubricant detergents. Lubricant detergents comprise additives that contain a soap portion incorporated with a metal carbonate in an oil matrix. It is observed from IS measurements and data analyses that there is a relationship between the interfacial impedance and the amount of carbonate present. This is observed in experiments where the detergent concentration is varied in a new oil (e. g, see TABLE 2) as well as when the carbonate is reacted out in an oxidation experiment (e. g., see FIGURE 27).

As the amount of carbonate is depleted, a corresponding decrease in the interfacial impedance is observed. These data may be interpreted to indicate that the anti-wear properties of the overbased sulfonate-detergent are related to the interfacial behavior of the carbonate incorporated in the additive, and not necessarily the soap portion.

0125 In an embodiment of the inventive concept, formulations may be designed wherein the properties influenced by the sulfonate-detergent additives are selectively modified. For example, the additives may be selectively modified to improve the anti-wear behavior, or to adjust interface reactivity effects. Adjustments to the carbonate incorporation methods may be used to make such modifications and improvements. These adjustments may include changing the soap-to-carbonate ratio, the detergent overbase ratio, the carbonate-metal counter ion, or the hydrocarbon size of the soap. For each additive adjustment, the inventive

methods taught herein may be used to determine the desired effect of the adjustment, for example, by determining the corresponding change in the interface reactivity fluid property.

An improved additive and formulation may thereby be developed in accordance with the inventive process. Many other implementations of the present inventive teachings used to improve the design of lubricant formulations will be readily apparent to persons skilled in the arts of lubricating fluid development and production.

Exemplary Formulations 0126 Although the practice of the present inventive concept is not limited to lubricating fluids, many advantageous embodiments may include lubricating fluid formulations. A lubricating fluid formulation may consist of a base oil (BO) incorporating one or more lubricant additives. Exemplary BO types may include, without limitation, the following categories: mineral base and synthetic base. Exemplary lubricant additive types may include, without limitation, the following lubricant additive categories: viscosity modifiers, pour point depressants, stabilizers, seal swell agents, anti-static additives, antioxidants, metal deactivators, anti-foam agents, detergents, dispersants, anti-wear additives, and corrosion inhibitors.

0127 Other exemplary fluids that may be used with embodiments of the present inventive concept may include, without limitation, lubricant additive packages, fuel treatment additives and top treatments. These exemplary fluids may include a dilutant (such as kerosene or other fuel), and at least one additive, wherein a typical additive may include, without limitation, the following categories: viscosity modifiers, pour point depressants, stabilizers, seal swell agents, anti-static additives, antioxidants, metal deactivators, anti-foam agents, detergents, dispersants, anti-wear additives, and corrosion inhibitors. A top treatment may include an additive (such a DI, VII, etc.) and a dilutant or BO.

Exemplary Fluid Properties 0128 Many examples of fluid properties have been presented hereinabove. In reference to the present inventive concept, fluid properties may include any physicochemical metric indicative of the physical or chemical properties of the fluids. Examples of such physicochemical metrics include, without limitation, the following: an interface reaction rate; a bulk resistivity; an ion concentration ; an ion mobility; a surface film formation; a total oxidation level; a P-O-C bond concentration; a P=S bond concentration; a ZDDP concentration ; and a carbonate concentration. Persons skilled in the arts of physics, chemistry and physical chemistry will understand that these are just a few of the possible fluid properties encompassed by the present teachings.

Exemplary Method of Operation 0129 In accordance with the present invention, an exemplary method for evaluating and improving a fluid formulation is described. The STEPS 3202 to 3210 described below are illustrated in the flowchart of the inventive method 3200 shown inFIGURE 32.

0130 At a STEP 3202, a set of fluids representative of a fluid formulation may be selected. In one exemplary embodiment, each fluid in the set may comprise a base fluid incorporating one or more additives at one or more concentration levels. In another exemplary embodiment, each fluid may comprise a sample from a HOOT process series. The method proceeds to a STEP 3204.

0131 At the STEP 3204, IS measurements are performed on each of the fluids to provide IS data. The IS data may include at least three points, and typically include tens or hundreds of points. More than a few hundred points are not usually required. In general, the IS data include frequencies both above and below 1 Hz, over a range of frequencies sufficient to define IS parameters associated with both the bulk and interface of the fluids. The method proceeds to a STEP 3206.

0132 At the STEP 3206, the IS data are analyzed using statistical techniques, equivalent circuit modeling techniques, or a combination thereof. The data analysis provides IS parameters indicative of at least one fluid property for the fluids. The method proceeds to a STEP 3208.

0133 At the STEP 3208, a correlation is determined between selected IS parameters and selected properties of the fluids. As one example, a correlation may be determined between the IS parameter dZmaglow and the carbonate concentrations in the fluids. As another example, a plurality of IS parameters may be correlated to total oxidation levels or values for the fluids. As shown by the teachings herein, a plurality IS parameters may be obtained, and these parameters may be correlated singly or in combination to a plurality of fluid properties.

The method then proceeds to a STEP 3210.

0134 At the STEP 3210, a new fluid formulation is developed, responsive to the correlation between the selected IS parameters and the selected properties of the fluids.

Examples for producing a new, or modified, fluid formulation include, without limitation, the following: adding at least one additive; removing at least one additive; modifying the concentration of at least one additive; modifying a soap-to-carbonate ratio of a detergent additive; modifying a detergent overbase ratio of a detergent additive; modifying a carbonate- metal counter ion of a detergent additive; and modifying a hydrocarbon size of a detergent additive. Many other examples in accordance with the present teachings will be apparent to persons skilled in the arts of developing industrial fluid formulations.

0135 The above description is intended to be illustrative and not restrictive. Many embodiments and applications other than the examples provided would be apparent to those of skill in the art upon reading the above description. The scope of the invention should be determined, not with reference to the above description, but should instead be determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled. It is anticipated and intended that future developments will occur in the field of determining fluid properties and IS parameters using impedance spectroscopy and that the disclosed systems and methods will be incorporated into such future embodiments.

Accordingly, it will be understood that the invention is capable of modification and variation and is limited only by the following claims. v v 0136 System Overview 0137 Figure 33 provides an overview of an exemplary system 100* by which the condition of a substance such as a fluid is determined in real or near real time. Although the invention is described herein with reference to system 100*, those skilled in the art will appreciate that other configurations and other components could support the claimed systems and methods so long as impedance spectroscopy is used to generate spectra for resistive and reactive impedance, which spectra could be used to perform calculations as described below to predict fluid conditions in situ and generally while the fluid is being actively used.

0138 Electrode mechanism 102* is disposed in fluid 104* contained in vessel 106*.

The electrode mechanism may be any of those that are known in the art of impedance spectroscopy, including those described in aforementioned U. S. Patent Nos. 6,433, 560, 6,380, 746,6, 377,052, and 6,278, 281, and U. S. Published Application 2003/0141882 of Zou et al. , all of which are hereby incorporated herein by reference. Present practice is to use the cylindrical probe mechanism disclosed in U. S. Published Application 2003/0141882 of Zou et al. In some embodiments of the present invention fluid 104* is an engine lubricant and vessel 106* is an engine crankcase.

0139 Power supply 114* powers oscillator 112*, a computing system or device such as microcomputer 116*, and excitation driver 108*. Excitation driver 108* receives an input from an oscillator 112*. In one embodiment, oscillator 112* provides voltage in a range from approximately 75 kilohertz and 0.0075 hertz. However, those skilled in the art will appreciate that the invention could be practiced using voltages at frequencies higher than 75 kilohertz and/or lower than 0.0075 hertz. In one embodiment, oscillator 112* sequentially provides voltages at a specified number of different frequencies to excitation driver 108*. Excitation driver 108*, upon receiving input from oscillator 112*, excites electrode mechanism 102*, and current sensor 118* provides input to microcomputer 116*. Temperature sensor 110* measures the temperature of fluid 104*, and provides input to microcomputer 116*.

0140 Microcomputer 116* calculates and stores in memory values for resistive and reactive impedance, not shown in Figure 33, based on the input from current sensor 118*.

Resistive and reactive impedance are sometimes referred to as real and imaginary impedance, respectively. Fluid temperature from temperature sensor 110* is input to computer 116*.

Baseline values 122*, comprising previously predicted values of fluid properties, are stored in information library 120*. Each of the baseline values 122* has as attributes the identity of the one fluid property with which it is associated, a fluid temperature, and a value representing the expended useful life of the fluid. Fluid temperature is usually expressed in degrees Celsius, and the value representing expended useful life may be expressed in hours. Baseline values 122* are determined in a laboratory external to system 100*, and comprise expected fluid property values for a fluid of a given age at a given temperature. In some embodiments baseline values 122* comprise expected fluid property values for a fluid that has been used in an engine for a given number of hours, and is currently at a given temperature.

0141 Microcomputer 116* uses information library 120* to determine the condition of fluid 104* by using the temperature input from temperature sensor 110* and the calculated values for resistive and reactive impedance to retrieve the correct prediction equations 124* from information library 120*. Prediction equations 124* are described in detail below.

Microcomputer 116* then uses prediction equations 124* to calculate fluid property values 128*. Microcomputer 116* then obtains one or more property predictions 126* of values relating to the condition of fluid 104* by comparing at least one predicted flud property value 128* to at least one of baseline values 122*. Property predictions 126* will be stored in the memory of microcomputer 116*, and may optionally be stored in information library 120*, although this optional configuration is not depicted in Figure 3 3.

0142 In some embodiments, information library 120* is a component of microcomputer 116*. In some embodiments, microcomputer 116* is further capable of providing as output an end of life (EOL) measurement, a remaining useful life (RUL) measurement, or both. In these embodiments, baseline values 122* are used in conjunction with the results from using the prediction equations 124*, that is, fluid property values 128*, to make a determination of whether fluid 104* is at or near the end of its useful life. Methods by which EOL and RUL determinations are made are discussed in more detail below.

0143 Figure 34 describes a system 200* that may be used for collecting data used to develop information library 120*, including baseline values 122*. Although the invention is described herein with reference to system 200*, those skilled in the art will appreciate that other configurations and other components could support the claimed systems and methods.

System 200* includes a sample fluid 204* at a predetermined, constant temperature contained in a vessel 206*. Electrode mechanism 202* is disposed in sample fluid 204*. In one embodiment, sample fluid 204* is an engine lubricant.

0144 Power supply 214* powers oscillator 212*, a computing system such as computer 216*, and excitation driver 208*. Excitation driver 208* receives a sequence of inputs from oscillator 212*. Oscillator 212* provides voltage at a specified number of frequencies in a range of frequencies. As currently practiced the invention uses frequencies in a range from approximately 75 kilohertz and 0.0075 hertz. However, those skilled in the art will appreciate that the invention could be practiced using voltages at frequencies higher than 75 kilohertz and/or lower than 0.0075 hertz. In one embodiment, the specified number of frequencies is 71. For each input from oscillator 212*, excitation driver 208* excites electrode mechanism 202*, and current sensor 218* provides input to computer 216*. Temperature sensor 210* measures the temperature of fluid 204*, and provides input to computer 216*. Computer 216* generates calculated values 228* for resistive and reactive impedance based on the input from current sensor 218*. Computer 216* stores the calculated values 228* in database 220* such that calculated values 228* are associated with the specified temperature and the frequency in the range of frequencies for which the calculated values for resistive and reactive impedance were obtained. In some embodiments, database 220* is a component of computer 216*.

0145 Computer 216* may be further configured to generate prediction equations 224* according to the method described in detail below. As part of this method, baseline values 222* are used to populate a result matrix, as described below. Prediction equations 224* may be stored in database 220*. At some point prior to use of system 100* for prediction of fluid conditions in real time, prediction equations 224* may be copied from database 220* into information library 120* in system 100*, wherein prediction equations 124* then comprise prediction equations 224*. Similarly, baseline values 222* may be copied to baseline values 122* 0146 Overview of the invention 0147 Figure 35 depicts a flowchart providing an overview of the method by which information library 120*, including prediction equations 124* and baseline values 122*, is developed and used to predict fluid conditions in real time. When reviewing Figure 35 it is helpful to bear in mind that the overall goal of the process being described is to develop a set of prediction equations 124*, each of which can be used to predict a fluid property. These prediction equations 124* will take the form: Yi, j = bO + blX1 + b2X2 +... + bfXf (2) 0148 where Yi, j represents the jth fluid property out of a given number of fluid properties being measured in the ith sample out of n fluid samples being used, each X1... Xf represents the value of an impedance reading at a particular frequency, and each bO... bf is what is called a loading coefficient. One goal of the inventive method is to solve for the loading coefficients using values for X1... Xf determined from a process using impedance spectroscopy, certain statistical techniques, and baseline values 122* for Yi, j that are known from prior laboratory testing. The developed equations, i. e. , prediction equations 124*, can then be used to determine a set of values Yi, j in real time, i. e. , the property predictions 126* described above with reference to Figure 33.

0149 Turning now to Figure 35, block 300* represents the process of assembling a matrix of spectral data relating to a set of n fluids. In one embodiment, each of the n fluids will represent fluid conditions for the fluid at n different ages, e. g. , number of hours of use.

The goal of the process represented by block 300* is to develop a matrix of data, hereinafter referred to as the spectral matrix, representing impedance spectra obtained using each of the n fluids, that can then be subjected to statistical analysis. The form of the spectral matrix is as follows: The dimensions of the spectral matrix will be n rows by p columns, where n is the number of fluid samples, as discussed above, and p is related to, and sometimes equal to, the number of frequencies for which calculations of impedance have been used. In one embodiment p is equal to one-hundred and forty-two, and represents two times the number of frequencies for which impedance calculations have been used. The process of creating the spectral matrix is described in detail below with reference to Figure 36.

0150 Block 302* represents the process of assembling a result matrix for the n fluids, i. e. , the same set of n fluids for which the spectral matrix was constructed in Block 300* above. As noted above baseline values 222* comprise data used in the result matrix. The data in the result matrix is achieved from traditional analytical laboratory procedures for measuring fluid properties. While by no means limited to the following, examples of the laboratory tests used to determine values for the result matrix include ASTM D-445 (40 or 100 degree vis), ASTM D-4739 (TBN), ASTM D-2869 (TBN, i. e. , Total Base Number), ASTM D-664 (TAN, i. e., Total Acid Number), ASTM D 5967 (per cent SOOT), and ASTM D-5185 (ICP Elementals). The form of the result matrix is as follows: The data in the result matrix represent fluid conditions with respect to f fluid properties in n sample fluids. Accordingly, the dimensions of the result matrix are n rows by f columns.

The result matrix is used to find the loading coefficients for the prediction equations 124* (2), as discussed in more detail below. The process of creating the result matrix is described in further detail below with reference to Figure 37.

0151 Returning to Figure 35, block 304* represents the optional process of applying pre-processing functions to the spectral matrix before subjecting the spectral matrix to statistical analysis. The pre-processing functions used in the present invention are all well known, and include but are not limited to mean-centering, taking the first or second derivative of the data, smoothing the data, sample averaging, and differencing. Differencing comprises taking the difference between two values related to a fluid sample, such as the difference between values for real and reactive impedance, or the difference between data values for a new fluid and a fluid that has been used, e. g. , a newengine lubricant and a lubricant that has been used in an engine operating for a number of hours.

0152 Block 306* represents the process of performing a Principal Component Analysis (PCA) on the spectral matrix. PCA is a technique for analyzing a set of data to determine underlying independent factors that influence the data. Applied in the context of the present invention, PCA provides the advantage of reducing the number of variables in the spectral matrix to a set of variables for which there are as few common variations in the data as possible. By using PCA to create a set of principal components that represent the major changes in the impedance spectra found in the spectral matrix, the present invention creates a simplified spectra that can be subjected to meaningful statistical analysis that would not be practical with the entire spectral matrix. The advantages of PCA can be seen by noting that in current practice of the present invention the number of columns in the spectral matrix is often reduced from one-hundred and forty-two to less than ten, allowing the creation of a set of meaningful data, wherein the size of the data set is such that it is practical to apply statistical techniques to the data. PCA in the context of the present invention is discussed in more detail below with reference to Figure 38.

0153 Block 308* represents the process of creating a reduced spectral matrix by redeveloping the spectral matrix with principal components selected from the output of the PCA. Specifically, as discussed below with reference to Figures 38 and 42, principal components are selected for the reduced spectral matrix if they show significant influence on the prediction of fluid properties. The selected principal components are then associated with the X values from the spectral matrix that correspond to them, and placed in the reduced spectral matrix. The process of creating a reduced spectral matrix is described in detail below with reference to Figure 38.

0154 Block 310* represents the process of performing a statistical analysis, using the result matrix and the reduced spectral matrix, to find loading coefficients that can be used in the prediction equations 124* (2) described above. Some embodiments of the present invention use Principal Component Regression (PCR) to determine the loading coefficients, while other embodiments use Multivariate Least Squares Regression (MLR) to determine the loading coefficients. PCR and MLR are both well known. Still other embodiments use well known nonlinear regression methods such as Group Methods of Data Handling. Additionally, other known methods of statistical analysis could be used to determine the loading coefficients. The use of statistical techniques in the context of the present invention is discussed in more detail below.

0155 Block 312* represents the process of using the prediction equations 124* (2) in system 100* of Figure 33 to obtain predicted property values 128*. Predicted property values 128* are in turn compared with baseline values 122* to obtain property predictions 126*.

This processing may be done by microcomputer 116*. In one embodiment, predicting fluid properties for a fluid in situ comprises determining the likely expended useful life of an engine lubricant by determining whether property predictions 126* either exceed or fail to meet predetermined threshold values, which may be done by microcomputer 116*. From this determination an estimate of Remaining Useful Life (RUL) of the lubricant can be made and output using microcomputer 116*. Similarly, if it is the case that, based on the determination of the likely age of the fluid, the fluid is near the end of its useful life, an End of Life (EOL) determination can be made and output from microcomputer 116*. The procedure for making an RUL or EOL estimation in the context of the present invention is discussed in more detail below.

0156 Creation of the Spectral Matrix 0157 Figure 36 provides a flowchart describing an approach for building a matrix of spectral data. In step 400*, a sample fluid 204* having a known expended useful life is provided at a specified temperature. In step 402*, the expended useful life in hours and the specified temperature are recorded in database 220*. It will be appreciated that recording the specified temperature is important inasmuch as many fluid measurements are a function at least in part of temperature. Similarly, many fluid measurements are a function at least in part of expended useful life. One of the objects of the present invention is to be able to determine a value for the expended useful life of a fluid when temperature is known.

0158 In step 404*, calculated values for real and reactive impedance 224* associated with predetermined properties of sample fluid 204* are recorded in database 220*. Baseline values 222* are recorded in database 220* such that they are associated with the specified temperature of sample fluid 204* as well as the known age in hours of sample fluid 204*.

The predetermined fluid properties associated with baseline values 222* may include bulk and/or interfacial properties of the sample fluid 204* comprising amounts of additives such as zinc dithiodialkylphosphates (ZDDPs) and contaminants such as soot as well as interfacial properties such as wear protection. Baseline values 222* will have been previously determined through one of a variety of testing means known to those skilled in the art, as discussed above.

0159 In step 406*, test parameters are selected, comprising a range of frequencies to be tested, a specified number of specific frequencies in the range of frequencies to be tested, and the specific frequencies that will be tested. Present practice is to select 71 frequencies in the range from 75 kilohertz to 0.0075 hertz. A listing, in hertz, of frequencies used in at least one instance of present practice of the invention is as follows: 75000,59574. 62,47321. 8, 37589.04, 29858.04, 23717.080000000000, 18839.15, 14964.47, 11886.7, 9441.941, 7500, 5957.462, 4732.18, 3758.904, 2985.804, 2371.708, 1883.915, 1496.447, 1188.67, 944.1941, 750,595. 7462,473. 218,375. 8904,298. 5804,237. 1708,188. 3915,149. 6447,118. 867, 94.41941, 75,59. 57462, 47. 3218,37. 58904,29. 85804,23. 71708, 18. 83915, 14.96447, 11.88670, 9.44194, 7.50000, 5.95746, 4.73218, 3.75890, 2.98580, 237171,1. 88391,1. 49645, 1.18867, 0.94419, 0.75000, 0.59575, 0.47322, 0.37589, 0.29858, 0.23717, 0.18839, 0.14964, 0.11887, 0.09442, 0.07500, 0.05957, 0.04732, 0.03759, 0.02986, 0.02372, 0.01884, 0.01496, 0.01189, 0.00944, and 0.0075. It should be understood that the invention is not limited to any particular frequencies, number of frequencies, or range of frequencies, and that the above list of frequencies is given for illustrative purposes only. The above-listed frequencies were chosen because they are somewhat evenly distributed across the impedance spectrum and have been found to yield good laboratory results.

0160 In step 408*, oscillator 220* inputs a sequence of voltages at each of the specific frequencies selected to excitation driver 208*, which sends current at each of the specific frequencies to electrode mechanism 202*. Current sensor 218* detects the amount of current at each of the specific frequencies from electrode mechanism 202*, and sends an amount of current as an input to computer 216*.

0161 In step 410*, computer 216* calculates resistive and reactive impedance values for each of the specific frequencies based on the current at each of the specific frequencies, and stores the calculated values 224* in database 220*.

0162 In step 412*, a Bode plot of the resistive impedance spectra is created. Bode plots are well known in the art. A Bode plot comprises a two-dimensional graph in which the x axis is comprised of the logarithms of the frequencies against which impedance is plotted, and the y axis is the value for impedance. In step 414*, a Bode plot of the reactive impedance spectra is created.

0163 Figure 39 shows Bode plots of real and reactive impedance spectra superimposed on the same graph E00. X axis 702* comprises values for the logarithms of frequencies in the impedance spectra. Y axis 704* comprises calculated values for impedance corresponding to the frequencies whose logarithms are represented in x axis 702*. Plots 706* comprise plots of the resistive impedance spectrum for different fluid samples. Plots 708* comprise plots of the reactive impedance spectrum for different fluid samples.

0164 Returning to Figure 36, in step 416* the values on the x axis of the Bode plot of the resistive impedance spectra are replaced with integers ranging from zero to a number one less than the specified number of frequencies that were selected in step 406*. In step 418*, the values on the x axis of the Bode plot of the reactive impedance spectra are replaced with integers ranging from the specified number of frequencies that were selected in step 406* to a number that is one less than two times the specified number of frequencies. For example, if the specified number of frequencies was seventy-one, values on the x-axis would range from 0 to one-hundred and forty-one. As discussed below, the number of columns p in the spectral matrix is the number identified in the present step that is two times the specified number of frequencies. Present practice is to include one-hundred and forty-two columns in the spectral matrix.

0165 In step 420*, the Bode plot of the resistive impedance spectra created in step 412* and modified in step 416* is combined with the Bode plot of the reactive impedance spectra created in step 414* and modified in step 418*. In one embodiment, the Bode plots of the resistive and reactive spectra are combined so that the plots are laid"head to tail", with the maximum plotted value on the x axis of the plot of the resistive impedance spectra one unit to the left of the minimum value on the x axis of the plot of the reactive impedance spectra. It should be understood that the order in which the Bode plots of resistive and reactive impedance are combined is not essential to the practice of the invention. For convenience, the approach and examples discussed herein place a Bode plot of resistive impedance to the left of a Bode plot of reactive impedance on a combined graph. This order could just as easily be reversed. In fact, points from the Bode plots of the resistive and reactive spectra could be placed on the combined plot in any order without making a difference to the results achieved by practicing the invention.

0166 Figure 40 shows an example of combined Bode plots on a graph 800*. Y axis 804*, which comprises the same scale as found on Y axis 704*, comprises calculated values for impedance corresponding to the frequencies whose logarithms are represented in x axis 702*. With respect to x axis 802*, however, the logarithms of the frequencies in the impedance spectra have been replaced by integers ranging from zero to one-hundred and forty-one. Resistive impedance spectra 806* lie above the points less than or equal to seventy. Reactive impedance spectra 808* lie above the points greater than or equal to seventy-one.

0167 Returning to Figure 36, in step 422* a test is done to determine whether there are any other data points to be added to the combined plot. It has been found that adding certain data points to the combined plot can increase the ability of the invention to predict fluid properties. Examples of data points that may be added to the combined plot are discussed below with respect to Figures 44 and 45.

0168 If there are other data points to be added to the combined plot, then they are added in step 424* by placing them either to the left of the minimum x value presently graphed on the combined plot, or to the right of the maximum x value currently graphed on the combined plot. If there are not any data points to be added to the combined plot, then the process continues to step 426*.

0169 In step 426*, a test is done to determine if there is another sample fluid 204* to be tested. If the answer is yes, the method returns to step 00*. If the answer is no, the method proceeds to step 428*. In general, there will be n iterations of the method, n being the number of fluid samples at different ages that are being tested.

0170 In step 428* a spectral matrix, taking the form described above, is constructed. As discussed above, the spectral matrix will have n rows and p columns. The number p generally represents the number of frequencies for which impedance readings are included in the matrix. However, as mentioned above, in the present invention, p is actually twice the number of frequencies for which impedance measurements were taken because resistive impedance and reactive impedance are placed together on combined plots. For example, in one embodiment impedance measurements are taken at 71 different frequencies, and p is therefore 142 when resistive and reactive impedance plots are combined. Each number X, or X value, in the spectral matrix represents an integer from zero to p, and each number Y represents an impedance value on the combined graph associated with its corresponding X value.

0171 ADDING DATA TO THE COMBINED PLOT and spectral matrix 0172 Figure 45 shows graph 1300* comprising three Nyquist plots of the impedance spectra for three different lubricating fluid samples. X-axis 02 comprises values for resistive impedance, also known as resistive impedance, denoted Z'. Y-axis 04 comprises values for reactive impedance, also known as reactive impedance, denoted Z". As will be understood by those skilled in the art, for each of the three plots on graph 1300* the data points to the left of the minimum value for Z"represent the fluid bulk. Likewise, those skilled in the art will understand that the data points to the right of the minimum value for Z"for each of the three plots on graph 1300* represent the interfacial region between the fluid and metal. In one embodiment, the interfacial region would represent where the fluid is in contact with an engine. Some of the data points discussed below are not contained on the graph 1200* but will be known from data recorded from generating the impedance spectra.

0173 Figure 44 depicts chart 1200* giving examples of data that can be added to the combined plot derived from a Nyquist plot of resistive impedance versus reactive impedance.

Column 1202* lists the identifiers for each of the samples for which data is provided. Note that the data in chart 1200* was taken from the same fluid samples as was used to generate the Bode plots shown in Figures 39 and 40. Further, although graph 1300* shows only three Nyquist plots, in actual practice of the invention a Nyquist plot would be generated for each of the data samples in column 1202*.

0174 Column 1204* represents the Z'value where Z"is minimum. Column 1206* represents the Z"value where Z"is minimum. As noted above, the point at which Z"is minimum denotes the boundary between the bulk and interface regions of the Nyquist spectrum.

0175 Column 1208* represents the frequency in the spectra at which Z"is minimum.

Graph 1300* does not represent this value. However, all of the values for Z"willbe contained in the spectral matrix, and moreover, it is possible to determine from the spectral matrix the frequency in the impedance spectrum at which each value was recorded, inasmuch as the Bode plots used to create the spectral matrix originally contained the logarithm of frequency on the x axis.

0176 Column 1210* represents the maximum Z'value within the total data set. In some cases this value is the"Nyquist Max,"that is, the Z'value between the bulk and interfacial regions of the Nyquist spectrum. In other cases this value could be the Z'value for data points associated with the lowest or highest frequencies.

0177 Column 1212* represents the minimum Z'value within the total data set. In some cases this value is the"Nyquist Min, "that is, the Z'value between the bulk and interfacial regions of the Nyquist spectrum. In other cases this value could be the Z'value for data points associated with the lowest or highest frequencies.

0178 Columns 1214*, 1216*, 1218*, and 1220* all contain data associated with the points in the bulk region of the Nyquist spectrum, that is, the points between the origin of the graph and the minimum value for Z". As is known, these points describe a semicircle.

Certain information about the semicircle, or the circle that would result from completing the semicircle, can be helpful in predicting fluid properties.

0179 Column 1214* represents the Z'value for the centerpoint of the circle in the bulk region of the Nyquist spectrum, that is, the circle completed bythe semicircle drawn from the leftmost point on the x axis on which data is plotted to the point on the x axis at which Z"is plotted. This circle is sometimes referred to as the bulk circle.

0180 Column 1216* represents the Z"value for the centerpoint of the centerpoint of the bulk circle.

0181 Column 1218* represents a measurement in radians of the angle between the x axis and a line drawn through the origin of the graph and the centerpoint of the bulk circle. This measurement is referred to as the depression angle of the bulk circle.

0182 Column 1220* represents a calculation of the radius of the bulk circle.

0183 Columns 1222*, 1224*, 1226*, and 1228* all contain data associated with the points in the interfacial region of the Nyquist spectrum, that is by the points to the right on the x axis of the minimum value for Z". As is known, these points describe a semicircle. Certain information about the semicircle, or the circle that would result from completing the semicircle, can be helpful in predicting fluid properties.

0184 Column 1222* represents the Z'value for the centerpoint of the circle in the interfacial region of the Nyquist spectrum, that is, the circle completed by the semicircle drawn from the points to the right on the x axis of the minimum value for Z". This circle is sometimes referred to as the interface circle.

0185 Column 1224* represents the Z"value for the centerpoint of the interface circle.

0186 Column 1226* represents a measurement in radians of the angle between the x axis and a line drawn through the origin of the graph and the centerpoint of the interface circle.

This measurement is referred to as the depression angle of the interface circle.

0187 Column 1228* represents a calculation of the radius of the interface circle.

0188 The data in Figure 44 can be added to the spectral matrix and used to predict fluid properties alongside the selected values for resistive and reactive impedance already discussed. This data can be included in the spectral matrix by placing t on a graph either to the left or to the right of the combined Bode plots discussed above. Some, only one, or all of the data points discussed above may be used. In addition, skilled artisans will recognize that adding other data related to a scan of the impedance spectrum is within the scope and spirit of the present invention.

0189 Creation of the Result Matrix 0190 Figure 37 provides a flowchart describing a method for building a result matrix. In step 500*, a fluid sample having a known age is tested in a laboratory to determine values for a predetermined set of fluid conditions, and these values are recorded. As discussed above with respect to Figure 34 these values may be recorded in database 220* as baseline values 222*. Some of the laboratory tests used in present practice are mentioned above with reference to block 302* in Figure 35. In step 502*, a check is done to determine whether there are more fluid samples to be tested. If the answer is yes, the process returns to step 500*. If the answer is no, the method proceeds to step 504*. In general, there will be n iterations of the method, n being the number of fluid samples at different ages that are being tested.

0191 In step 504* a result matrix, taking the form described above with reference to block 302* of Figure 35, is constructed. Each number Yi, j in the result matrix represents a fluid condition with respect to the jth fluid property from a total of f fluid properties tested in the ith fluid sample from a total of n fluid samples.

0192 Figure 41 provides an example of a result matrix 900* populated with empirically achieved data representing properties of an engine lubricant. Sample column 902* contains an identifier for each of the fluids represented in the result matrix. Hours column 904* contains the number of hours during which the lubricating fluid has been in operation in the engine, i. e. , the age of the lubricating fluid. It should be understood that all of the data in the results matrix represents fluid conditions at a single temperature. Note that each value in column 904* contains the suffix"70", which indicates that the data in result matrix 900* represents fluid conditions at seventy degrees Celsius.

0193 Result columns 906*, 908*, 910*, 912*, 914*, 916*, and 918*each contain values representing the measurement of a particular fluid property determined by performing laboratory tests as described above. Result matrix 900* depicts fifteen sample lubricants in which seven fluid properties have been measured. Accordingly, when put into the form given above, result matrix 900* is a fifteen by seven matrix; that is, with respect to result matrix 900*, n is fifteen, because 15 sample fluids have been analyzed, and f is seven.

0194 Principal Component Analysis 0195 Figure 38 shows the process by which the present invention uses Principal Component analysis (PCA). PCA is a well known technique, described in a plethora of literature, including the articles"Principal Component Analysis Methods"and"Discriminant Analysis, The PCA/MDR Method", both found on the website of Thermo Galactic of Salem, New Hampshire at"www. galactic. com/Algorithms/pca. htm" and "www. galactic. com/Algorithms/discrim_pca. htm" respectively and incorporated by reference herein. PCA is also explained in the following, all of which are incorporated by reference herein : Michael Palmer, "Principal Component Analysis"found on the world wide web at "www. okstate. edu/artsci/botany/ordinate/PCA. htm" ; StatsSoft, Inc. ,"Principal Components and Factor Analysis"found on the world wide web at "www. statsoftinc. com/textbook/stfacan. html" ;"Principal Component Analysis"found on the world wide web at"www. casaxps. cwc. net/FactorAnalysis. htm". Because PCA is well known, it will be described herein only to the extentnecessary to explain how PCA is applied in the context of the present invention.

0196 The goal of PCA is to reduce the number of elements in a data set by selecting principal components that are associated with the most variation in the data set. PCA comprises iteratively removing independent variations from a data set. Thus, in the context of the present invention the goal of PCA is to create a reduced spectral matrix that comprises a subset of the spectral matrix discussed above with respect to Figure 36. PCA operating on spectral data uses the following relationship: A = SF + EA (3) where A is the n by p spectral matrix, S is an n by f score matrix, F is the f by p matrix containing the principal components, i. e. , a principal components matrix, and EA is an n by p error matrix. F is sometimes referred to as a matrix of eigenvectors because it is used to recreate the spectral matrix. As above, n is the number of spectral samples in the spectral matrix and p is the number of data points represented in the spectra. The number f represents the number of principal components. Thus, PCA applied to spectral data depends on the theory the expression SF can be used to recreate spectral matrix A.

0197 The actual calculations required for PCA in the present invention are performed by a statistical software package. Software packages that have been used to practice the invention include The Unscrambled from Camo Technologies of Woodbridge, New Jersey; Spectrum Quant+ from PerkinElmer, Inc. of Wellesley, Massachusetts; and MatLab from Mathworks, Inc. of Natick, Massachusetts. Those skilled in the art will appreciate and understand the usage of such packages. Accordingly, the process described in Figure 38 is relatively simple. At step, 600*, the spectral matrix is provided as an input to the PCA process. At step 602*, the aforementioned statistical software package provides output of the PCA process in the form of a principal component matrix.

0198 At step 604*, a set of regression coefficients associated with the principal components matrix is generated. This step serves the purpose of enabling selection of the principal components that appear to have a significant influence on the spectral matrix once the principal components are identified in the principal component matrix. That is, it is necessary to determine which principal components will be useful in recreating the spectral matrix, i. e. , which principal components have significant influence over variations in the data set.

0199 There are different known methods of identifying principal components, and the particular method used to select principal components is not critical to practicing the invention. However, present practice is to select principal components by generating a set of regression coefficients and plotting the regression coefficients over the points in the impedance spectra; this plot is called a regression spectrum. Regression coefficients are simply the set of coefficients obtained by regressing values in the impedance spectra against values representing known fluid properties, i. e. , select values from the result matrix. One skilled in the art will appreciate that standard statistical software packages, such as those discussed above, and a variety of regression methodologies, could be used to generate the regression spectrum.

0200 Step D_04 can be made clear by way of example. One fluid property measured in engine lubricants is Total Base Number (TBN). A predicted value for TBN with respect to a fluid sample at a given temperature and age can be retrieved from the result matrix, discussed above with reference to Figures 37 and 41, and regressed against the original impedance spectra generated as part of the process discussed above with reference to Figure 36.

Performing this regression generates a set of regression coefficients for TBN. These regression coefficients can then be plotted as shown in plot 1006* on graph 1000* of Figure 42. X axis 1002* comprises integers in the range of the integers contained on the x axis of the combined Bode plot discussed above with reference to Figures 36 and 40. Consistent with one embodiment referenced in examples previously given, plot 1006* is plotted from points zero to one-hundred and forty-one in relation to x axis 1002*. Y axis 1004* comprises a range of values of coefficients generated from the regression of TBN against the impedance spectra.

0201 A visual inspection of plot 1006* reveals that positive or negative peaks occur at points 18,25, 52,63, 70,85, 92,129, and 141 lying above x axis 1002*. That is, the regression coefficients at these points have a significant magnitude, and these regression coefficients thus exert a relatively higher degree of influence over the prediction of a value for TBN than other regression coefficients in plot 1006*. Accordingly, columns corresponding to points 18,25, 52,63, 70,85, 92,129, and 141 are selected from the spectral matrix and placed in the reduced spectral matrix in step 606*.

0202 It should be noted that, with respect to the above example, the present invention does not require using the regression spectrum for TBN, as opposed to other potentially available regression spectra, to pick points to be included in the spectral matrix. Other fluid properties, either individually as was the case with TBN in the above example, or in combination with one another, could have been used for this purpose. In this case, the regression spectrum for TBN was used because that regression spectrum was determined to lead to relatively accurate predictions of fluid properties. Other regression spectra may yield better results for other data sets.

0203 Figure 43 shows reduced spectral matrix 1100*. Header row 1122* lists the points in the combined spectrum that have been identified as principal components. Column 1102* lists the identifiers for the fifteen different lubricants for which combined impedance spectra were generated. Note that the fifteen lubricants identified in column 1102* are the same fifteen samples identified in column 902* of result matrix 900*. Columns 1104*, 1106*, 1108*, 1110*, 1112*, 1114*, 1116*, and 1118* each contain values from the combined plots of the resistive and reactive impedance spectra for each of the fifteen lubricants. The identifiers in column 1102* are the same as the identifiers in column 902* contained in result matrix 900* depicted in Figure 41; Figures 41 and 43 are based on the same laboratory test of the invention, and thus represent a result matrix and reduced spectral matrix respectively based on the same lubricating fluids.

0204 Statistical Analysis of the Reduced Spectral Matrix 0205 As discussed above with respect to block 310* of Figure 35, a number of different statistical techniques, some of which use linear relationships and some of which use nonlinear relationships, may be used to identify coefficients for use in prediction equations (2) above.

Linear regression methods with which the present invention may be practiced include but are not necessarily limited to Multivariate Least Squares Regression (MLR), also known as Multiple Linear Regression, Inverse Least Squares, or P-Matrix, as well as Principal Component Regression (PCR).

0206 The present invention takes advantage of the following linear relationship: R=PA+E (4) where result matrix R represents a concentration of a property or properties in a fluid, A represents a reduced spectral matrix, and coefficient matrix P is a matrix of loading coefficients. The matrix E is an error matrix, also known as the residual matrix, because it represents the difference between fluid properties estimated by the expression PA and the true value of fluid properties. It will be readily apparent to one skilled in the art that, where R and A are known, MLR can be performed to determine P. The matrix P in turn comprises coefficients that can be used in prediction equations 124* (2) to obtain predicted property values 128*.

0207 PCR extends MLR to regress against the scores obtained from PCA as opposed to spectral data itself. Recall that the theory of PCA scores matrix S is that S can be used to reconstruct the spectral matrix A. Assuming that there is a linear relationship between spectral matrix A and concentration C, represented in the present invention by result matrix R, it is true that R=BS+E (5) It will be readily apparent to one skilled in the art that, where R and S are known, MLR can be performed to determine B. The matrix B in turn comprises coefficients that can be used in prediction equations (2) to obtain predicted property values 128*.

0208 Nonlinear regression techniques with which the invention has been practiced include pattern recognition analyses, cluster analyses, and neural network analyses. One nonlinear regression technique that has been successfully applied in the present invention is the Group Method of Data Handling (GMDH) algorithm using the software program KnowledgeMiner available on the World Wide Web from Script Software. First, a software program such as Mathematica from Wolfram Research, Inc. of Champaign, Illinois, is used to read selected resistive and reactive impedance values for a given set of frequencies from a set of files. For example, approximately one frequency at each decade of the frequency range could be used. In one instance of practicing the invention impedance values at lOkHz, lkHz, 100 Hz, lHz, O. lHz, 0.01 Hz were selected.

0209 The next step is to calculate internal variables that assist in describing a geometric shape profile of each impedance spectra. These may consist of the length vector between each possible pair of two frequencies that have been collected. The angle, delta, of each selected point on the Nyquist representation of the spectra can be included; this angle is generally referred to as the dissipation factor, and the tangent of angle delta is calculated by dividing resistive impedance by reactive impedance. The tangent of angle delta represents the ratio of energy loss versus energy stored at a given frequency. Also, the inverse values of the afore-mentioned measured and internal variables are also calculated, allowing for variable combinations that include both quotients and products.

0210 In order to determine the presence of redundancy in the variables a cluster analysis can be preformed. Cluster analysis is a statistical procedure that separates and groups a set of data into smaller sets of similar data. If multiple variables are found to be statistically similar, one variable from this variable cluster is selected. The resulting reduced variable set is then paired with an appropriate physio-chemical parameter to which a correlation is to be determined. This data is then modeled using Group Method for Data Handling, which is found in commercially available software, i. e. , KnowledgeMiner. The resulting correlation could be considered a multi-layered neural network composed of connection weights that are polynomial, (including linear) functions.

0211 Other application of nonlinear regression techniques to the reduced spectral matrix may be apparent to skilled artisans. Further, skilled artisans will understand that the aforementioned linear regression techniques will be used to find loading coefficients for prediction equations 124* (2) that can in turn be used to obtain predicted property values 128*.

0212 RUL and EOL Determinations 0213 As discussed above with respect to step 312* described in Figure 35, predicted fluid properties can be used to estimate either the Remaining Useful Life (RUL) of the fluid, or when the fluid will reach its End of Life (EOL). Referring to Figure 33, microcomputer 116* can be programmed with a variety of logical functions comparing baseline values 122* to the results of prediction equations (2), namely property predictions 126*, to determine a fluid's RUL and/or EOL.

0214 As discussed above, each of the baseline values 122* has as attributes the fluid property for which the value was recorded, the age of the fluid sample with which the value is associated, and the temperature with which the value is associated. Measurements of RUL and EOL may take a variety of forms. For example, RUL may be expressed as a percentage of remaining useful life, as a percentage of useful life expended, as the number of miles a vehicle can be run before the fluid will reach EOL status, or as the number of hours for which an engine can be run before a fluid will reach EOL status. Similarly, EOLmay represent a value for a fluid property at which the fluid has reached the end of its useful life or an age in hours at which the fluid has reached the end of its useful life.

0215 In general, microcomputer 116* will be programmed with at least one logical function that evaluates an expression comparing one or more of the baseline values 122* with one or more of the property predictions 126*. It should also be understood that the determination of RUL and EOL may be based on one or a plurality of fluid properties.

0216 In the simplest case, determining RUL or EOL based on one fluid property, programming of microcomputer 116* could require iteratively evaluating the expressions (Y > vl, Y > v2,..., Y > vn) where Y is one of the property predictions 126*, each value vi is one of the baseline values 122*, and n is the number of fluid samples representing different fluid ages for which baseline values 122* have been stored in information library 120*.

Iterations would continue until one of the expressions (Y > vi) evaluated to FALSE, i. e. , until the program established that prediction of the fluid property value made in real time exceeded a threshold value, thus enabling the program to predict the age of the fluid based on the threshold value exceeded.

0217 In a more general case, programming could iteratively evaluate a compound logical expression such as (Yl > vl, i OR Y2 > v2, i OR... OR Yk > vk, n) where Yl... Yk are property predictions 126*, vl, i... vk, n are baseline values 122*, k is the number of fluid properties being evaluated, and n is the number of fluid samples representing different fluid ages for which baseline values 124* have been stored in information library 120*. Thus, if any of the fluid properties of interest fail to meet a threshold established by the relevant baseline values 122*, the expression will evaluate to FALSE, and an RUL determination can be made based on the age attribute associated with vl, i... vk, n. Similarly, microcomputer 116* could be programmed to make an EOL determination when any given threshold represented by one of EOL/RUL values 124* failed to be met. Also, the logical operator OR could be replaced by the logical operator AND in the logical expression, requiring that certain thresholds be met for all or at least a subset of property predictions 126* before and EOL and/or RUL determination is made.

0218 The preceding discussion of the programming of microcomputer 116* is meant to be illustrative rather than limiting, inasmuch as a skilled artisan would recognize that a number of different algorithms could be implemented to make EOL and RUL determinations.

Any determination of EOL and RUL using property predictions 126* or predicted property values 128* would be within the scope of the present invention.

0219 Although not shown on Figure 33, microcomputer 116* provides output of both RUL and EOL determinations. In one embodiment, for example, output of EOL determinations is manifested in the activation of a warning light on a vehicle console. In another embodiment, an RUL determination is manifested on a digital display embedded in a vehicle console. These examples are meant to be illustrative rather than limiting, inasmuch as any mode of outputting and displaying EOL and RUL determinations would be consistent with the present invention.

0220 Alternative Embodiments 0221 The preceding disclosures of the invention assume that the spectral matrix is created as described in Figure 36. According to the embodiment described in Figure 36, Bode plots of resistive and reactive impedance spectra for a plurality of fluid samples are laid"head to tail"on a combined plot, and the data from the combined plot is then placed into the spectral matrix. Figure 46 describes an alternative embodiment in which data from each of the Bode plots of resistive and reactive impedance spectra are processed separately, with results then combined for use in predicting fluid properties.

0222 Prior to step 1400*, steps 400* through 418* will have been performed as described above with reference to Figure 36. However, instead of combining Bode plots as described with reference to step 420* and then creating a reduced spectral matrix as described with reference to step 428*, in step 1400* a first spectral matrix is created solely from the resistive impedance spectra. Similarly, in step 1402*, a second spectral matrix is created solely from the reactive impedance spectra.

0223 Step 1404* follows the process discussed above with reference to figure 38 to perform PCA with respect to the first spectral matrix to create a first reduced spectral matrix.

Similarly, step 1406* performs PCA with respect to the second spectral matrix to create a second reduced spectral matrix.

0224 After step 1406* is complete, the invention can be practiced either by combining the first and second reduced spectral matrices and performing a regression on the resultant combined reduced spectral matrix, or by performing regressions on each of the first and second reduced spectral matrices and combining the results. Accordingly at step 1408* a check is done to determine whether regression is to be performed on the first and second reduced spectral matrices together or separately.

0225 If regression is to be performed on the first and second reduced spectral matrices together, they are combined in step 1410*. The first and second reduced spectral matrices will each have n rows, n being the number of fluid samples being tested, but may have different numbers of columns because the number of principal components identified for each of the first and second reduced spectral matrices may or may not be the same. The first and second reduced spectral matrices are combined simply by placing the data in the two matrices side by side, so that the combined reduced spectral matrix has n rows, and has the number of columns that is the sum of the number of columns in each of the first and second reduced spectral matrices.

0226 In step 1412*, MLR is performed with respect to the combined reduced spectral matrix in the same manner as described above with respect to the reduced spectral matrix.

Similarly, the scores from the PCAs performed on the first and second spectral matrices could have been combined and a PCR performed on the combined scores in the manner described above.

0227 If the response to the check done in step 1408* is that regression is to be performed on the first and second reduced spectral matrices separately, a first MLR is performed on the first spectral matrix in step 1414*. In step 1416*, a second MLR is performed on the second spectral matrix. In step 1418* the results of the first and second MLRs are then used in the prediction equations 124* (2) in the manner described above, and the results of the prediction equations are then combined to yield final predictions of fluid properties. The results of prediction equations might be combined in a number of ways; in one embodiment they are averaged. Again, PCR also could have been performed in this manner.

0228 Embodiments of the invention discussed thus far comprise using at least one Bode plot to create a spectral matrix. However, some embodiments of the invention create a spectral matrix without using data from a Bode plot, i. e. , make a spectral matrix containing only Nyquist-derived datum values that are processed using the techniques that have been described above with respect to Bode plots. As discussed above with reference to Figure 44, there is a considerable amount of data incidental to practicing impedance spectroscopy that is not contained in or derived from a Bode plot. For example, as discussed above regarding Figure 44, data from Nyquist plots of impedance spectra has also been found to be useful when added to a combined Bode plot of real and reactive impedance, and then used to create a spectral matrix. Some embodiments of the present invention forgo the use of data from Bode plots and construct the spectral matrix solely from data such as the data discussed with reference to Figure 44. Processing otherwise proceeds as described above with reference to Figure 36.

0229 The above description is intended to be illustrative and not restrictive. Many embodiments and applications other than the examples provided would be apparent to those of skill in the art upon reading the above description. The scope of the invention should be determined, not with reference to the above description, but should instead be determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled. It is anticipated and intended that future developments will occur in the field of determining fluid conditions using impedance spectroscopy and that the disclosed systems and methods will be incorporated into such future embodiments. Accordingly, it will be understood that the invention is capable of modification and variation and is limited only by the following claims.