PROGNOSTICS AND HEALTH MANAGEMENT

BACKGROUND

[oooi] A variety of complex device are in widespread use today. Many of those devices have various operating parameters that indicate whether the device is functioning properly or if there may be a problem with the operation of the device. For example, fuel cell systems have specified threshold limits for certain performance variables. For example, there are temperature limits for various portions of a fuel cell system during acceptable operating conditions. There are also limits on output voltage or current for many fuel cell systems.

[0002] Significant study has been devoted to prognostics and health management (PHM) and principle component analysis (PCA) for detecting when a device is operating under conditions that depart from an expected or desired operating state. One limitation on such approaches is that the analysis is done with respect to the normal or baseline operation of the device instead of basing the analysis on threshold limits on the operating parameters.

SUMMARY

[0003] According to an embodiment, a method of monitoring the operation of a device includes determining a plurality of operational parameters that are indicative of an operation condition of the device. A difference between each operational parameter and a corresponding limit on that parameter is determined. Each limit indicates a value of the corresponding operational parameter that corresponds to an undesirable operation condition of the device. An action index is determined based on at least a smallest one of the determined differences. A determination is made whether the action index is within a range corresponding to desirable operation of the device.

[000 ] According to an embodiment, a system for monitoring device operation includes a plurality of detectors that provide respective indications of operational parameters that are indicative of an operation condition of the device. The system includes a processor that is configured to determine a difference between each operational parameter and a corresponding limit on that parameter. Each of the limits indicates a value of the corresponding operational parameter that corresponds to an undesirable operation condition of the device. The processor is configured to determine an action index based on at least a smallest one of the determined differences. The processor is configured to determine whether the action index is within a range corresponding to desirable operation of the device.

[0005] The various features and advantages of a disclosed example embodiment will become apparent to those skilled in the art from the following detailed description. The drawings that accompany the detailed description can be briefly described as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] Figure 1 schematically illustrates a system for monitoring operation of an example device.

[0007] Figure 2 is a flowchart diagram summarizing an example approach.

[0008] Figure 3 graphically illustrates a relationship between an action index and the shutdown limits related to the action index.

DETAILED DESCRIPTION

[0009] Figure 1 schematically illustrates a system 20 for monitoring operation of a device 30. The disclosed system and method of this description is not necessarily limited to any particular device 30. In the illustrated example, the device 30 comprises a fuel cell power plant. Selected portions of an example fuel cell power plant are illustrated for discussion purposes. A cell stack assembly (CSA) 32 contains a plurality of fuel cells that generate electrical power in a known manner. A source of fuel 34 supplies the CSA 32. A coolant assembly 36 selectively provides coolant for controlling a temperature of the CSA in a known manner.

[oooio] The system 20 for monitoring operation of the device 30 includes a plurality of sensors or detectors that are situated for detecting a plurality of operational parameters that provide an indication regarding the operation condition of the device 30. In the illustrated example, a temperature sensor 40 provides an indication of a temperature of exhaust from the CSA 32. Another detector 42 provides an indication of a voltage or temperature within the CSA 32. Another detector 44 provides an indication regarding the content or amount of fuel provided by the fuel supply system 34 to the CSA 32. Another detector 46 provides information regarding the coolant system 36 such as information regarding a temperature of the coolant or a concentration of a particular component within the coolant. Another detector 48 provides an indication of coolant temperature exiting the CSA 32.

[oooii] Given this description, those skilled in the art who are dealing with a particular device of interest will be able to configure a set of detectors to provide the necessary operational parameter information for monitoring the operating condition of the device with which they are dealing. The illustrated detectors are provided for discussion purposes. The disclosed example embodiment is not necessarily limited to any particular device or any particular arrangement of detectors.

[00012] A processor 50 collects information from the detectors 40-48 for monitoring the operating condition of the device 30. Based on determinations made by the processor 50, an output is provided to a user through a user interface 52. The output may be a visible or audible alarm or some indication that there is reason to adjust or shutdown operation of the device 30. Depending on the particular device and the conditions that are being monitored, the output provided by the user interface 52 may be customized to meet the needs of a particular situation.

[00013] Figure 2 includes a flowchart diagram 60 that summarizes an example approach for monitoring an operating condition of the device 30 in Figure 1. At 62, a plurality of operational parameters are determined. This occurs by gathering information from the detectors 40-48 and processing them within the processor 50 to place the detector information into a usable form. At 64, a determination is made regarding the distance between each operational parameter and a corresponding limit. In one example, the distance is determined as the Euclidean distance between the operational parameter value indicated by the corresponding detector and a limit that is determined for that particular parameter. In the case of a fuel cell as the example device 30, there are limits on temperature of various components within the device 30. Acceptable limits on temperature are determined based on known information regarding acceptable operation parameters to ensure proper operation of the device 30 and to facilitate achieving a desired lifetime for the device. Given a particular device 30, the particular limits on the operational parameters at issue will either be known or can be determined to meet the needs of a particular situation.

[0001 ] One way in which the disclosed example departs from previous PHM techniques is that the difference or distance between an operational parameter and a limit on that parameter is determined instead of determining how much an observed operation parameter differs from a base line or expected value for that parameter. With the disclosed example, taking into account the distance or difference between the operational parameters and their corresponding limits allows for obtaining an early warning of a condition that may lead to a desire or need to shutdown operation of the device 30. Obtaining an early warning allows for being more proactive in addressing a condition of a device 30 before having to shut it down, for example. Another feature of taking the approach of the disclosed example is that it allows for being more lenient in setting thresholds that place limits on operational parameters. For situations in which a device would be shut down when a limit on a particular parameter is met, those limits must be strictly set to avoid catastrophic failure of the device. Taking the approach of the disclosed example and obtaining an early warning of an operational parameter approaching a limit on that parameter provides more leeway in setting a threshold as it becomes possible to address an operating condition of the device before an absolute threshold on that particular is met.

[00015] Once the distance between each operational parameter and its corresponding limit has been determined, the processor 50 determines which of those distances is the smallest. In other words, the processor 50 identifies which of the operational parameters is closest to the limit on that parameter.

[00016] At 66, an action index is determined based at least on the smallest one of the determined distances. In one example, the action index comprises a shutdown index. For a situation in which device operation is monitored for purposes of shutting down the device to avoid failure, the action index provides information regarding taking action to shut down the device. There are other possible action indices such as a coolant replacement index, a recharge index, a fuel adjustment index, among many others. A shutdown index is used for an example for discussion purposes.

[00017] In one example, the action index is determined based upon modified observation values. Each of the observed operation parameters is modified by combining the observed operational parameter value and a weighted Euclidean distance between that value and the corresponding limit. In other words, the Euclidean distance for each operational parameter is multiplied by a weight and then combined with the observed parameter value. In one example, the weighted Euclidean distance is added to the observed operational parameter value to obtain the modified value.

[00018] Determining the smallest of the distances between the operational parameters and their corresponding limits is useful for setting the weighting of the distances for purposes of obtaining modified operational parameter values. In one example, the smallest distance is weighted the most significantly. That way, when the modified operational parameter values are used for determining the action index, the one that is closest to its corresponding limit has the most significant impact on the action index.

[00019] In one example, the Euclidean distance that is the smallest of the determined distances receives a weight of approximately one and all other Euclidean distances are weighted with a factor of zero.

[00020] The action index is computed in one example using the following relationship

action index = (x ^SD - x) P λ P ^T (x ^SD - x) ^T

[00021] wherein (x ^SD - x) is a vector matrix of distances between the modified observation values and the corresponding limits on those values; P is a principle component vector matrix containing the normal or expected operation values for each operational parameter; λ is a diagonal matrix of principle component Eigen values; P ^T is a transpose of the principle component vector matrix; and (x ^SD - x) ^T is a transpose of the vector matrix of the distances between the modified operational parameter values and the corresponding limits on those operation parameters. Further definition of the meaning of P and the topic of principal component analysis approaches can be found in Fault detection and diagnosis in industrial systems, ISBN 1-85233-327-8.

[00022] The matrix multiplication operation enabled by the P λ P ^T term in the equation serves to project the "x" vector onto the principal component subspace. It appropriately rotates and rescales the "x" vector in order to calculate a appropriately scaled action index consistent with the selection of the P principle components. In the absence of such scaling, the risk of multiple false alarms is significantly higher.

[00023] In Figure 2, at 68 a determination is made whether the action index is within an acceptable range. In one example, an appropriate lower limit is placed on the action index. For example, if the index has a value less than 0.1, that is an indication that some action should be taken based upon the current operating condition of the device 30. In an example where the action index is a shutdown index, when the action index has a value that is less than 0.1, that is an indication that the device 30 should be shutdown. The user interface 52 provides an indication indicating that the device 30 should be shutdown in one example. In another example, the processor 50 automatically shuts down the device 30 and the user interface 52 provides an indication that shutdown has occurred. The user interface 52 may also provide information regarding the operational parameter value, the action index value or a combination of them and any other information that would be useful to an individual for troubleshooting operation of the device 30, for example.

[0002 ] Figure 3 schematically shows the upper and lower limit of the action index value. The dashed line 70 indicates the higher action index limit (HAIL) and the solid line 72 indicates the lower action index limit (LAIL). A plurality of index values based on system performance are shown at 74. As long as the action index calculated is outside of the HAIL and LAIL limits, the system is judged as being in control. The actual values of the HAIL and LAIL are determined based on the balance between risk and false alarms associated with the system. Additionally, an escalating alert strategy might be employed depending on how closely the actual action index comes to HAIL and LAIL limits. Note that in this case, the action index value of 0 signifies arrival at the shutdown limit.

[00025] The preceding description is exemplary rather than limiting in nature.

Variations and modifications to the disclosed examples may become apparent to those skilled in the art that do not necessarily depart from the essence of this invention. The scope of legal protection given to this invention can only be determined by studying the following claims.