A METHOD FOR PREDICTING A REMAINING LIFETIME PARAMETER OF A COMPONENT

Abstract

A method for predicting a remaining lifetime parameter of a component installed in a system is provided, in particular of an engine component and/or a filter, the method comprising: repeatedly sensing at least one parameter of the system to obtain a history of data values;

fitting an aging pattern to the data values; and

determining a remaining lifetime parameter of the component from the aging pattern, wherein at least some data values are erased with time such that the fitting is based on a subset of the data values determined since an initialization of the algorithm, wherein data values from an initial phase are not erased but retained as anchor values for the fitting throughout the lifetime determination of the component.

Claims

1. A method for predicting a remaining lifetime parameter of a component installed in a system, the method comprising: repeatedly sensing at least one parameter of the system to obtain a history of data values; fitting an aging pattern to the data values; and determining the remaining lifetime parameter of the component from the aging pattern, wherein at least some data values are erased with time such that the fitting is based on a subset of the data values determined since an initialization of an algorithm, wherein data values from an initial phase are not erased but retained as anchor values for the fitting throughout a lifetime determination of the component.

2. The method of claim 1, wherein in addition to the anchor values, second data values from a most recent time period and/or third data values from at least one intermediate period are used in the fitting.

3. The method of claim 1, wherein the data are only erased when the algorithm is re-initialized.

4. A method for predicting a remaining lifetime parameter of a component installed in a system, the method comprising: repeatedly sensing at least one parameter of the system to obtain a history of data values; fitting an aging pattern to the data values; and determining the remaining lifetime parameter of the component from the aging pattern, wherein the aging pattern for determining the remaining lifetime parameter is automatically selected from a predefined set of different aging patterns.

5. The method of claim 4, wherein the method starts with a default aging pattern and/or automatically switches to a different aging pattern if the aging pattern does not fulfill a quality criterion.

6. A method for predicting a remaining lifetime parameter of a component installed in a system, the method comprising: repeatedly sensing at least one parameter of the system to obtain a history of data values; fitting an aging pattern to the data values; and determining the remaining lifetime parameter of the component from the aging pattern, wherein a change of the component is automatically detected by evaluating the data values.

7. The method of claim 6, wherein the change of the component is detected by monitoring a change in the data values.

8. The method of claim 6, wherein the remaining lifetime parameter determination is reset when a change in the component is detected.

9. The method of claim 8, wherein the reset comprises removing data values that were obtained before the change of the component from evaluation and/or resetting parameters of the aging pattern, wherein the reset comprises resetting an algorithm to an initial state and/or default values.

10. The method of claim 1, wherein during an initialization period, the determined remaining lifetime parameter is not output to a user.

11. The method of claim 1, wherein the fitting of the data values to the aging pattern comprises determining at least one parameter of the aging pattern.

12. The method of claim 1, wherein the fitting of the data values to the aging pattern comprises determining parameters of the aging pattern that minimize an error between the data values and values provided by the aging pattern.

13. The method of claim 1, wherein the component is a filter and the at least one parameter of the system is a pressure loss over the filter.

14. A system for predicting the remaining lifetime parameter of the component installed in the system, the system comprising a controller configured and programmed to determine the remaining lifetime parameter of the component using the method of claim 1.

15. The system of claim 14, the system being connected to a user interface and programmed to inform a user of a remaining lifetime of the component and/or output a servicing requirement of the component.

16. The method of claim 2, wherein the second and third data values are regularly replaced by new data values and may be saved and erased on a first in first out basis.

17. The method of claim 5, wherein the method starts with the default aging pattern and/or automatically switches to the different aging pattern if it does not provide a fit to the data values with a predefined accuracy.

18. The method of claim 7, wherein the change of the component is detected by monitoring a time derivate of the data values.

19. The method of claim 10, wherein an end of the initialization period is automatically determined by monitoring a change in the determined remaining lifetime parameter and a gradient of the determined remaining lifetime parameter.

20. The method of claim 11, wherein the aging pattern is defined as a formula, with coefficients of the formula being determined in the fitting.

Description

BRIEF DESCRIPTION OF THE FIGURES

[0081] The present disclosure will now describe in more detail on the basis of the following figures and embodiments.

[0082] The figures show:

[0083] FIG. 1A diagram showing steps to be performed for setting up an embodiment of the method of the present disclosure,

[0084] FIG. 2A flow diagram showing steps of an embodiment of the method of the present disclosure, and

[0085] FIG. 3A diagram showing an embodiment of the fitting of the aging pattern.

DETAILED DESCRIPTION

[0086] A more detailed description of an embodiment of the method of the present disclosure is described below with respect to FIGS. 1 to 3. The present disclosure, and in particular the embodiment described, is especially well suited to predict the remaining time of a filter before it is clogged.

1. Ageing Pattern Definition

[0087] The behaviour of the monitored component 10 over time, as described by the aging pattern, is the main input to the algorithm. FIG. 1 shows such an aging pattern 1 in the form of a curve 2 describing the dependency of a parameter value 5 of the system 20 on time 5.

[0088] If the pattern is not a priori known, it has to be determined before the implementation of the function.

[0089] The pattern can be obtained by learning from experimental data if they are available. If not, literature reviews or physical descriptions could help to define the pattern.

[0090] The algorithm does not require a full pattern description as pattern parameters are tuned during the component life. Nevertheless, it shall be possible to “inverse” the representation of the ageing pattern, i.e. to be able to find a time from any computed component condition.

[0091] In an embodiment, the pattern is defined by a mathematical expression. In alternative embodiments, the pattern is defined as a matrix, a model in the loop, etc.

[0092] If it is not possible to foresee a single pattern before the implementation of the function, it is also possible to predefine a list of typical patterns. The algorithm can then automatically select the most appropriate pattern.

[0093] FIG. 1 therefore shows the input necessary for setting up the method, i.e. for programming a controller to perform the method. The steps performed during the runtime of the inventive method on the controller are shown in FIG. 2.

2. Algorithm Initialisation 30

[0094] In addition to the predefined ageing pattern, the algorithm needs to sense and save the history of the monitored component. It requires a minimum of points to build and consolidate the computed ageing pattern, i.e. to fit the aging pattern to the sensed data values.

[0095] Especially when it is the first time that the algorithm is running or after each replacement of the component, the algorithm need to reset its parameters. This induces a period of initialisation 30. The duration of this period is depending on the pattern but it is possible to leave this period automatically by computing the gradient of the calculated remaining time and to wait for stabilised computations.

3. Component/System Condition Measurement 31

[0096] The system is sensed in step 31, i.e. a value of at least one system parameter is determined by measurement or from control values. The sensed parameter is the parameter described by the aging pattern or related to the parameter described by the aging pattern.

[0097] When the system is in a condition that allows the measurement, or when the system is in a referenced state (stabilised points for example), which is checked in step 32, it is the most appropriate time to sense the system. Therefore, if the check provides a positive result, all the necessary points (including the time) are saved in memory.

4. Data Conditioning 33

[0098] Depending on the frequency/amplitude of oscillations contained in the measured signals, performances of the algorithm might be affected. To overcome this situation, it might be necessary to filter the data in step 33.

[0099] Moreover, it may happen that the monitored component/system is replaced, without explicitly informing the algorithm of the change. This situation is handled by monitoring in step 34 whether a sudden change of the parameter/condition of the component/system occurred (by computing the derivative of the measured data for example). If so, the algorithm has to reset to its initial state and goes back to step 30.

5. Fitting Process 35

[0100] During the fitting process 35, the algorithm has to compute the parameters of the ageing pattern. For that purpose, the algorithm minimises the error between the ageing pattern and the measurements (cost function).

[0101] Many numerical methods exists in the literature to estimate the minimum of such a cost function. The choice has to be made wisely depending on the available computation power and memory.

[0102] Some methods require iterations to find the fitted parameters. For example, the Nelder-Mead algorithm is well suited for embedded system because of its development simplicity. Non-recursive non-linear regression algorithms also works. The later converges in one iteration and does not require any initial guess.

[0103] If the algorithm fails to find parameters within the allocated time and/or accuracy (maximum number of iterations reached and/or accuracy not accepted), the algorithm could automatically switch to another type of pattern (if a list of possible pattern is a priori defined and implemented). This may happen if the pattern is not known or if the system behaviour has changed (after a failure or a hardware change for example).

[0104] Once a new pattern is found that finds parameters within the allocated time and/or accuracy, this pattern may be used as the new default pattern for the next fitting procedure.

[0105] To improve the efficiency of the method, the disclosure may implement a data selection that will allow to get faster the trend of the ageing parameter. Whenever possible, the measurement history that is used for the fit should contain different type of data values named “tail points”, “middle points” and “head points”.

[0106] Illustration of the strategy is shown in FIG. 3 and are summarised below:

[0107] “Tail points” are first data values 40 which are measured within a relative short period of time after a new component is installed or during and/or after the initialisation of the algorithm. These points serve as an anchor for the ageing trend and are consequently not erased or replaced during the trend identification process.

[0108] “Head points” are second data values corresponding to the latest measured points and indicate the latest trend. They are especially useful when the component or the system could be differently solicited. They are stored in a “first-in first-out” memory.

[0109] “Middle points” are third data values corresponding to intermediary points in between “tail” and “head” points. They give the overall trend and are also stored in a “first-in first-out” memory.

[0110] In an embodiment, middle and head points may have different sampling rates. In particular, middle points may have a lower sampling rate (for example 1 hour) whereas head points have may have a faster sampling rate (example 10 minutes). Sampling rates have to be wisely selected depending on the component to monitor.

[0111] Further, in an embodiment, middle and tail points may have different sampling rates. In particular, middle points may have a lower sampling rate than tail points.

[0112] It is not mandatory to use all the above points, but numbers and data acquisition have to be wisely defined depending on the ageing trend.

6. Remaining Time Computation 36

[0113] As soon as new parameters are obtained from the fitting process 35, and/or at each activation of the monitoring function, the remaining time has to be updated from the inverse of the fitted model in step 36.

[0114] The end condition, which is the condition when the component has to be replaced or repaired, has to be known a priori. It could be for example a threshold 3, which has not to be exceeded, as shown in FIG. 1.

[0115] The end time t.sub.end corresponding to the end condition is deduced from the inverse of the fitted model, as indicated in FIG. 3. From that, the remaining time t.sub.rem can be predicted using the current time t.sub.our and the predicted end time tend.

7. Algorithm Ending 37

[0116] The algorithm ends in step 37 when the power supply is switched off. If possible, all the points, which have already been measured and stored to the memory, have to be saved in a non-volatile memory. This will avoid the requirement of a new initialisation phase and therefore significantly reduce the time required for the start-up for the next activation of the function.

[0117] The inventive method to predict a remaining lifetime of a system/component contains the following innovations:

[0118] 1. The method does not have to embed big data history of many similar previous systems, as it only requires the ageing pattern.

[0119] 2. If the ageing pattern is not known, the algorithm tests the most appropriate one from a predefined list of most probable or typical ageing patterns.

[0120] 3. Depending on the real use of the system, the ageing pattern parameters evolve at each iteration: the trend is built based on the current condition but also using strategic key points (named tail, head, and middle points). Using the proposed key points also reduces the running time before getting the trend of the pattern.

[0121] 4. The algorithm is autonomous, as it does not require any information from the user. If the monitored system/component is replaced, the algorithm detects it automatically and reset its parameters.

[0122] The method can be implemented on a controller of the system or on a separate controller receiving the sensed parameter from the controller of the system.

[0123] The following claims particularly point out certain combinations and sub-combinations regarded as novel and non-obvious. These claims may refer to “an” element or “a first” element or the equivalent thereof. Such claims should be understood to include incorporation of one or more such elements, neither requiring nor excluding two or more such elements. Other combinations and sub-combinations of the disclosed features, functions, elements, and/or properties may be claimed through amendment of the present claims or through presentation of new claims in this or a related application. Such claims, whether broader, narrower, equal, or different in scope to the original claims, also are regarded as included within the subject matter of the present disclosure.