Crude oil, also known as petroleum, is a complex mixture of hydrocarbons. The component hydrocarbons are separated from one another to produce commercially valuable petroleum products. The various physical and chemical processing steps for separating crude oil into component hydrocarbons are collectively referred to as"refining." Because the component hydrocarbons differ in their volatility, an effective technique for separating them from one another is fractional distillation. In this technique, heated petroleum is placed in a distillation tower having a temperature gradient that progressively decreases from a high at the base of the tower to a low at the top of the tower.

Petroleum vapor, which consists of the component hydrocarbons in their vapor phase, rises through the distillation tower. As it rises, the petroleum vapor encounters progressively lower temperatures. When the petroleum vapor reaches a height at which the temperature in the distillation tower is equal to the condensation temperature of one of the vapor's components, that component condenses. A tray placed at that level of the tower collects that condensed hydrocarbon component.

The relative amounts of each petroleum fraction generated in the distillation tower depend, to a great extent, on the composition of the petroleum. A refiner's control of the composition of any particular variety of petroleum to be processed is limited. Refiners must often process equity crudes or those available on an attractive commercial basis from the short term or spot markets. Therefore, there is limited ability to regulate the yields of different petroleum fractions generated by distillation. In addition, demands for light, clean petroleum products normally exceed the quantity of these products appearing naturally in crude oils which are recoverable by distillation only.

At the dawn of the petroleum refining industry, the scarcity of electric lighting made clean-burning lamp oils such as kerosene the petroleum fraction in highest demand.

However, with the advent of the internal combustion engine, the demand for lighter petroleum fractions, such as gasoline, increased significantly at the expense of heavier petroleum fractions, such as kerosene. In response, the petroleum refining industry has

developed methods for increasing the gasoline yield obtained from a particular varie of petroleum. Catalytic cracking is one such method.

In catalytic cracking, a petroleum fraction heavier than gasoline, commonly called gasoil, is fed into a reaction chamber and vaporized into a hydrocarbon vapor at an extremely high temperature. Ceramic pellets containing catalytic agents are introduced into the hydrocarbon vapor. A high velocity vapor stream propels the ceramic catalyst pellets into the reaction chamber and disperses them throughout the hydrocarbon vapor. The catalyst carried by the pellets facilitates a chemical reaction in which the hydrocarbon molecules split, or"crack,"into lighter molecules. These lighter molecules pass to a distillation tower called the main fractionator for recovery. Because the vapor now contains lighter hydrocarbon molecules, distillation of the vapor yields additional gasoline and other light hydrocarbons.

The extent to which hydrocarbon vapor is cracked into lighter molecules depends, in part, on the hydrocarbon composition supplied to the reaction chamber and on the rate at which it is supplied.

In conventional methods for controlling the operation of a catalytic cracker, measurements of the gasoil composition and products from the main fractionator are obtained by periodically sampling the feed and product streams, and testing the samples in a laboratory. This, however, is costly and time-consuming. As a result, it is difficult to perform laboratory testing frequently enough to attain up-to-date values of selected quality properties of the feedstock and all the distillates which the cracked hydrocarbon mixture yields. Consequently, this known method results in poor control over the catalytic cracking process.

SUMMARY The method of the invention overcomes deficiencies in the art by performing on-line measurements of the values of selected properties of one or more hydrocarbon materials involved in the catalytic process. In one instance, these are materials fed to the catalytic cracker, i. e. the input feedstock. In another instance, these materials involved in the catalytic process are at least one output fraction from the main fractionator tower fed by a fluid catalytic cracker. The results of these on-line measurements, together with the desired values of the selected properties of the materials being measured, are provided to an

automated controller. On the basis of these measurements, the controller calculates values of manipulated process variables to apply to the fluid catalytic cracker to obtain output fractions having the desired values of the selected properties. Because the estimates are performed on-line, a system incorporating the invention responds rapidly to variations in the measured materials.

In most practices of the invention, it is deemed preferable to secure both gasoil feedstock measurements and output product measurements, and to manipulate process variables in response to both groups of measurements. However, features of the invention may extend to practices where only the gasoil feedstock is measured, or where only one or more output fractions are measured, and one or more manipulated process variables are adjusted in response to such measurement. The process variables are adjusted for feed-forward control in response to measurements of gasoil feedstock material, and are adjusted for feed-back control in response to measurements of product material.

A preferred practice of the invention employs nuclear magnetic resonance measurements in conjunction with values of selected material properties, e. g. of the gasoil feedstock or of one or more product fractions. NMR measurements do not rely on optical or infrared radiation, and are therefore not affected by high opacity. Additionally, NMR measurements are relatively stable as a function of temperature. Consequently, reliable measurements can be made across the broad range of temperatures of the gasoil feedstock and of the various product fractions produced in the main fractionator.

One illustrative practice of the invention includes the steps of imposing a steady magnetic field on a material involved in the catalytic process, for example on a sample of a hydrocarbon output fraction. With the steady magnetic field in place, a nuclear magnetic resonance (NMR) sensor imposes a transient magnetic field on the fraction and measures its response to the transient magnetic field. An NMR sensor generally does not directly provide values of the selected properties of the fraction. Instead, the NMR sensor provides the chemical composition of the fraction. For this reason, the method of the invention includes the step of ascertaining values of the properties of the measured material on the basis of its measured composition. These values are then used to selectively control process variables such as the gasoil flow rate into the fluid catalytic cracker, the regenerator catalyst slide valve position, reactor dense bed level, regenerator pressure, feed preheat temperature, air or oxygen injection rate, reactor/regenerator differential pressure, wet gas

compressor suction pressure, main fractionator product draw rates, steam injection rates, overhead reflux duty, overhead accumulator release, and slurry to steam generators, thereby generating a product fraction having desired values of selected quality properties. The exact set of process variables available for manipulation will depend upon the specific configuration of the individual fluid catalytic cracking unit. Examples of these selected quality properties include aromaticity, boiling point, flash point, cloud point, viscosity, pour point, API gravity, freeze point, octane, PIONA, RVP, and other chemical or physical properties.

The information obtained from the NMR measurements can, in accord with the invention, be used to explicitly control both the conversion in a fluidized catalytic cracking unit, and the selectivity of the unit to the desired levels. This practice of the invention involves NMR measurements on materials directly involved in the catalytic conversion, and are made on-line and essentially in real time.

The desired product qualities are generally dictated by competitive market forces, environmental regulations and economic factors including feed and product prices and operating costs. They may also be constrained by specific characteristics of the refinery including equipment configuration and availability and operating constraints.

The method of the invention can thus provide real-time measurements of the values of selected properties of gasoil feed and of a variety of fractions produced by the main fractionator. These real-time measurements enable the control system of the invention promptly to adjust the operation of the catalytic cracker to compensate for fluctuation in the feedstock and/or in the output of the main fractionator. Because the method of the invention relies on NMR rather than on optical techniques, the accuracy of these measurements can be essentially independent of the opacity or the temperature of the measured material.

A system for practice of the invention includes an optimizer for specifying desired long-term values of selected fluid catalytic cracker operating variables that produce products of the desired qualities while optimizing economic operation of the unit considering among other things feed and operating cost and product values. The optimizer also takes into account specific characteristics of the refinery including equipment configuration and availability and operating constraints including product production limits.

A system for practice of the invention further includes a sensor for estimating the values of selected product properties. These estimates, together with the desired long term values of selected operating variables specified by the optimizer, are provided to a controller. On the basis of the values from the sensor, desired product properties values and the long-term desired values from the optimizer, the controller determines current values of the operating variables needed to form products having the desired values of the selected properties, while optimizing economic performance of the refinery, and in particular of the catalytic cracker.

Practice of the invention also includes measuring, with nuclear magnetic resonance, parameters regarding the composition of at least one material or substance directly involved in a catalytic converter in a petrochemical process, and applying the measured information to a controller of the petrochemical process. The measured material or substance can be the feedstock, or an ingredient of the feedstock, and can be one or more petrochemical product fractions. A preferred practice performs both feedstock and product measurements. The controller determines, in response to the measured information and to further target information it stores or receives regarding desired parameters of the petrochemical process, including of materials produced or otherwise involved, control signals for adjusting the process to attain selected values of the measured information relative to the target information.

DESCRIPTION OF THE FIGURES These and other features and advantages of the invention will be apparent from the following detailed description and the accompanying figures, in which : FIG. 1 is a schematic block drawing of a multivariate distillation control system embodying features of the invention ; FIG. 2 is a schematic block drawing showing details of the distillation system of FIG. l ; and FIG. 3 is a schematic block drawing of another distillation system according to the invention.

DESCRIPTION OF ILLUSTRATED EMBODMENTS FIG. 1 shows a multivariate control system 10 incorporating principles of the invention. The multivariate control system 10 includes a petrochemical cracking system 12 that receives an input feedstock and that is in communication signal-wise with a multivariate controller 14. The output product of the cracking system 12 includes a plurality of hydrocarbon products each characterized by values of selected properties. These values, which are represented in FIG. 1 by an output vector y that is fed back to the multivariate controller 14, are controlled variables for the multivariate control system 10.

The effectiveness of the illustrated controller 14 is enhanced by also applying to it an input vector z representing values of selected properties that characterize the input gasoil feedstock. The multivariate control system 10 includes an optimizer 16 for generating desired long-term values of selected operating variables that produce the desired product qualities, while optimizing economic operation. These desired values are represented in FIG. 1 by a setpoint vector r generated by the optimizer 16 and supplied to the multivariate controller 14. The elements of vector r are specified taking into account feed and operating cost, product economic values, and specific characteristics of the refinery including equipment configuration and availability and operating constraints including product production limits.

In response to the set-point vector r and one or more differences between desired product qualities and the output vector y and/or changes in feedstock vector z, the multivariate controller 14 generates a vector of manipulated variables x that, when applied to the cracking system 12, adjusts the cracking system operation to change the values of the elements in the output vector y to approach the corresponding desired values.

FIG. 2 shows that the illustrated cracking system 12 includes an input pipeline 18 leading to a reaction chamber 20. A pump 33 coupled to the input pipeline 18 propels a gasoil feedstock mixture into the reaction chamber 20 through a control valve 26 that is opened and shut by a valve actuator 28 under the control of the multivariate controller 14.

The reaction chamber 20 is also in fluid communication with a regeneration chamber 32 by means of a catalyst feed pipe 34 and a catalyst removal pipe 36. This regeneration chamber 32 has an air or oxygen intake 38 and an exhaust 40 for use with the catalyst regeneration process. A control valve 22 in the catalyst feed pipe 34 is controlled by an actuator 23 that responds to input control signals from the controller 14. Hot regenerated ceramic catalyst

particles from the regeneration chamber 32 are added, by way of the catalyst feed pipe 34, to the gasoil feedstock in the riser prior to delivery to the reactor chamber 20. The heat from the catalyst particles causes the gasoil feedstock to vaporize and the resultant high velocity vapor stream carries the catalyst particles into the reactor chamber 20.

The reaction chamber 20 is in fluid communication with the main fractionator tower 30 of the type commonly found in petroleum refineries. Such a distillation tower 30 typically includes a plurality of tower outputs 31 located at different heights along the tower 30. Each output corresponds to a particular fraction distilled from the petroleum input.

Outputs located at the lower portions of the tower 30 correspond to heavier fractions, such as heating oil or kerosene. Outputs located at the higher portions of the tower 30 correspond to lighter fractions, such as gasoline or naphtha. The number of tower outputs affects the computational burden sustained by the multivariate controller 14, and is not the subject matter of the invention. Hence, for the sake of clarity and ease of exposition. FIG. 2 shows only four tower outputs 31 a-d.

A nuclear magnetic resonance (NMR) sensor 42 is coupled to various selected stages of the cracking system 12 to measure chemical properties of material at those selected stages in the process. The illustrated sensor 42 of FIG. 2 is coupled to the input pipeline 18 to sample the feedstock, and is coupled to each tower output 3 la-d to sample each product, or fraction, produced by the main fractionator tower 30. Although the cracking system 12 shown in Figure 2 has a single NMR sensor 42 responding, for example, on a time-shared basis to each of multiple inputs, it will be understood that one or more additional NMR sensors can be provided to respond to the sampled materials. The sensed information output from the NMR sensor 42 is applied to a calibrator 44 that transforms it to a format suitable for the multivariate controller 14.

One preferred sensor employs the technology of the I/A Series Process NMR equipment available from The Foxboro Company of Foxboro, Massachusetts, however, a wide variety of NMR sensors may be used for chemical analysis.

A suitable calibrator is preferably a chemometic modeling unit. A chemometic modeling unit for the calibrator 44 and suitable for practice of the invention is implemented by a digital processor executing instructions for estimating values of selected physical properties on the basis of the measured chemistry of a sample, as determined by an NMR sensor. These instructions implement procedures that are well-known to those of ordinary

skill in the art. Such procedures include establishing look-up tables, interpolating between values in a look-up table, and implementing mathematical models for estimating values of the selected properties. FIG. 3 shows the calibrator 44 as a separate element for clarity, however, the calibrator 44 may be integrated into software anywhere in the system.

Both the optimizer 16 and the multivariate controller 14 are preferably implemented as software instructions executed on a programmable digital processor. In one practice, these instructions are executed on a general purpose digital computer. However, particularly to meet demanding performance requirements, the optimizer 16 and the multivariate controller 14 can be implemented in hardware, for example, with application specific integrated circuits, in software, or in a combination of hardware and software. A suitable optimizer and multivariate controller for practice of the invention are sold by Simulation Sciences of Brea, California, USA under the trade designations ROMeo and Connoisseur, respectively. The specific implementation details of the multivariate controller 14 and the optimizer 16 are within the purview of one of ordinary skill in the art and do not affect the scope of this invention.

In operation of the system shown in FIGS. 1 and 2, the pump 33 propels gasoil feedstock through the control valve 26 and into the reaction chamber 20. At the same time, small hot pellets of catalyst material are added through the catalyst feed pipe 34 and the slide valve 22, from the regeneration chamber 32. The heat from the catalyst particles causes the gasoil feedstock to vaporize and the resultant high velocity vapor stream propels the catalyst particles into the reaction chamber 20. The catalyst is typically a ceramic material, such as zeolyte. However, other catalysts are known and used in connection with fluid catalytic cracking. The presence of the catalyst within the reaction chamber 20 facilitates breaking the covalent bonds of the molecules in the gasoil mixture to produce lower molecular weight hydrocarbons.

Heat is produced by burning carbon off catalyst in the regenerator 32. This heat is transferred to the reactor via the catalyst feed pipe 34 to heat the gasoil feed to sufficient temperature to catalytically crack the gasoil to lighter molecules. This results in the generation of a vaporous mixture of numerous hydrocarbon species having generally lower molecular weights than the hydrocarbon species present in the feedstock. This vaporous mixture of low molecular weight hydrocarbons flows into the main fractionator 30 for

separation into a plurality of products. These products exit the main fractionator 30 through a plurality of tower outputs, four of which are shown in FIG. 2.

A byproduct of the cracking process is the deposition of carbon on the catalyst pellets. These carbon deposits suppress the interaction between the catalyst and the gasoil vapor, thereby reducing the effectiveness of the catalyst. As a result, catalyst pellets are continuously drawn off the reaction chamber 20, blown through the catalyst removal pipe 36, and returned to the regeneration chamber 32. Air or oxygen from the intake 38 is injected to burn carbon off the catalyst pellets to renew catalytic activity and to generate heat to sustain the cracking process. The gaseous byproducts of this regeneration process are vented from the regeneration chamber through the exhaust 40.

As noted above, the illustrated NMR sensor samples the feedstock and samples the products from the main fractionator 30. The NMR sensor 42 imposes a steady magnetic field on each such sample to align the magnetic dipole moments associated with the molecules in the sample. With the steady magnetic field in place, the NMR sensor 42 imposes a transient magnetic field having a direction different from, and preferably orthogonal to, that of the steady magnetic field. This transient magnetic field temporarily aligns the dipoles from the sample in a direction other than that in which the static magnetic field aligns dipoles. When the transient magnetic field is turned off, the dipoles in the sample spring back into the alignment imposed upon them by the steady magnetic field. As they do so, the dipoles generate a RF signal. The rate at which a particular dipole springs back to alignment with the steady magnetic field, and hence the frequency of the resulting RF signal, is characteristic of the sample's molecular structure. The resultant RF spectrum is responsive to, and hence provides a way of determining, the chemical composition of the sample.

The NMR sensor thus provides information on the chemical composition of materials or substances directly involved in the process stream of the cracking system 12.

As illustrated, these include the feedstock and the fractions formed in the main fractionator 30. It is known in the art to predict the corresponding values of selected properties from this measured composition of the sample. This operation of converting the measured sample composition to values of selected properties is carried out, in the embodiment of FIG. 2, by a calibrator 44 in communication with both the multivariate controller 14 and the NMR sensor 42. The input information to the calibrator 44 includes the chemical composition of

the samples as measured by the NMR sensor 42. The output of the calibrator 44 is a corresponding set of measured values of the selected properties for each measured sample.

These selected properties can include aromaticity, boiling point, flash point, cloud point, viscosity, pour point, API gravity, freeze point, octane, PIONA, RVP, or other chemical or physical properties.

In response to the data provided by the NMR sensor 42, together with other conventional data, as from other sensors (not shown) of level, pressure, pressure differential, temperature, time, and flow rate as conventional and known in the art, the multivariate controller 14 determines values of the manipulated variables which minimize the difference between the desired values of the product qualities and the values of the corresponding selected product qualities, as provided by the NMR sensor 42. Desired long- term values of the operating variables as determined by the optimizer 16 are also taken into account by the controller 14. The controller 14 determines the values of the manipulated variables using methods known in the art. These methods include the use of a look-up table and the implementation of empirically derived mathematical models. These manipulated variables include the flow rate of gasoil feed into the reaction chamber, regenerator catalyst slide valve position, reactor dense bed level, regenerator pressure, feed preheat temperature, air or oxygen injection rate, reactor/regenerator differential pressure, wet gas compressor suction pressure, main fractionator product draw rates, steam injection rates, overhead reflux duty, overhead accumulator release, and slurry to steam generation. The exact set of process variables available for manipulation will depend upon the specific configuration of the individual fluid catalytic cranking unit.

Having determined the values of the manipulated variables, the multivariate controller 14 transmits control signals to the actuator 28 to marginally open or close the control valve 26 to adjust the flow rate of the gasoil feed into the reaction chamber 20.

Similar signals are sent to control the other manipulated variables.

Because the NMR sensor 42 and the calibrator 44 rapidly provide information concerning the process stream to the multivariate controller 14, the controller 14 will readily respond to fluctuations in properties of the feedstock and of the distillates. This enables the controller 14 to change the settings associated with the catalytic cracking process to reduce any difference between measured properties and corresponding desired properties.

FIG. 3 shows an NMR stage 42'that can be used in the system of FIG. 2 in place of the NMR 42 of that drawing. The illustrated NMR stage 42'has a set of multiple NMR sensors 42a, 42a, each connected to receive information regarding the different product sample. The inputs to each NMR sensor 42a, 42a, can be a separate tower output 31 a, 31b... of FIG. 2. The output signals from the NMR sensors are applied to an interface selector 42b that operates in conjunction with the multivariate controller 14 of FIG. 2. The illustrated interface selector 42b applies the output from any selected one NMR sensor 42a to a calibrator 44, typically the calibrator 44 of FIG. 2.

As also shown in FIG. 3, the illustrated NMR stage 42'has a further NMR sensor 42c, the output of which is applied to the input selector 42b. The illustrated NMR sensor 42c receives information from either of two feedstock samples, by way of sample lines 43a and 43b and a controllable selector 42d. The two input lines 43a and 43b illustrate for purposes of example only, that the invention can be practiced both with measured information from multiple products and with information from multiple feedstock samples.

The illustrated NMR stage 42'further illustrates that the invention can be practiced with one or more NMR sensors that receive information from a single sample or from any of two or more material samples, depending on the optimum instrumentation strategy for a particular petrochemical conversion facility.

The system configurations described above illustrate the chemical cracking techniques described herein. However, a wide variety of other configurations can also take advantage of these techniques.

Having described the invention, and a preferred embodiment thereof, what is claimed as new and secured by Letters Patent is :