The present disclosure relates to the field of predictive maintenance for components of road vehicle.
Preventive maintenance of a road vehicle consists in performing scheduled maintenance so that a component can be replaced before it fails. Maintenance operations are generally scheduled according to component operating times, generally expressed in number of operating hours, calculated to allow for replacement of the part before it fails.
This method is not very efficient because it is based on nominal values established for a set of components and does not take the real state of the component into account. Thus, components are systematically replaced while some of them could still be used.
Moreover, in case of a sudden failure of the component's performance, the user can only notice failure of their component, without having been alerted of the component degradation.
Document US20180204393A1 proposes to predict a time of use or a distance that can be traveled by the vehicle before a maintenance operation of a component, such as an air filter, is necessary. This prediction is implemented when mean values and/or standard deviation of collected data representative of filter fouling are above thresholds. The prediction is based on a variation in the mean value and/or standard deviation of data representative of filter fouling. These thresholds can be predetermined or estimated according to parameters such as the distance traveled by the vehicle, a driving history of the vehicle, a calibration value obtained upon installing the filter for example.
This method has a drawback. Data that are acquired concerning filter fouling are not necessarily correlated to a context of use of the component, in addition to a context of use of the vehicle, which does not allow to have a great accuracy in the detection and in the prediction made.
The present disclosure improves this situation.
In particular, one purpose of the present disclosure is to predict future behavior, such as future failure, of the component with a better reliability.
There is provided a method for predictive maintenance of a component of a road vehicle, the method being implemented by at least one calculator connected to the component, the method comprising the following steps:
According to another aspect, there is provided a device for predictive maintenance of at least one component of a road vehicle, the device comprising:
According to another aspect, there is provided a computer program comprising instructions for implementing all or part of a method as defined herein when such program is executed by a processor. According to another aspect, there is provided a non-transitory, computer-readable recording medium on which such a program is recorded.
Characteristics set out in the following paragraphs can optionally be implemented. They can be implemented independently of each other or in combination with each other:
Further characteristics, details and advantages will become apparent from the detailed description below, and from an analysis of the appended drawings, in which:
The following drawings and description essentially contain definite elements. Therefore, they may not only serve to further the understanding of the present disclosure, but also contribute to its definition, where applicable.
The vehicle 10 comprises at least one component 11, at least one calculator 12, at least one electronic controller 13 of the component and a plurality of sensors 14. The calculator 12 is connected to the component 11, the electronic controller 13 and the plurality of sensors 14, for example, via a Controller Area Network (CAN) or FlexRAy type data communication bus. The calculator 12 can be configured to communicate directly with the remote server 20, when the calculator 12 includes adapted communication interfaces, or indirectly, via another calculator including adapted communication interfaces. In this case, data is transmitted between the two separate calculators via the aforementioned data communication bus. it will be noted that the electronic controller 13 of the component may be separate (as shown here) or may be integrated into the calculator 12.
In the example described here, the calculator 12 is an engine controller (electronic control unit) and comprises at least one processor, a memory and communication interfaces with various actuators and sensors of the vehicle and more particularly of the engine or motor as well as communication interfaces with the remote server 20. The electronic controller 13 of the component 12 is configured to implement a predetermined control law as a function of different parameters measured by different sensors specific but not limited to the component. The electronic controller 13 also comprises at least one processor, a memory and communication interfaces with the various sensors if necessary.
In addition, the electronic controller 13, upon controlling the component 11, obtains a measurement of a parameter involved in the control law and representative of wear of the component, also called the control parameter of the component. This value is advantageously periodically collected by the calculator 12 in order to know the course of the wear parameter of the component as a function of the number of kilometers traveled.
Some of the sensors 14 make it possible to acquire physical characteristics describing the dynamic behavior of the vehicle 10. In one exemplary embodiment, the sensors 14 allow acquisition of usage parameters of the vehicle such as speed, engine (or motor) torque, engine temperature, accelerator pedal position, vehicle acceleration or deceleration, steering wheel angle or steering angle for example.
Another part of the sensors 14 can be used to obtain usage parameters of the component considered. Advantageously, the usage parameters of the component are parameters having an influence on wear of the component. The usage parameters of the component can also be obtained by means of the electronic controller 13 when the latter is distinct from the calculator 12 and by means of the calculator 12, in this case the engine controller. These may include, for example, in the case of an automobile vehicle fuel injector, parameters such as fuel injection pressure, fuel temperature, fuel injection quantity and injection pump speed. Of course, other types of components and therefore other component usage parameters can be considered.
An odometer may also be included among the sensors 14 in order to determine the number of kilometers traveled by the vehicle.
The calculator 12 is configured to collect data relating to the use of the vehicle, the usage of the component and the course of wear of the component as a function of the number of kilometers traveled.
In particular, data relating to the vehicle usage and component usage may correspond to data relating to the use of predetermined combinations of component usage and vehicle usage parameters, respectively.
The collected data relating to the component usage may comprise frequencies of use of predetermined combinations of usage parameters of the component. As illustrated in
Advantageously, as illustrated in
The collected data related to the vehicle usage may comprise frequencies of use of predetermined combinations of usage parameters of the vehicle. As illustrated in
Furthermore, the data relating to the course of wear of the component correspond to data relating to the course of at least one wear parameter P3 of the component as a function of the number of kilometers traveled as schematically illustrated with reference to
As is visible in
Furthermore, as mentioned previously, the predictive maintenance device comprising the calculator 12 is able to communicate with the remote server 20. The remote server 20 is configured to collect, for a plurality of vehicles, data relating to the usage of the vehicle, the usage of the component and the course of wear of the component as a function of the number of kilometers traveled such as those previously described with reference to
The remote server 20 is also configured to determine a plurality of classes through implementation of an unsupervised classification algorithm, based on component usage and vehicle usage data collected as described above for a plurality of vehicles. The implementation of the unsupervised classification algorithm makes it possible to identify vehicles with similar component usage and vehicle usage profiles. The vehicle usage and component usage profiles are associated with the distribution of frequencies of use of different combinations CP1(1), . . . , CP1(N) and CP2(1), . . . , CP2(M) of component usage and vehicle usage parameters, respectively. The unsupervised classification can thus be done based on the frequencies of use of the combinations of component wear and vehicle wear parameters collected for each vehicle, for example.
The remote server 20 is also configured to determine reference data associated with each of the classes. The reference data relate to the course of the wear parameter of the component as a function of the number of kilometers traveled and are obtained from data on the course of the wear parameter of the component of each of the vehicles of one class.
Finally, the remote server 20 is configured to transmit, to the vehicle predictive maintenance device, data regarding the pre-established classes and associated reference data to be stored in the memory of the calculator 12.
Thus, the calculator 12 is able, from the data collected by the vehicle, i.e. data relating to the usage of the vehicle, the usage of the component and the course of wear of the component, and from the data relating to the plurality of pre-established classes stored in its memory, including the reference data associated with each of the pre-established classes, to deduce a future behavior of the component as described in more detail with reference to
In one embodiment, the vehicle 10 may also include a display (not represented) connected to the calculator 12 for displaying an alert message to the driver or a dedicated maintenance service, for example when the behavior of the component is abnormal in terms of wear with respect to wear data collected for vehicles with similar usage parameters of the component and the vehicle, in order to indicate that maintenance of this component is to be scheduled.
With reference to what has been previously described, the predictive maintenance method includes a step S100 of receiving and storing in memory data concerning the classes pre-established by the remote server 20 and reference data associated with each class.
The data concerning the pre-established classes allow identification of vehicles with similar component usage and vehicle usage profiles, and the reference data are obtained from the data on the course of the component wear parameter for each of the vehicles in the class considered. This reference data can be obtained by a statistical analysis of the course data collected for each of the vehicles of the class considered. Thus, the reference data may include, for example, positional characteristics (for example, mode, median, arithmetic mean, quantiles) and dispersion characteristics (for example, range, mean deviation, interquartile deviation, variance, standard deviation, and coefficient of variation). In the example of
The predictive maintenance method also includes a step S200 of collecting data relating to the usage of the vehicle, data relating to the usage of the component and data relating to the course of wear of the component as a function of the number of kilometers traveled as described previously with reference to
In particular, the number of times each predetermined combination CP2(1), . . . , CP2(M) of usage parameters of the vehicle and the number of times each predetermined combination CP1(1), . . . , CP1(N) of usage parameters of the component is used are collected. This can be made by incrementing a counter, specific to each combination of parameters, each time each of the usage parameters of the component or the vehicle belongs to a range of values of the combination considered. The course of the wear parameter, i.e. the control parameter representative of wear, is also collected as a function of the number of kilometers traveled. In one embodiment, in order to reduce the volume of data collected, a mean value of the wear parameter obtained over a predetermined distance traveled, for example, every 100 km, can be collected.
The predictive maintenance method also includes a step S300 of selecting, from the plurality of pre-established classes stored in memory, a class for which the data relating to component usage and the data relating to vehicle usage collected in step S200 are similar to those of the vehicles of the class selected.
More precisely, the selection is made so as to select a class for which the vehicle usage and component usage profiles are similar using data concerning the distribution of frequency of use of different combinations of component usage parameters CP1 and different combinations of vehicle usage parameters CP2 and data concerning the pre-established classes stored in memory. For example, the data concerning the pre-established classes may correspond to the position of a centroid for each pre-established class and the class with its closest centroid is selected.
Then, in a step S400, the data relating to the course of the vehicle wear parameter collected in step S200 and the reference data associated with the class selected in step S300 are compared.
In one exemplary embodiment, each value of the wear parameter collected in step S200 is compared to one or more threshold values defined based on reference data for the class considered. The threshold values S1 and S2 can be determined from position and/or dispersion characteristics previously described, for example.
In the example of
It is reminded that in the case of a fuel injector, the wear parameter, i.e. the control parameter representative of the component wear, can correspond to the injector closing time for example.
Alternatively or additionally, step S400 can include a sub-step of calculating a sliding average from the data relating to the course of wear of the component collected during step S200, here the mean value of the wear parameter collected every 100 km, and a sub-step of comparing the value of the sliding average with the average μ.
Next, in a step S450, it is determined whether the behavior of the vehicle component 10 is abnormal relative to components of vehicles of the class selected. For example, abnormal behavior of the component is detected when the number of times a representative value of the collected wear parameter is outside a reference range. In the example of
Then, during a step S500, a future behavior of the component is deduced therefrom.
In one embodiment, when abnormal behavior is detected in step S450, a future failure of the component is deduced therefrom. In step S600, an alert message is then issued to the driver or to a maintenance service in order to indicate that maintenance of this component is to be scheduled.
Step S200 can be implemented continuously, while steps S300, S400, S450 and S500 can be periodically implemented, depending on the number of kilometers traveled by the vehicle.
Step S200 of collecting a data set of the vehicle may thus further comprise collecting data relating to the number of kilometers traveled by the vehicle and step S300, as well as the successive steps, is implemented when a predetermined number of kilometers has been traveled by the vehicle for example when the number of kilometers traveled by the vehicle since the last iteration of these steps is a number less than 10,000 km for example.
Furthermore, in one embodiment, the step S300 of selecting a class from the plurality of classes pre-established by the remote server 20 and stored in memory can comprise:
Indeed, as illustrated with reference to
In the example of
Each subclass gathers vehicles with similar vehicle and component usage profiles for the predetermined range of kilometers traveled. Likewise, the reference data associated with these subclasses are calculated for the corresponding ranges of kilometers traveled.
In the example of
Thus, dividing the set of pre-established classes into different groups of subclasses according to the number of kilometers traveled by the vehicles allows the unsupervised classification algorithm to be implemented based on the currently available data. This is particularly advantageous when not all the vehicles used to collect the data have traveled long distances. It is then possible to obtain reference data for different ranges of kilometers traveled, with the reference data being more accurate for the first ranges of kilometers traveled. This is because a greater number of vehicles have traveled in the first ranges of kilometers, as illustrated in
Advantageously, the step S100 of receiving and storing in memory data concerning the pre-established classes and the reference data associated with each class can be regularly updated, in particular in order to refine the model, i.e. the classes and the associated reference data, when the number of kilometers traveled by the different vehicles increases.
Furthermore, it will be noted that the data collected by the calculator 12 can also be transmitted to the remote server in order to be taken into account by the unsupervised classification algorithm for the establishment of classes and for the calculation of the reference data associated with these classes. The establishment of the classes then includes a sub-step of selecting the data set collected by the different vehicles as a function of the number of kilometers traveled. Thus, the collected data used by the remote server 20 are advantageously transmitted when a predetermined number of kilometers has been traveled by each vehicle. In the example described here, every 10,000 km for example.
In one alternative embodiment, in step S450, if no abnormal behavior of the component is detected, it is deduced therefrom that the component behaves like all components of the class selected and, in step S500, a future behavior of the component can be deduced therefrom by predicting the value of the wear parameter of the component after the vehicle has traveled a determined additional distance. In particular, the value of at least one wear parameter of the component, for predetermined kilometers traveled, can be predicted using reference data obtained for a subclass established for a range of kilometers traveled comprising the predetermined kilometers traveled.
Advantageously, the subclass considered is selected so that it includes a majority of vehicles with a similar vehicle wear profile and component wear profile as the vehicle considered for the number of kilometers currently traveled by the vehicle upon implementing step S300. Step S100 may then include receiving and storing in memory data regarding the subclasses, and more particularly the reference data associated, to be used for predicting the value of the wear parameter at the predetermined kilometers traveled as a function of the subclass selected in step S320. For example, according to the previous example, if subclass C4,3 has been selected in step S320 and the value of the wear parameter is to be predicted for a distance of 100,000 km, subclass C4,3 will be associated with subclass C10,k established for vehicles that have been run between 0 and 100,000 km for which the maximum number of vehicles in subclass C4,3 have a vehicle usage and component usage profile similar to vehicles in class C10,k.
It is then possible, using the at least one wear parameter, to deduce therefrom an end-of-life time for the component, for example.
Number | Date | Country | Kind |
---|---|---|---|
FR2007414 | Jul 2020 | FR | national |
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/EP2021/066256 | 6/16/2021 | WO |