Method and device for estimating the wear rate of components of a vehicle traveling on a road

Abstract

The invention relates to methods and devices for estimating wear rates of a set of at least one component of a vehicle (10) traveling on a first road (1001). To achieve this objective, the method comprises receiving first data representative of the characteristics of the first road from a first set of detection devices and identifying a road model associated with the first road using a road model identification model, the parameters of the road model being associated with this road model. The method also comprises receiving a set of control parameters of at least one on-board active system and estimating the wear rates of the set of at least one component of the vehicle from at least one wear prediction model associated with the set of components receiving as input the parameters of the road configuration and the set of control parameters. Figure 1

Description

Title of the invention: Method and device for estimating the wear rate of components of a vehicle traveling on a road Technical field

[0001] The present application generally relates to the field of a method for estimating the wear rate of the components of a first vehicle traveling on a road. The present application also relates to a method and a device for assisting a driver in controlling a vehicle, for example an autonomous or semi-autonomous vehicle. The present invention also relates to a method for learning a wear prediction model. Background

[0002] The purpose of this section is to introduce the reader to various aspects of the art, which may be related to various aspects of at least one exemplary embodiment of the present application that is described and / or claimed below. This discussion is deemed helpful to provide the reader with background information that will facilitate a better understanding of the various aspects of the present application.

[0003] Many modern vehicles are equipped with advanced driver assistance systems (AD AS). AD AS are passive and active systems designed to eliminate human error in driving all types of vehicles and / or to improve driving comfort, for example. ADAS systems use advanced technologies to assist the driver while driving, thereby improving performance. ADAS systems use a combination of sensor technologies to perceive the environment around a vehicle and then provide information to the driver or act on vehicle systems called active on-board systems by setting their parameters.

[0004] There are several levels of ADAS, acting in particular on the active on-board systems of the vehicle, for example adaptive cruise control, suspension control or power steering, by setting parameters to control certain actuators of each active on-board system.

[0005] ADAS systems embedded in a vehicle are powered by data obtained from one or more embedded sensors such as, for example, cameras, or by data received from a remote server. This data can in particular be used to detect obstacles in front of the vehicle, such as a pothole, in order to adapt the vehicle's suspension parameters as it passes over them.

[0006] The quality of the vehicle's response to driving conditions depends on its parameters, which can be set at the factory or by a user, for example at the start of a journey. However, the settings of the vehicle's active on-board systems are not always optimized, as driving conditions can change during a journey.

[0007] Furthermore, to improve vehicle reliability and passenger safety, vehicles require regular maintenance. Such operations follow a schedule published by a manufacturer, for example, in which it recommends how often certain mechanical parts should be replaced. These intervals are estimated based on average wear rates, for example, but do not necessarily take into account the actual lifespan of the vehicle. As a result, maintenance operations are carried out more frequently, sometimes more often than necessary, to avoid any risks, making vehicle maintenance expensive. In addition, it may happen that the vehicle suffers unusual damage due to certain driving conditions. Unfortunately, these difficult conditions can only be identified by the vehicle users, and this depends on their own sensitivity.It is therefore difficult to determine the actual state of wear of a vehicle's components without specific analysis. Summary

[0008] The following section presents a simplified summary of at least one exemplary embodiment to provide a basic understanding of certain aspects of the present application. This summary is not an exhaustive overview of an exemplary embodiment. It is not intended to identify critical elements of an embodiment. The following summary merely presents certain aspects of at least one of the exemplary embodiments in a simplified form as a prelude to the more detailed description provided elsewhere in the document.

[0009] According to a first aspect of the present application, there is provided a method for estimating the wear rate of a set of at least one component of a first vehicle traveling on a first road, the method comprising the following steps: - receiving first data representative of characteristics of the first road from a first set of detection devices of said first vehicle; - identifying a road model associated with said first road by feeding said first data into a road model identification model, said road model identification model being learned during a learning phase from learning data representative of characteristics of a set of secondary roads obtained from a set of secondary vehicles, the parameters of the road model being associated with said road model; - receiving a set of control parameters from at least one active on-board system; and - estimation of the wear rates of said set of at least one component of said first vehicle from at least one wear prediction model associated with said set of components receiving as input said road model parameters and said set of control parameters.

[0010] Such a method makes it possible to determine the wear rate or the fatigue state of the various components of the vehicle based on the road patterns encountered, these being representative of the actual driving conditions encountered by the vehicle. A relevant preventive maintenance operation can then be easily planned based on this wear rate.

[0011] In an exemplary embodiment, the wear prediction model is associated with a vehicle behavior model associated with the first vehicle, the vehicle behavior model receiving as input the parameters of the road model.

[0012] In another embodiment, the method further comprises receiving second data representative of intrinsic characteristics of said first vehicle (10), said set of control parameters being further determined from the behavior model of the vehicle receiving said second data as input.

[0013] In another embodiment, the set of control parameters is determined following a minimization function minimizing the wear rates of the set of at least one component of the first vehicle estimated from the wear prediction model receiving as input a first set of loads applied to at least one component. The first set of loads is calculated by the vehicle behavior model as a function of the parameters of the road model.

[0014] In another exemplary embodiment, the wear prediction model is learned in a training phase from training data representative of a first set of historical maintenance data obtained from a set of third-party vehicles, the first set of historical maintenance data being associated with a third set of loads applied to component types corresponding to at least one component of the first vehicle.

[0015] In another embodiment, the wear prediction model is fed by:- a computer-aided design file, and / or a finite element analysis result, and / or road test simulation results, and / or ground test results, and / or mechanical calculations, and / or slice fatigue damage calculations.

[0016] Another example of embodiment, the method further consists of predicting the maintenance operation as a function of the wear rates of all of at least one component of the first vehicle.

[0017] According to a second aspect of the present application, there is provided a device for estimating wear rates of a set of at least one component of a vehicle, wherein the device comprises a memory associated with at least one processor configured to implement the method in accordance with the first aspect of the present application.

[0018] According to a third aspect of the present application, there is provided a computer program product comprising instructions which, when the program is executed by one or more processors, cause one or more processors to execute a method according to the first aspect of the present application.

[0019] According to a fourth aspect of the present application, there is provided a non-transitory storage medium containing program code instructions for executing a method according to the first aspect of the present application.

[0020] The specific nature of at least one of the exemplary embodiments as well as other objects, advantages, features and uses of at least one of the exemplary embodiments will become apparent from the following description of examples taken in conjunction with the accompanying drawings. Brief description of the drawings

[0021] It will now be necessary to refer, by way of example, to the attached drawings, which show exemplary embodiments of the present application and in which:

[0022] [Fig.l] schematically illustrates a vehicle traveling on a road and its environment according to a particular and non-limiting embodiment of the present invention;

[0023] [Fig. 2] illustrates a flowchart of the different steps of a method for estimating the wear rate of a set of at least one component of the vehicle of [Fig. 1], according to a particular and non-limiting example of the present invention; and

[0024] [Fig. 3] illustrates a functional diagram of an example of a device configured to estimate wear rates of a set of at least one component of the vehicle of [Fig. 1], in accordance with at least one exemplary embodiment. Description of exemplary embodiments

[0025] At least one of the exemplary embodiments is described in more detail below with reference to the accompanying figures, in which examples of at least one of the exemplary embodiments are illustrated. An exemplary embodiment, however, may be embodied in numerous alternative forms and should not be construed as limited to the examples set forth herein. Accordingly, it should be understood that there is no intention to limit the exemplary embodiments to the particular forms disclosed. Rather, the disclosure is intended to cover all modifications, equivalents, and alternatives within the spirit and the scope of this application.

[0026] [Fig.l] schematically illustrates an environment 1 comprising a first vehicle 10, according to a particular and non-limiting embodiment of the present invention.

[0027] The first vehicle 10 corresponds for example to a vehicle with a thermal engine, with electric motor(s) or even to a hybrid vehicle with a thermal engine and one or more electric motors, for example a rechargeable hybrid vehicle. The first vehicle 10 therefore corresponds for example to a land vehicle, for example a car, a truck, a bus, or a motorcycle.

[0028] The first vehicle 10 may be equipped with various advanced driver assistance systems (ADAS) that are configured to assist the driver of the first vehicle 10 in controlling the first vehicle 10. The first vehicle 10 may be a semi-autonomous vehicle or an autonomous vehicle, for example, a vehicle traveling with an autonomy level that is for example greater than or equal to 2 in an autonomy range between 1 and 5, level 1 corresponding to a vehicle having a minimum autonomy level and level 5 corresponding to a vehicle having a maximum autonomy level, for example, a fully autonomous vehicle.

[0029] The first vehicle 10 is equipped with a road condition detection module 11, the objective of which is to process one or more time-stamped input signals, periodic or not, and not necessarily originally aligned in the time or frequency domain. A range of signals extends, but is not limited to, readings from the first set of detection devices comprising at least one sensor from: - inertial sensors, - wheel speed sensors, - steering wheel sensors, - radars, - lidars, - laser scanners, - the cameras, - acceleration sensors, - vibration sensors, - noise sensors, - rain sensors, - the external thermometer and - the weather station. Please note that this list is not exhaustive.

[0030] The road surface detection module 11 periodically analyzes the recent part of the available signals, preprocesses them as needed, for example by sampling, filtering, imputation of missing values ​​or deletion of values outliers, and aligns them, for example by interpolation, to obtain regularly spaced multidimensional data points, called time series, characterizing the evolution of the road characteristics attributable to the location over the distance recently traveled by the first vehicle 10. Each of these time series is, for example, also associated with the vehicle identification number, called VIN number, to allow the attribution of derived information to this vehicle at a later stage of processing. The length of the time series may depend on the ability of the vehicle to cache the recent part of the data and may range from a small value, for example the last minute, to the duration of the ignition cycle.

[0031] According to an exemplary embodiment, the first vehicle 10 is equipped with a communication unit, corresponding for example to a communication module of the TCU (Telematic Control Unit), of the ATB (“Autonomous Telematic Box”) or of the RFSB (“Radio Frequency Service Box”). Such a communication unit is advantageously connected to one or more antennas for, for example, transmitting and / or receiving data to and / or from a remote device, for example a data center 120 of the “cloud” 100, or a mobile communication device, such as a smartphone or a tablet, via a wireless connection, according to OTA (“Over The Air”) technology for example.The wireless communication between the first vehicle 10 and the remote device 120 is established for example via a network infrastructure comprising for example one or more relay antennas or base stations 110, or according to a direct communication mode (for example when the mobile communication device is at a distance from the first vehicle 10 allowing such a communication mode, for example at a distance of less than 10, 20 or 50 m). The wireless connection is based for example on one or more wireless communication protocols such as Bluetooth®, Wi-Fi® (based on IEEE 802.11), LTE (“Long-Term Evolution”), LTE-Advanced, 3GPP (“3rd Generation Partnership Project”) of the fourth or fifth generation, called 3GPP 4G or 5G, satellite communication such as Starlink®.

[0032] The first vehicle 10 is also equipped with at least one active on-board system. Such an active on-board system acts, for example, on the steering of the first vehicle, the brakes, the speed, the stabilizer bar, the shock absorbers, the suspension parameters or other parameters of the mechatronic system.

[0033] According to an exemplary embodiment, the first vehicle 10 is equipped with a second set of detection devices comprising, for example, a thermometer, a pressure sensor, a force sensor or a strain sensor with strain gauge. The second set of sensors makes it possible to know the intrinsic parameters of the vehicle or to capture the physical reactions of the vehicle or of its components in its environment.

[0034] According to an exemplary embodiment, the first vehicle 10 is equipped with a navigation and geolocation system, also called a GNSS (Geolocation and Navigation by a Satellite System) system, for example a satellite positioning system of the Global Positioning System (GPS) type. A geolocation signal is available to provide information on the momentary physical position of the vehicle.

[0035] According to another embodiment, the arbitrary combination of input signals is not assumed to be accompanied by a geolocation signal, but only by the vehicle identification, for example the VIN number. In this case, the vehicle signals, called non-geolocated signals, after being preprocessed and aligned, are converted into multidimensional data points that cannot be associated with any physical location. In the case of these non-geolocated signals from several vehicles, it is not possible to determine at which locations the roads coexist.

[0036] The on-board systems, such as the active on-board system, the communication unit, the ADAS and the road surface detection module 11 are each controlled by one or more computers, for example computers (electronic control unit). These computers form, for example, a multiplexed architecture for the realization of various services useful for the proper functioning of the first vehicle 10 and for assisting the driver and / or the passengers of the vehicle in the control of the first vehicle 10 via the control of the on-board systems of the first vehicle 10. The computers communicate and exchange data with each other via one or more computer buses, for example a communication bus such as CAN (Controller Area Network), CAN FD (Controller Area Network Flexible Data-Rate data bus type), FlexRay (according to the ISO 17458 standard), LIN (Local Interconnect Network) or Ethernet (according to the ISO / IEC 802-3 standard).

[0037] The first vehicle 10 travels on the first road 1001. The first road is, for example, an open road, such as a county road or a national road, or an off-road path or trail. The first road 1001 has certain characteristics associated with features. These characteristics include, for example, road roughness, road surface type, slope, oscillations of the road surface, but also turns, intersections, stops, regulations and the environment. The road characteristics are then classified into groups called road models, each road model comprising road model parameters.

[0038] A method for controlling an on-board active system of the first vehicle 10 and / or for estimating the wear rate of a set of at least one component of the first vehicle 10 traveling on the first road 1001 is advantageously implemented by one or more processors of one or more on-board computers of the first vehicle 10.

[0039] During a first operation, the computer receives first data representative of characteristics of the first road from the first set of sensors of the first vehicle 10.

[0040] In a second operation, a road model associated with said first road 1001 is identified by inputting said first data into a road model identification model, also called a road model identifier. The road model identification model has been learned in a learning phase from learning data representative of the characteristics of a set of secondary roads obtained from a set of secondary vehicles, the parameters of the road model being associated with each road model.

[0041] During the training phase, appropriate signals indicating various characteristics of the road are collected in a distributed manner, involving all of the second vehicles, and over a longer period of time. The data representative of these signals may be physically hosted in a data center 120 with several co-located data processing servers, or in a cloud-based or other suitable distributed infrastructure. The means and protocols for data transmission are not limited and may use any modern technology such as over-the-air (OTA) transmission over wireless or cellular networks.

[0042] For example, data collection takes from one to several months, and the data is collected from hundreds of vehicles of different categories, these vehicles constituting a third set of sensing devices similar to the first set of sensing devices. The aligned multidimensional signals, each from a vehicle of the set of second vehicles, must be transformed into a format that allows robust comparisons with each other. Therefore, a generic feature extraction operation is performed on each signal and the outputs are merged into a single data set. Vehicle identification is preserved, e.g., after a feature extraction step, the identity of the vehicle that emitted the source signal is still unambiguously assigned to each output record.For computational efficiency, the above transformation should be performed, and the results persist near the storage of the aforementioned signals, keeping the data at rest to avoid large data transmission overhead. Therefore, when enough data is collected, the assumptions of some road models in the analyzed signaling feature space representing named or unnamed classes. specific can be verified with more statistical significance.

[0043] The objective of the road pattern identification model is to repeatedly analyze the first data comprising feature-encoded data points in order to identify the coherent groups of road features or, alternatively, to confirm the existence of predetermined categories thereof. If no prior knowledge of the signatures of the road features is available, an unsupervised technique (clustering) is used. Otherwise, the task can be formulated as pattern matching / classification which will require additional training using independent samples of data with ground truth class labels assigned to them. In other words, the pattern identification model aims to find a function mapping an input multidimensional feature X into a cluster identifier or one of the predefined class labels.

[0044] If geolocation information, for example latitude, longitude and possibly altitude, is present among the received data, then the solution to the above clustering or classification problem will reflect to some extent the association of the road patterns with particular locations on the Earth's surface. The strength of this association will increase with the relative weight assigned to the location variables. In one possible embodiment of this invention, the location variables can be completely separated, and 2-dimensional clustering (or 3-dimensional clustering if altitude is also to be treated as a variable) can be performed only on these in order to first find only those locations visited by many vehicles.Then, the second-step clustering can be performed on the remaining variables, but only in the spatial neighborhood of the most common geographic locations previously found.

[0045] According to an exemplary embodiment, the a priori signatures of the road model categories are available, and it is possible to apply look-alike modeling in the case where the prior class signatures have a different representation than the actual features extracted from the available signals. This is possible, for example, in situations where the signatures are constructed under different conditions, for example in a controlled experimental environment or involving more sensors than those available in the vehicles of the set of second vehicles. This look-alike modeling is responsible for optimally assigning the extracted signal features to the closest model, but only taking into account the common data dimensions.For example, if the aggregated features extracted by the second vehicle set only include aggregated readings of the X and Y axis acceleration, but the . signature features also contain the aggregated Z-axis acceleration readings, pattern matching can still be performed in 2-dimensional feature space with the Z-axis acceleration data ignored.

[0046] According to another exemplary embodiment, assuming the availability of predefined pattern signatures and location variables in the data, a variant of the non-maximum suppression technique known from computer vision is proposed to integrate multiple model hypotheses densely generated around the actual locations of road conditions. The use of this technique implies that only the predictions with the maximum degree of confidence are initially selected and all other predictions having an overlap with the selected predictions greater than a threshold are removed. It is particularly useful to avoid the generation of redundant (almost perfectly co-located) models or too many models and efficiently achieves two objectives: clustering followed by classification.Non-maximum suppression is performed for each subset of predictions that produced the same model labels. If multiple model hypotheses were generated around the same physical location, this may be an indication of low overall confidence in the model identifier. Such predictions may therefore be excluded (or postponed until other signals sampled near that location are available) when there is no clearly dominant model label ultimately produced.

[0047] According to an exemplary embodiment, during a third operation, the first data is transmitted to the data center 120 or to the cloud 100 by the first communication unit of the vehicle 10. The data collected by the first vehicle 10 can then be used to enrich the database and update the identification model of the road model, or even refine the identification model of the road trajectory.

[0048] According to an exemplary embodiment, during a fourth operation, the position of the first vehicle 10 is acquired by the geolocation system, the first data then being associated with the current position of the first vehicle 10. According to this exemplary embodiment, the first vehicle 10 receives data from the data center 120 or from the cloud 100, the data comprising information guiding the identification of the road pattern of the second operation.

[0049] According to an exemplary embodiment, during a fifth operation, a second data representative of intrinsic characteristics of the first vehicle 10 is received. This data is acquired, for example, by the second set of detection devices. This second data is, for example, used as input for a behavior model of the vehicle. This second data com take for example data relating to the number of passengers or an on-board mass, to adjustment ranges of the actuators, to a type of tires, or even to the state of a part of the first vehicle 10. Optionally, this data is transmitted to the data center 120 or to the cloud 100 by the first communication unit of the vehicle 10.

[0050] According to a first embodiment, a set of control parameters is received, for example from the data center 120 or from the cloud 100, the set of control parameters being determined as a function of the parameters of the route associated with the first route 1001 and possibly as a function of the second data using an online vehicle behavior model or using tables, such as look-up tables (LUT).

[0051] According to a second embodiment, a set of control parameters is determined from a behavior model of the vehicle associated with said first vehicle 10 receiving as input said road model parameters associated with the first road 1001 and possibly as a function of the second data.

[0052] For the first and second embodiments, the vehicle behavior model is constructed, for example, from CAD files, physical calculations and / or multi-body dynamics simulations. Many models are constructed for different vehicle types and sizes. The vehicle behavior model used is the one associated with a vehicle whose characteristics are closest to the characteristics of the first vehicle 10. In particular, the vehicle behavior model is learned or improved using data collected via sensors embedded in the vehicles, for example via the second set of detection devices of the first vehicle 10 or similar embedded sensors in other vehicles.The vehicle behavior model is, for example, learned in a learning phase from learning data representative of a first set of physical parameter values ​​obtained from a set of third-party vehicles, said first set of physical parameter values ​​being associated with said road diagram and said set of control parameters.

[0053] According to an exemplary embodiment, the set of control parameters is determined in order to optimize the operating parameters for driving and / or steering the first vehicle 10. The set of control parameters is then determined following a maximization function maximizing the user comfort in the first vehicle 10, the user comfort being estimated from a user comfort prediction model receiving as input a second set of loads. This second set of loads is calculated by the vehicle behavior model as a function of the parameters of the road model. In a first example, the vehicle behavior model determines the vibration of the passenger compartment of the first vehicle 10 as a function of the speed of the first vehicle, a road profile and the stiffness of the suspension system of the first vehicle 10. In a second example, the vehicle behavior model determines an inertial force applied to the passengers as a function of the radius of curvature of a turn and the speed of the first vehicle 10. In a third example, the vehicle behavior model determines a torque to be applied by a driver to a steering wheel as a function of the radius of curvature of a turn, the speed of the first vehicle 10 and the power steering parameters.

[0054] According to an exemplary embodiment, the set of control parameters is further determined following a minimization function minimizing the wear rates of a set of components of the first vehicle 10, the wear rate being estimated from at least one wear prediction model associated with a set of components. The wear rate model receives as input a first set of loads to determine the wear rates, the first set of loads being calculated by the vehicle behavior model as a function of the parameters of the road model.

[0055] The set of components includes, for example, all or part of the following components: - suspension components such as the control arm, spring and / or shock absorber, - steering components such as the power steering system, steering shaft and / or steering rack, and / or - tires, and / or - braking system components such as brake pads and / or brake discs, and / or - engine components, and / or - chassis components, and / or - shock absorbers and / or silent blocks.

[0056] According to an exemplary embodiment, the wear prediction model is fed by: - a computer-aided design file, and / or - a finite element analysis result, and / or - road test simulation results, and / or - ground test results, and / or - mechanical calculations, and / or - calculations of fatigue damage in slices.

[0057] The wear prediction model has, for example, been learned in a learning phase from learning data representative of a first historical maintenance data set obtained from a set of third-party vehicles, the first historical maintenance data set is associated with a third set of loads applied to different types of components corresponding to at least one component of the first vehicle. For example, the wear prediction model is learned by matching the third set of loads or simulation results with the analysis of parts, parts recovered from vehicles from the set of third-party vehicles.

[0058] With the output of the pattern identification subsystem, e.g., each historical signal being assigned a distinct road pattern that it likely represents, the following method is proposed for estimating the propensity of vehicles to fail due to a potentially negative impact of certain road conditions or, conversely, the expected duration of their operation remaining fault-free.

[0059] For newly designed vehicles, there are no or few historical warranty claim records or maintenance reports from which knowledge could be obtained to guide any statistical learning approach for life estimation. In this case, computer-aided engineering (CAE), virtual simulation, and durability calculation combined with proving ground (PG) testing or road test simulation (RTS) are used to estimate the life of the vehicle system and / or components. For a given vehicle, with a given labeled road layout, along with given operational parameters for steering, braking, speed, and / or suspension, the shear damage attributable to the most recent time period or distance traveled can be calculated.Then, it can be correlated with the proving ground (PG) test results or road test simulation (RTS) results obtained over a longer period or mileage and using different models to determine the total accumulated durability damage. The total damage can finally be matched with the vehicle / component life expectancy and / or the time of the next scheduled maintenance can be estimated. In the wear prediction model, the vehicle conditions corresponding to the road patterns and vehicle operational parameters are included and considered. The material properties and the component / system mechanism can be obtained from physical tests, assumptions or simulations.

[0060] For vehicles in production for a sufficient period of time, the wear prediction model approach described above can be combined with a statistical learning approach to better estimate the expected durability of the vehicle. It is assumed that there is a record of historical warranty claims, maintenance reports, or other documentation of a similar nature when the facts of repairs or part replacements performed on vehicles of a given identity can be unambiguously linked to their root causes. From this record, a subset is selected where the root cause is related to structural durability over a predefined time period or distance traveled. Then, for the same list of vehicles, the algorithm constructs predictor variables based on the information if, when, and / or how often during the same analysis period each particular vehicle emitted signals that were assigned to the classes or groups found by the road pattern identifier. These variables can range from total counts for all road pattern classes / clusters, counts per class / cluster, frequencies of occurrence, and other simple aggregations to arbitrarily complex transformations and / or combinations thereof.Auxiliary variables can be used to describe vehicle characteristics, depending on their availability, provided that they contribute to greater accuracy of the estimated model.

[0061] Accordingly, one or more statistical models are trained to quantify the risk of failure of the vehicle system / component during its use on the roads. In other words, the following hazard functions are learned:

[0062] [Math.l] •> [Math.l] hj~p(y - failure|X, 0j)

[0063] where hj is the expected hazard function of the occurrence of the j-th type of failure in a predefined time window or distance to be traveled, X is the vector of predictive variables characterizing the historical state of a vehicle, with at least one dimension of this vector reflecting whether, when and how often it has emitted signals attributed to one of the recognized categories or groups of road models, and 0j is a vector of parameters of the model to be optimized.

[0064] Any state-of-the-art statistical learning approach can be used to estimate the above functions. In addition, the notion of conditional probability of failure of a vehicle system or component can have different, but related, formulations. For example, the function being optimized can integrate the temporal aspect of the prediction directly by modeling:

[0065] [Math.2] hj ^failure < Or [Math.2] ^J~p(-^failure “

[0066] where we estimate the probability that a failure of type j occurs before or after a determined time / mileage t.

[0067] Once the above functions are estimated, historical data from other vehicles, including the first vehicle 10, which were not seen in the model training stage, can be used to predict if or when a failure of each type considered may occur and thus trigger possible mitigation or prevention activities. In addition, for certain classes of learning algorithms, such as logistic regression, ensemble learning algorithms, or the use of model-independent techniques such as LIME or Shapley values, it is possible to quantify the contribution of the road model interaction variables to the model's decisions, both at the level of the entire population of vehicles being scored and at the level of individual vehicles.

[0068] During a sixth operation, the active on-board system is controlled according to all of the control parameters.

[0069] For example, for a first vehicle 10 with level 1 or higher autonomous driving capabilities or with an active suspension system, with signals carrying road information collected and their patterns identified, the autonomous driving system or the active suspension system can be used and adjusted to optimize driver handling / passenger comfort, increase safety, and reduce structural damage caused by different road conditions. The speed, steering wheel angle, acceleration, deceleration, preset driving mode, suspension height and anti-roll stiffness, shock absorber damping coefficient, and / or stabilizer bar can all be modified and optimized according to different road patterns.

[0070] If the previously known road patterns and the first data acquired by the first vehicle 10 are geolocated, and the road pattern categories are a priori known (they have predetermined semantics), taking into account the momentary location, the horizontal speed of a vehicle, possibly a map with the local road network, and the known precalculated locations of the road patterns detected from the pattern identification operation, it is possible to predict the time required to reach the nearest pattern in space. Therefore, the adjustment of the operating parameters can be triggered some time before the occurrence of the predicted condition. This adjustment can be guided by the response of the vehicle structure to particular conditions, obtained from the vehicle behavior model, as described below.

[0071] If both the already known road patterns and the first data acquired by the first vehicle 10 are geolocated, but the road pattern categories are not known a priori, it is possible to estimate the time required to reach the expected dangerous road section and take action in advance. However, since the nature of the condition is initially unknown, a generic and conservative adjustment of the vehicle operating parameters may be made, such as a reduction in speed, or an alert may be issued to the driver to minimize the expected negative impact on the vehicle as well as on the driver's comfort and safety.Over time, however, the impact of the "unnamed" road conditions can be empirically quantified and the rules for updating the respective operating parameters determined in a data-driven manner, for example by using reinforcement learning with a real-valued action space, and a reward function minimizing the negative impact of the parameter update on the durability of the selected components. Reinforcement learning with the real-valued action space is, for example, explained in US patent US10776692B2 on "Continuous control with deep reinforcement learning", written by T. P. Lillicrap, J. J. Hunt, A. Pritzel, N. M. O. Heess, T. Erez, Y. Tassa, D. Silver and D. P. Wierstra, and published in 2020.

[0072] If one does not geolocate the already known road patterns or the first data acquired by the first vehicle 10, it is generally impossible to establish a relationship between the physical location of a vehicle and the location of the patterns. In this case, the first vehicle 10 will only be able to detect the road condition associated with the pattern using its detection instruments if it is already very close or in direct contact with this condition. This means that the assignment of the pattern to one of the named or unnamed classes will also be delayed to the very last moment and the operating parameters of the vehicle can therefore only be adapted in advance when the available sensors allow it, an example of such sensors being the front cameras.Otherwise, the detected road condition will usually be classified and therefore the operating parameters adapted once the vehicle has already started to be affected, making the whole system reactive rather than predictive. For prolonged impacts, such as long sections of rough surface, it is always advantageous for the durability of the first vehicle.

[0073] In a seventh operation, the wear rates of all of at least one component of the first vehicle 10 are estimated from at least the wear prediction model associated with all of the components receiving as input the parameters of the road configuration and all of the control parameters.

[0074] During an eighth operation, a maintenance operation is planned according to wear rates of all at least one component of the first vehicle 10. Capable of reasoning on the expected time to failure, the process aims to facilitate decision-making when it comes to managing the cost of the first vehicle 10 maintenance or planning the supply of spare parts. An estimate of the vehicle condition model is used directly to drive the selection of vehicles to be checked.

[0075] For example, vehicles that have been assigned a failure risk or a cumulative damage level above a predefined threshold or those whose expected time to failure is below a certain threshold can be placed on the list of vehicles awaiting action.If the component most likely to malfunction is critical, for example if it could jeopardize the safety of the driver or passengers, or result in a significant negative experience or reputational damage, a customer will normally be asked to carry out a technical inspection of the first 10 vehicles, but this is subject to the customer's agreement. If the affected component is less critical, information about the predicted problem can be used in many other ways, for example to put the affected vehicle part or system on the to-do list for the next regular inspection or maintenance. In addition, the possible repair or replacement of a component is conditioned on confirmation of the existence of a defect or a general aging effect of the components after the first 10 vehicles have already been recalled for inspection.

[0076] According to an exemplary embodiment, the following operations are implemented based on an estimated value of the wear rate. In a ninth operation, a first user of the vehicle 10 is informed of a required inspection. For example, by using the in-vehicle infotainment (IVI) system to convey a message by displaying graphical content on a screen or by playing an audio message via in-vehicle speakers. The required inspection may include recalling the first vehicle 10 for examination at a workshop or requesting the user to perform a visual inspection.

[0077] In addition, there are other less tangible aspects of the above process that can nevertheless have a crucial effect on the parties involved. Among others, early prediction of a defect can prevent a vehicle breakdown in an unexpected road situation and thus minimize both driver stress and the costs associated with towing the vehicle to the garage. Secondly, if the condition of the vehicle poses a danger to the health or safety of the driver or passengers, mitigating the risks by recalling the vehicle early or during a regular check or maintenance can generate far more long-term benefits for all parties than the potential expenses it generates.

[0078] In order to better support the decision-making process related to business management of predictive maintenance of vehicles, a separate embodiment of the present invention comprises an optimization module that quantifies the impact of all the above factors and maps it on a common scale to finally find the best setting of the prediction model in the sense of maximizing the expected gain. In principle, the formula for the expected gain is designed to allow for as low a maintenance frequency as possible in order to avoid unplanned reactive maintenance, without incurring excessive costs associated with too much preventive maintenance.

[0079] The proposed new way to perform the optimization of the parameters of the failure and wear prediction model is to connect the desired cost / benefit factors directly to the loss function optimized by the wear prediction model in the training process, as well as to the scoring function used to judge the quality of the model on a particular data set. More specifically, assuming that the cost / benefit factors can each be assigned to true positives (TP), false positives (FP), true negatives (TN), or false negatives (FN), and expressed on a common scale, e.g., an amount of money lost / gained, the following formulas can be used:

[0080] [Math.3] , _y / ,(0 rVpp — L^y'pp •> [Math.3] ' WFP “ -l^pp, WFN~ '^i^FN'

[0081] a = ~~.

[0082] where b; and ci denote the unit amounts (per vehicle) associated with different benefit or cost factors and related to each true positive, true negative, false positive and false negative (respectively) produced by the wear prediction model, and a is a measure of the increased importance that the wear prediction model should give to reducing the number of false positive cases (vehicles with no inherent defect or not subject to aging, still classified as such) than to reducing the number of false negative cases (vehicles with the true problem but classified as no problem). With the above, the wear prediction model is trained using a weighted binary cross-entropy loss function:

[0083] [Math.4] L = E. - ay Iog(p.)-( 1 -v ,)log( 1 -p.)

[0084] where y, are the ground truth class labels (1 for a vehicle requiring a maintenance or 0 otherwise) and -.Pi are the probabilities of each training set vehicle requiring maintenance as evaluated by the model. Additionally, when evaluating each validation dataset during the routine model cross-validation procedure, the following scoring function can be used:

[0085] [Math.5] s=J)++

[0086] where:

[0087] _ (1 if the i-th instance is true positive, 1 t 0 otherwise

[0088] ^oO) fl if the i - th instance is true negative, 1 l 0 otherwise

[0089] ^o,i) _ f 1 if the i - th instance is false positive, t 0 otherwise

[0090] _ P if the i ' i^me instance is false negative. i 0 otherwise

[0091] With such a scoring function, the wear prediction model will automatically favor parameter sets that maximize the hypothetical seal gain and minimize the hypothetical seal loss, regardless of which factors contribute most and in which direction.

[0092] [Fig. 3] shows a block diagram illustrating an exemplary system or device 3 in which various exemplary aspects and embodiments are implemented.

[0093] The device 3 may be integrated as one or more devices comprising the various components described below. In various embodiments, the device 3 may be configured to implement one or more of the aspects described in the present application.

[0094] Examples of equipment that may constitute all or part of the device 3 include a calculator, an electronic control unit (ECU), a data processing device, or other communication devices. The elements of the device 3, alone or in combination, may be incorporated into a single integrated circuit (IC), multiple integrated circuits, and / or discrete components. For example, in at least one embodiment, the processing and encoder / decoder elements of the device 3 may be distributed across multiple integrated circuits and / or discrete components. In various embodiments, the device 3 may be communicatively coupled to other similar systems, or other electronic devices, via, for example, a communication bus 320 through dedicated input and / or output interfaces 32.

[0095] The device 3 may include at least one processor 30 configured to execute instructions loaded therein to implement, for example, the various aspects described in the present application. The processor 30 may include integrated memory, an input and output interface, and various other circuits known in the art. The device 3 may include at least one memory 31 (e.g., a volatile memory device and / or a non-volatile memory device). The device 3 may include a storage device, which may include non-volatile memory and / or volatile memory, including, but not limited to, electrically erasable programmable EEPROM, read-only memory (ROM), programmable read-only memory (PROM), random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), flash memory, magnetic disk drive, and / or optical disk drive.The storage device may include an internal storage device, an attached storage device, and / or a network accessible storage device, by way of non-limiting examples.

[0096] The device 3 may include an encoder / decoder module configured, for example, to process image data to provide encoded / decoded text data, and the encoder / decoder module may include its own processor and memory. The encoder / decoder module may represent one or more modules that may be included in a device to perform the encoding and / or decoding functions. As is known, a device may include either or both of the encoding and decoding modules. In addition, the encoder / decoder module may be implemented as a separate element or may be incorporated into the processor 30 as a combination of hardware and software, as is known to those skilled in the art.

[0097] The program code to be loaded onto the processor 30 or the encoder / decoder to perform the various aspects described in the present application may be stored in a storage device and then loaded onto the memory 31 for execution by the processor 30. In accordance with the various embodiments, one or more of the processors 30, the memory 31, the storage devices, and the encoder / decoder modules may store one or more of various elements during execution of the processes described in the present application. These stored elements may include, but are not limited to, data representative of the audio content, data representative of the video content, haptics-related data, a bitstream, matrices, variables, and intermediate or final results of processing equations, formulas, operations, and operational logic.

[0098] In several embodiments, memory within the processor 30 and / or the encoder / decoder module may be used to store instructions and provide working memory for processing that may be performed during processing, the encoding or decoding of data.

[0099] In other embodiments, however, memory external to the processing device (e.g., the processing device may be the processor) may be used for one or more of these functions. The external memory may be memory 31 and / or the storage device, e.g., dynamic volatile memory and / or non-volatile flash memory. In at least one embodiment, fast external dynamic volatile memory such as RAM may be used as working memory for processing the data.

[0100] [Fig.2] shows a functional diagram of the steps of a method 2 for estimating the wear rate of a set of at least one component of a vehicle, for example of the first vehicle 10 traveling on a first road, in accordance with at least one exemplary embodiment.

[0101] The method can, for example, be implemented by an on-board device of the first vehicle 10 or by the device 3 of [Fig.3].

[0102] In a first step 21, first data representative of characteristics of the first road are received from a first set of detection devices of the first vehicle 10.

[0103] In a second step, step 22 consists of identifying a road model associated with the first road 1001 by entering the first data in a road model identification model. The parameters of the road model are associated with the road model.

[0104] The road pattern identification model was learned in a learning phase from learning data representative of the characteristics of a set of secondary roads obtained from a set of secondary vehicles.

[0105] In a third step 23, a set of control parameters of at least one active on-board system is received as a function of the parameters of the route configuration.

[0106] In a fourth step 24, the wear rates of all of at least one component of the first vehicle 10 are estimated from at least one wear prediction model associated with all of the components receiving as input the parameters of the road configuration and all of the control parameters.

[0107] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting. It will also be understood that the terms "includes / includes" and / or "comprising / comprising", when used in this specification, may specify the presence of indicated features, elements and / or components, for example, but do not exclude the presence or addition of one or more other features, elements, components, and / or groups thereof. Furthermore, when an element is said to be "reactive" or "connected" to another element, it may be directly reactive or connected to the other element, or intermediate elements may be present. In contrast, when an element is said to be "directly reactive" or "directly connected" to another element, there are no intermediate elements present.

[0108] It should be noted that the use of any of the symbols / terms " / ", "and / or" and "at least one of", for example, in the cases of "A / B", "A and / or B" and "at least one of A and B", may be intended to encompass the selection of the first listed option (A) only, or the selection of the second listed option (B) only, or the selection of both options (A and B).As a further example, in the cases of "A, B and / or C" and "at least one of A, B and C", this formulation is intended to encompass the selection of the first listed option (A) only, or the selection of the second listed option (B) only, or the selection of the third listed option (C) only, or the selection of the first and second listed options (A and B) only, or the selection of the first and third listed options (A and C) only, or the selection of the second and third listed options (B and C) only, or the selection of all three options (A and B and C). This may be extended, as is clear to one of ordinary skill in this and related arts, to as many items as there are enumerations.

[0109] Different numerical values ​​may be used in the present application. The specific values ​​may be exemplary and the aspects described are not limited to these specific values.

[0110] It will be understood that, although the terms first, second, etc. may be used herein to describe various elements, such elements are not limited by these terms. These terms are used only to distinguish one element from another. For example, a first element could be called a second element, and, likewise, a second element could be called a first element without departing from the teachings of this application. No order is implied between a first element and a second element.

[0111] Reference to "an exemplary embodiment" or "an exemplary embodiment" or "an implementation" or "an implementation", as well as other variations thereof, is frequently used to indicate that a particular feature, structure, characteristic, etc. (described in connection with the embodiment / implementation) is included in at least one embodiment / implementation. Thus, occurrences of the phrase "in an exemplary embodiment" or "in an exemplary embodiment" or "in an implementation" or "in an implementation", as well as any other variations, appearing at various places in the present application do not necessarily refer to all to the same embodiment.

[0112] Similarly, reference herein to "in accordance with an exemplary embodiment / example / implementation" or "in an exemplary embodiment / example / implementation", as well as other variations thereof, is frequently used to indicate that a particular feature, structure, or characteristic (described in connection with the exemplary embodiment / example / implementation) may be included in at least one exemplary embodiment / example / implementation.Thus, occurrences of the phrase "in accordance with an exemplary embodiment / example / implementation" or "in an exemplary embodiment / example / implementation" at various locations in the specification do not necessarily all refer to the same exemplary embodiment / example / implementation, and separate or alternative exemplary embodiments / examples / implementations are not necessarily mutually exclusive of other exemplary embodiments / examples / implementations.

[0113] The reference numerals in the claims are given for illustrative purposes only and have no limiting effect on the scope of the claims. Although not explicitly described, the embodiments / examples and variations present may be used in any combination or subcombination.

[0114] A number of implementations have been described. Nevertheless, it will be understood that various modifications may be made. For example, elements of different implementations may be combined, supplemented, modified, or deleted to produce other implementations. In addition, a person of ordinary skill in the art will understand that other structures and processes may be substituted for those disclosed and that the resulting implementations will perform at least substantially the same function(s), in at least substantially the same manner, to achieve at least substantially the same results as the disclosed implementations. Accordingly, these and other implementations are contemplated in this application.

Claims

Claims

1. Method for estimating the wear rate of a set of at least one component of a first vehicle (10) traveling on a first road, the method comprising the following steps: - receiving (21) first data representative of characteristics of the first road from a first set of detection devices of said first vehicle (10); - identifying (22) a road model associated with said first road by feeding said first data into a road model identification model, said road model identification model being learned during a learning phase from learning data representative of characteristics of a set of secondary roads obtained from a set of secondary vehicles, the parameters of the road model being associated with said road model; - receiving (23) a set of control parameters of at least one active on-board system;and - estimation (24) of the wear rates of said set of at least one component of said first vehicle (10) from at least one wear prediction model associated with said set of components receiving as input said road model parameters and said set of control parameters.;

2. The method of claim 1, wherein said wear prediction model is associated with a vehicle behavior model associated with said first vehicle (10), said vehicle behavior model receiving as input said road model parameters.

3. Method according to claim 2, further comprising receiving second data representative of intrinsic characteristics of said first vehicle (10), said set of control parameters being further determined from the behavior model of the vehicle receiving said second data as input.

4. A method according to claim 2 or 3, wherein said set of control parameters is further determined following a minimization function minimizing the wear rates of said set of at least one component of said first vehicle (10) estimated from said wear prediction model receiving as input a first set of loads applied to said at least one component, said first set of loads being calculated by said vehicle behavior model based on said road model parameters.

5. The method of claim 4, wherein said wear prediction model is learned in a learning phase from learning data representative of a first set of historical maintenance data obtained from a set of third-party vehicles, said first set of historical maintenance data being associated with a third set of loads applied to component types corresponding to at least one component.

6. A method according to claim 5, wherein said wear prediction model is fed by: - a computer-aided design file, and / or - a finite element analysis result, and / or - road test simulation results, and / or - ground test results, and / or mechanical calculations, and / or - slice fatigue damage calculations.

7. A method according to any one of claims 1 to 6, further comprising predicting a maintenance operation based on said wear rates of said set of at least one component of said first vehicle (10).

8. A computer program product comprising program code instructions for executing the method of any one of claims 1 to 7, when said program is executed on a computer.

9. Device (3) for estimating the wear rate of a set of at least one component of a vehicle, in which said device comprises a memory (31) associated with at least one processor (30) configured to implement the method according to any one of claims 1 to 7

10. d / . Vehicle (10) comprising the device (3) according to claim 9.

Images