EP2339318B1 - Method for diagnosing a malfunction of a mechatronic system - Google Patents

Description

The present invention relates to a method for diagnosing a malfunction of a mechatronic system, a tool for carrying out this method, and a method for generating a signature table for implementing the method of diagnostic.

Electronics is used in the automotive industry for the implementation of a multitude of features. On-board electronic systems in vehicles are today structured around Electronic Control Units (ECUs). These UECs are connected to each other via computer networks, usually CAN networks or buses, and are connected to physical systems via sensors and actuators.

Physical components, whether electronic, mechanical, hydraulic or thermal, for example, are constantly monitored and controlled by the ECUs. The UECs thus form control systems, and the physical components the controlled systems.

The entire vehicle electronics allows the realization of different functions, the realization of a function that may require one or more UEC. This notion of function is important because it is in terms of a defect on this function that a user describes his problem to a garage mechanic. This is the first information the mechanic receives. A function is the set of resources (UEC, part of software, networks, electronic and mechanical components ...) necessary for the realization of a service.

The maintenance of the vehicles, and in particular the fault diagnosis in the garage, has thus become a real problem. The known diagnostic systems use methods based on the modeling of the physical components, which determine static values, to be compared with the measurement carried out during a test. These methods remain rather rudimentary so that today, in garages, the diagnosis of mechatronic systems based on these methods relies mainly on the knowledge of the mechanic or manually built diagnostic trees.

In particular, we know of the document DE 103 15 344 A1 a method of diagnosing a vehicle malfunction, implementing comparisons of an instantaneous test value to instantaneous reference values.

Document is known US 5,592,614 a system and method for identifying failures in an electrical circuit, for example a motor vehicle, also based on comparisons of an instantaneous test value to instantaneous reference values.

The document DE 10 2008 016 801 A1 also describes a fault diagnosis method comprising a comparison between an instantaneous parameter and reference values.

It is essential to give operators more and more efficient maintenance tools. The present invention aims at providing a high-performance method for diagnosing a mechatronic system and a tool adapted to the implementation of this method.

1. a) en réponse à l'observation d'un disfonctionnement d'une fonction dudit système, on choisit un premier parmi une pluralité de tests (Ti) prédéterminés de ladite fonction,
2. b) on met en oeuvre ledit premier test choisi (Ti) pour recueillir un signal temporel de test représentatif d'une première variable observable de ladite fonction,
3. c) on compare le signal temporel de test recueilli à un ensemble de signaux temporels de référence associés chacun à l'un de plusieurs modes de disfonctionnement (Fi) connus de ladite fonction,
4. d) on sélectionne, par ressemblance entre ledit signal temporel de test et lesdits signaux de référence, auquel desdits signaux de référence correspond le signal temporel de test recueilli,
5. e) on identifie ledit disfonctionnement observé si le signal de référence sélectionné est spécifique à un seul desdits modes de disfonctionnement (Fi) connus, et
6. f) dans le cas contraire, on répète les étapes a) à d) avec un nouveau test (Ti) jusqu'à l'obtention d'une séquence de signaux de référence sélectionnés qui, pour les tests mis en oeuvre successivement, est spécifique à un seul desdits modes de disfonctionnement connus.

For this purpose, the present invention proposes a method for diagnosing a mechatronic system, according to which:

1. a) in response to the observation of a malfunction of a function of said system, a first of a plurality of predetermined tests (Ti) of said function is selected,
2. b) implementing said first selected test (Ti) to collect a test time signal representative of a first observable variable of said function,
3. c) comparing the collected test time signal with a set of reference time signals each associated with one of several known malfunction modes (Fi) of said function,
4. d) selecting, by similarity between said test time signal and said reference signals, to which said reference signals correspond the collected test time signal,
5. e) identifying said observed malfunction if the selected reference signal is specific to only one of said known malfunction modes (Fi), and
6. f) in the opposite case, the steps a) to d) are repeated with a new test (Ti) until a sequence of selected reference signals is obtained which, for the tests implemented successively, is specific. to only one of said known modes of malfunction.

Thanks to the method according to the invention, the comparison made between a signal representing the evolution over time of an observable variable and a reference signal makes it possible, by the wealth of information that it brings into play, to report with a much greater relevance of a much greater number of malfunctions than the simple conventional comparison between the value taken at a given moment by this observable variable and a value, a range or a reference threshold. Indeed, such a comparison according to the state of the art relates to a static value and ignores any transient or periodic regime.

• on compare également ledit signal temporel de test à un signal de référence représentatif d'un mode de bon fonctionnement de ladite fonction ;
• le procédé comprend également une étape de sélection de l'un desdits signaux de référence en réponse à une observation directe par un opérateur de ladite variable observable (Oi).
• ladite variable observable (Oi) est un paramètre de ladite fonction ;
• ladite variable observable (Oi) est une grandeur physique mesurée sur un composant physique de ladite fonction;
• chaque test de ladite pluralité de tests (Ti) consiste à recueillir ledit signal de test dans une configuration donnée de ladite variable observable (Oi) ;
• ledit nouveau test porte sur ladite première variable observable (Oi) dans une configuration différente de celle dudit premier test ou sur une variable observable (Oi) différente de ladite première variable observable ;
• chacun desdits signaux de référence est représentatif d'une classe d'équivalence (M_ij) de signaux de signature individuelle regroupés par ressemblance et dont chacun est la signature de l'une desdites variables observables dans une configuration donnée et un mode de fonctionnement donné ;
• lesdits signaux de référence sont stockés dans une table dans laquelle l'un desdits tests est associé à chacune desdites variables observables dans chacune de ses configurations, dans ses différents modes de fonctionnement anticipés.

According to other characteristics of the diagnostic method according to the invention, considered alone or in combination:

• said test time signal is also compared to a reference signal representative of a mode of operation of said function;
• the method also comprises a step of selecting one of said reference signals in response to a direct observation by an operator of said observable variable (Oi).
• said observable variable (Oi) is a parameter of said function;
• said observable variable (Oi) is a physical quantity measured on a physical component of said function;
• each of said plurality of tests (Ti) comprises collecting said test signal in a given configuration of said observable variable (Oi);
• said new test relates to said first observable variable (Oi) in a configuration different from that of said first test or to an observable variable (Oi) different from said first observable variable;
• each of said reference signals is representative of an equivalence class (M _ij ) of individual signature signals grouped by similarity and each of which is the signature of one of said observable variables in a given configuration and a given mode of operation;
• said reference signals are stored in a table in which one of said tests is associated with each of said observable variables in each of its configurations, in its various anticipated operating modes.

• lesdits signaux temporels de référence sont stockés dans ladite mémoire (6),
• ledit calculateur est adapté pour mettre en oeuvre ledit test et acquérir ledit signal temporel de test à partir desdits moyens d'acquisition,
• ledit calculateur est adapté pour remplir l'une au moins des deux fonctions comprenant :
• l'affichage dudit premier signal temporel de test et ledit ensemble de signaux de référence associés audit test,
• la comparaison dudit premier signal temporel de test avec ledit ensemble de signaux de référence associés audit test et la détermination de ladite classe d'équivalence dudit premier signal temporel de test.

The invention also relates to a diagnostic tool, for the implementation of the aforementioned diagnostic method, comprising a computer, at least one memory, man-machine interface means comprising at least display means, and means acquisition of variables of said system connected to the computer, this tool being remarkable in that:

• said reference time signals are stored in said memory (6),
• said computer is adapted to implement said test and acquire said test time signal from said acquisition means,
• said calculator is adapted to fulfill at least one of the two functions comprising:
• displaying said first test time signal and said set of reference signals associated with said test,
• comparing said first test time signal with said set of reference signals associated with said test and determining said equivalence class of said first test time signal.

• ledit premier signal temporel de test et ledit ensemble de signaux de référence associés audit test sont affichés sur lesdits moyens d'affichage pour permettre visuellement ladite comparaison par un opérateur
• la comparaison dudit premier signal temporel de test avec ledit ensemble de signaux de référence associés audit test est faite visuellement par l'opérateur, par l'intermédiaire des moyens d'affichage de l'outil ou, avantageusement, le calculateur est programmé pour opérer la comparaison dudit premier signal temporel de test avec ledit ensemble de signaux de référence associés audit test ;
• les moyens d'acquisitions de variables comportent un circuit électronique de diagnostic doté d'une prise de diagnostic, et/ou un circuit électronique de mesure de grandeurs physiques.

According to other characteristics of the diagnostic tool according to the invention, considered alone or in combination:

• said first test time signal and said set of reference signals associated with said test are displayed on said display means to visually enable said comparison by an operator
• the comparison of said first test time signal with said set of reference signals associated with said test is done visually by the operator, via the display means of the tool or, advantageously, the computer is programmed to operate the comparison of said first test time signal with said set of reference signals associated with said test;
• the variable acquisition means comprise an electronic diagnostic circuit with a diagnostic socket, and / or an electronic circuit for measuring physical quantities.

• on modélise l'ensemble des ressources nécessaires à la réalisation d'une fonction du système mécatronique,
• on simule un ensemble de tests de ladite fonction en mode de bon fonctionnement, et dans les divers modes de disfonctionnement anticipés, pour toutes les configurations possibles des ressources de la fonction,
• on génère un signal de signature individuelle pour chacune des simulations, on compare ces signaux de signature individuelle entre eux,
• on regroupe, en fonction de leur ressemblance, lesdits signaux de signature individuelle en une pluralité de classes d'équivalence dont chacune est représentée par un signal de référence, et
• on construit ladite table en associant à chacun desdits modes de fonctionnement, pour chacun desdits tests, l'un desdits signaux de référence.

The invention further relates to a method for obtaining a signature table for the implementation of the aforementioned diagnostic method, this method being remarkable in that:

• we model all the resources necessary for the realization of a function of the mechatronic system,
• a set of tests of said function is simulated in the operating mode, and in the various malfunction modes anticipated, for all the possible configurations of the resources of the function,
• an individual signature signal is generated for each of the simulations, these individual signature signals are compared with each other,
• according to their resemblance, said individual signature signals are grouped into a plurality of equivalence classes each of which is represented by a reference signal, and
• said table is constructed by associating with each of said operating modes, for each of said tests, one of said reference signals.

• la modélisation à l'étape i) est réalisée hiérarchiquement, et décomposée jusqu'aux composants élémentaires de la fonction ;
• la modélisation à l'étape i) est une modélisation causale des systèmes de commande, et une modélisation acausale des systèmes commandés.
• la simulation à l'étape ii) est réalisée uniquement pour des variables observables par l'opérateur.
• le regroupement des signaux de signature individuelle à l'étape v) est opéré par un algorithme de ressemblance de signaux.
• un signal temporel de référence représentatif d'une première classe d'équivalence doit être distinct du signal temporel de référence représentatif d'une deuxième classe d'équivalence, afin de pouvoir être distingués visuellement par le garagiste.

According to other characteristics of the method for obtaining a signature table according to the invention, considered alone or in combination:

• the modeling in step i) is carried out hierarchically, and decomposed to the elementary components of the function;
• the modeling in step i) is a causal modeling of the control systems, and an acausal modeling of the controlled systems.
• the simulation in step ii) is performed only for variables observable by the operator.
• the grouping of the individual signature signals in step v) is performed by a signal similarity algorithm.
• a reference time signal representative of a first equivalence class must be distinct from the reference temporal signal representative of a second equivalence class, in order to be visually distinguishable by the mechanic.

• les figures 1, 1a, 1b sont des tables de signature selon un mode de réalisation de la présente invention ;
• les figures 2, 2a sont des représentations schématiques d'un outil de diagnostic selon l'invention ;
• la figure 3 est une représentation du modèle global de la fonction d'essuyage de la glace arrière d'un véhicule;
• la figure 4 est une représentation du modèle du calculateur habitacle de la fonction modélisé à la figure 3 ;
• la figure 5 est une représentation du modèle du logiciel du calculateur modélisé à la figure 4 ;
• la figure 6 est une représentation du modèle du moteur d'essuyage de la fonction modélisé à la figure 3 ;
• la figure 7 est un exemple d'une modélisation par automate d'un composant de la fonction modélisé à la figure 3 ;
• la figure 8 représente une matrice de résultats obtenue par une simulation de la fonction d'essuyage de la glace arrière d'un véhicule; et
• la figure 9 représente une matrice de signature de fautes dérivée de la matrice de résultats de la figure 8.

The invention will now be described in more detail, by way of example only and in a nonlimiting manner, with reference to the appended figures and in which:

• the Figures 1, 1a, 1b are signature tables according to an embodiment of the present invention;
• the Figures 2, 2a are schematic representations of a diagnostic tool according to the invention;
• the figure 3 is a representation of the overall model of the wiping function of the rear window of a vehicle;
• the figure 4 is a representation of the cabin calculator model of the function modeled at the figure 3 ;
• the figure 5 is a representation of the model of the calculator software modeled at the figure 4 ;
• the figure 6 is a representation of the wiper motor model of the function modeled at the figure 3 ;
• the figure 7 is an example of an automaton modeling of a component of the function modeled at the figure 3 ;
• the figure 8 represents a matrix of results obtained by a simulation of the wiping function of the rear window of a vehicle; and
• the figure 9 represents a signature matrix of faults derived from the result matrix of the figure 8 .

Mechatronic systems, such as motor vehicles, include electronic components associated with mechanical, and / or electromechanical, and / or hydraulic, and / or pneumatic components, etc., and more and more frequently to software components. The malfunction of such a mechatronic system is difficult to diagnose.

Since motor vehicles have a lot of electronics, when a user detects a malfunction of his vehicle and reports it to a mechanic, the latter proceeds to the diagnosis by means of an electronic diagnostic tool that can be connected to an electrical outlet. diagnosis of the vehicle to implement a diagnostic method.

The diagnostic method according to the invention will now be described in connection with a motor vehicle, but it should be understood that the invention is applicable to other mechatronic systems than motor vehicles.

In a motor vehicle, a function can be defined as the set of resources (UEC, part of software, networks, physical, electronic, mechanical, thermal components, etc.) needed to perform a service. These resources are variables of the function that can take different configurations. By example, being a switch, it can be in the closed or open state. The configuration of the function is the set consisting of the configuration of each resource.

The variables of the mechatronics system that are likely to be observed and tested by the garage mechanic are hereinafter referred to as observable or simply observable variables.

These observables can be of three categories.

The functional observables are the abstractions of a physical variable, directly apprehensible by the mechanic. For example, a speed of rotation of the wiper blades whose signal is a sinusoid can be translated into functional observation by a two-parameter information, namely the brushes are in motion, or the brushes are at the same time. stop.

The parameter-type observables are all the parameters that can be obtained directly by means of a diagnostic tool connected to a diagnostic socket of the vehicle.

Finally, observables of the physical quantities type are all the physical quantities that can be measured on a controlled system, that is to say on physical components. These physical measurements can for example be electrical, such as potential measurements, hydraulic measurements, pressure measurements, etc.

The diagnostic method according to the invention involves carrying out one or more tests on one or more observables of a function of which a malfunction has been detected. A test is the observation of an observable in a configuration of the given function. It is therefore a couple (observable, configuration).

The functional observations relating to the detected malfunction made by a user or a mechanic make it possible to determine which test, or series of tests, is most appropriate for the study of this malfunction. A test Ti is initially chosen by the mechanic. The choice of this test to be performed may be left to the initiative of the garage owner, or be guided by an algorithm determining the relevance of a particular test based on previous functional observations.

During this test Ti, a first test time signal corresponding to an observable O _i in a configuration of the sensor is collected either by means of a diagnostic test, or by measurement of physical quantities, or by means of a functional observation. given system C _i . This test must be performed for a certain time. Its duration must be of the order of magnitude of twice the maximum period of periodic phenomena that exist in the system, and greater than the duration of all the transients.

This first test time signal is then compared to a set of reference time signals present in a table or matrix of signatures M, represented to the figure 1 . These reference time signals, which can be generated by simulation as described below, also have durations of the order of magnitude of twice the maximum period of periodic phenomena that exist in the system in order to be able to compare signals having substantially the same duration.

The signature table M, or modality table, associates a signature with each of the modes of operation, namely the mode of good functioning and all the anticipated malfunction modes, the observables of the function studied in each of their different configurations. .

In the sample signature table shown in the figure 1 three modes of malfunction F1, F2, F3 have been shown, as well as the mode of operation F0, for three observables O1, 02, 03 which can be observed in a configuration of the function C0 and a configuration of the function C1. Each given mode of operation of a given observable in a given configuration is associated in the signature table of the figure 1 a signature constituted by a reference time signal M _ij .

Each reference signal M _ij is representative of an equivalence class, also called modality and also designated by the reference M _ij, which groups by resemblance a certain number of individual signature time signals. These individual signature signals are the true signatures of the different modes of operation of the observables in the different configurations of the function. These individual signature signals are obtained preferably by simulation of the different functions to be diagnosed, according to a method which will be described below.

In the example shown in figure 1 the same reference signal M ₁₁ constitutes the signature of the operating mode F0 for the observable O1 in a configuration C0 and that of the operating mode F2 for the observable O1 in the same configuration C0. In other words, an observable can have the same reference signal or equivalence class M _ij, for several different operating modes in a given configuration.

The set of suspicious operating modes at a given moment of the diagnostic session, that is to say the good functioning F0 or the malfunctions F1, F2 or F3 in the case of figure 1 . At the beginning of a diagnostic session, all modes of operation are suspect. This is the set of total ambiguity. For the figure 1 we have {F0, F1, F2, F3}. Then the tests will reduce this set of ambiguity. The process stops when the ambiguity set has only one element, even if this element is functioning properly.

The test time signal collected during the test performed by the garage mechanic is compared with the reference signals contained in the signature table M for this test Ti to determine by resemblance to which equivalence class belongs this test signal. This similarity determination can be made by an algorithm such as the Dynamic Time Warping (DTW) algorithm, or visually by the garage mechanic, as described hereinafter.

If for this test Ti, the selected reference signal corresponds to a class of equivalence specific to only one of the modes of operation of the observable in the tested configuration, said operating mode Fi is directly identified. For example in the figure 1 , if the selected test is T1 and the signal corresponds to the modality M12 the diagnosis stops because the set of ambiguity is only {F1}. In the same way if the signal corresponds to M13, then the ambiguity set is {F3}. And the diagnosis stops on this fault.

In the case where the equivalence class selected from the test signal is common to several modes of operation, a new test must be performed to determine which of these modes of operation is to be retained. More generally, when the ambiguity set contains more than one operating mode, other tests must be carried out. For example, as shown in figure 1a , if for the T1 test the reference signal selected corresponds to the equivalence class M ₁₁ , the operating modes can be F0 and F2. The ambiguity set is then {F0, F2}. The diagnosis must continue, it is then necessary to carry out a new test, different from the previous test.

This new test may relate to an observable different from that tested previously, or may relate to the same observable in a different configuration. The choice of this new test to be carried out can be left to the initiative of the mechanic, or be guided by an algorithm determining the relevance of this or that test based on the previous test and results found.

During this new test, a second test time signal T _i , representative of an observable of said function, is collected in a given configuration. This second time signal is then compared with the reference signals present in the signature table for the modes of operation of the observable in the configuration which is that of the test.

It is then determined by resemblance to which reference signal, therefore to which equivalence class, the second time signal corresponds. This resemblance determination can be made by an algorithm such as the DTW (Dynamic Time Warping) algorithm, or visually by the garage mechanic.

The lessons of the successive tests practiced are combined to determine whether the sequence of the equivalence classes M _ij selected successively is specific to only one of the operating modes F1. If this is the case, said operating mode Fi is identified.

For example, as shown in figure 1b if the T1 test of the observable O1 in the C0 configuration led to the selection of the equivalence class M _11, the ambiguity set is {F0, F2}. Then, if the T2 test of the same observable in the C1 configuration led to select the equivalence class M _{23, then} the ambiguity set is only {F2}. The sequence M _11, M ₂₃ is specific to the F2 malfunction mode of the observable 01. A malfunction is identified, the diagnosis can stop.

The diagnostic method described above can be implemented by means of a diagnostic tool such as that shown in FIG. figure 2 . Advantageously, this diagnostic tool 2 is in the form of a laptop, easily transportable by the garage to the vehicle on which it must intervene.

Referring also to the figure 2a , this diagnostic tool 2 comprises a computer 4 and at least one memory 6, in which the signature table M is stored. It further comprises man-machine interface means such as a keyboard 8 and display means 10. The diagnostic tool 2 also comprises means 12 for acquiring variables of the mechatronic system of the vehicle, namely a circuit diagnostic electronics 14 having a diagnostic socket (not shown) that can be connected to a diagnostic socket complementary to a motor vehicle and an electronic circuit for measuring physical quantities 16, for example a voltage or a hydraulic pressure also provided with appropriate sensor means (not shown) The diagnostic electronic circuit and the electronic measuring circuit are connected by connection means 18 appropriate to the computer 4. The electronic diagnostic circuit 14 and the electronic measurement circuit 16 may physically separated from the computer or in the form of integrated electronic cards.

In operation, once in possession of a functional description of a malfunction of a function of a vehicle, and if necessary after a first functional observation, the mechanic chooses a first test on the computer, makes the required connections and performs this first test on one of the observable variables making it possible to perform the offending function.

The computer 4 controls the acquisition means 12 to acquire the test time signal.

According to a first embodiment, the computer 4 is adapted to allow simultaneous display, on the display means 10, on the one hand of the test time signal collected, and on the other hand of all the reference associated with this test given in the signatures table. This simultaneous display allows the mechanic to visually compare these time signals and determine itself which reference signal has the greatest resemblance to the test signal collected. Because they correspond to equivalence classes, the reference signals are limited in number so that they are distinguished from each other by characteristic forms freed from certain singularities, which facilitates their comparison with the signal of test and ranking it in one of several equivalence classes.

In an alternative embodiment, the computer 4 is programmed to make the comparison of the first test time signal collected with the set of reference signals corresponding to said test in the signature table M, and to determine which equivalence class belongs to the first time signal of test.

If the comparison between the first time signal and the reference signals results in an equivalence class specific to one of the malfunction modes associated with the test, the diagnostic tool 2 signals the operator by the display means 10 what is the malfunction is identified. Otherwise the tool signals the operator that a new test must be performed.

According to the embodiment, the new test may be chosen manually by the operator or proposed by the computer 4 by a calculation of local or overall optimization of the choice of the test.

Indeed, the computer 4 can be advantageously programmed to determine what is the sequence of tests to be implemented to perform with a minimum cost the diagnosis of the system function. This determination can be made using an algorithm that uses the signature table as input, and also takes into account information such as intrinsic test costs, configuration change costs, instrumentation costs, or still the probabilities of malfunctions.

A local optimization of the choice of the test, that is to say, decide which is the next best test, can be set up and updated after each test carried out by the mechanic. A second solution may be the establishment of a graph forming a general tree of tests, and a search algorithm in this graph, for the overall optimization of the choice of tests.

The sequence of equivalence classes selected during the successive tests is updated and stored at least until a diagnosis is obtained. A list of future tests in order of relevance may be presented to the garage on the display means 10, without the garage owner being forced into his choice. If the recommended test is a functional observation to be made by the garage, it can use the keyboard 8 to enter the calculator 4 the result observed visually.

The method for constructing, for a given function, the signature table or matrix by modeling and simulation of this function will now be described. The construction of the signature table requires, first of all, a modeling of all the resources (UEC, networks, physical and electronic components) necessary for the realization of a function of the mechatronic system.

The principle of modeling consists of an approach based on models of the whole function. It breaks down into a controlled system and a control system.

Two types of modeling are implemented for a mechatronic system, a causal modeling of the control system and an acausal modeling of the controlled system.

The controlled system, be it electronic, mechanical, hydraulic or thermal, for example, is modeled in an acausal way by the equations of the physics of the corresponding domain. The acausal modeling takes into account two types of information, the structural knowledge of the controlled system, and the behavioral knowledge of each physical component of this controlled system, in the various modes of operation, both in a mode of good functioning and in a mode malfunction.

In acausal modeling, the components are interconnected with acausal ports. For example, for electronic components, these ports are the pins of the component. The ports are the sharing points of the flow and effort variables, for example, for the electronics, the intensity of the current I which crosses the port and the potential U at the port level.

The equations corresponding to each port are independent of the type of component, they are equations of structure. Thus, for an electronic port, these equations relate to the law of the nodes as well as to the equality of potentials.

• R1.n.i + R.p.i = 0 ; R2.p.i + R.n.i = 0 ; R.p.v = R1.n.v ; R.n.v = r2.p.v

For example, a resistor R will be connected to the other components via a positive port p and a negative port n. Two variables are present at each of these ports, the current (pi and ni) and the potential (pv and nv). The direction of the currents is positive towards the dipole. In addition, the negative port R1.n of component R1 is connected to Rp and the positive port R2.p of the component R2 is connected to Rn This structural knowledge makes it possible to write four equations of structure, independent of the type of component:

• R1.ni + Rpi = 0; R2.pi + Rni = 0; Rpv = R1.nv; Rnv = r2.pv

• R.p.i + R.n.i = 0 ; R.p.v - R.n.v = R.R * R.p.i

As discussed above, the modeling must also consider the behavioral knowledge of each component. In the case of a resistance in good working, there are two equations specific to the dipole, namely the conservation of the current and the law of Ohm.

• Rpi + Rni = 0; Rpv - Rnv = RR * Rpi

The equations of structure must be considered as equations shared between two dipoles, and the behavioral equations as equations specific to each dipole. In the present case, the modeling of a resistance results in four equations and four variables.

The two behavior equations correspond to a particular mode of operation, the proper functioning. It is also necessary to integrate the operation of the component for anticipated malfunction modes. At the same time, the structure equations remain the same for all modes of operation.

The behavior of each component is modeled by a hybrid two-level controller.

The first level corresponds to the mode of operation and the different modes of malfunction. The transitions between the operating mode and a malfunction mode represent the expected failures for the component, for example a gate relay. In the modeling described, only the permanent malfunctions are considered, and the transitions from a malfunction mode to the operating mode are therefore not taken into account. At each transition is associated a condition, for example for the relay the intensity that exceeds a threshold.

The second level will describe each mode of the first level by an automaton. Thus a controlled switch will have a mode of operation that will be described by a two state machine, passing and blocked. The transition condition between these two states is related to a control signal cmd. The model of the switch can thus be represented by the figure 7 . In this figure, the signal cmd is a causal signal that determines the sub mode to use, closed or open. A first malfunction Fi corresponds to the switch blocked in open circuit, a second malfunction Fii corresponds to the open switch in closed circuit.

The control system is modeled in a causal manner with, among other things, finite state machines for the software parts. In addition, it is taken into account that certain physical components of the controlled system may have a causal input.

• des capteurs en entrée, qui proviennent des composants acausaux, ou des paramètres de configuration,
• des commandes en sortie, qui vont sur les composants acausaux, tels que des commutateurs commandés par logiciel,
• des bus en entrée/sortie qui regroupent des signaux causaux d'un point à un autre, par exemple d'une Unité Electronique de Contrôle à une autre, ou encore
• des éléments de traitement des signaux simples sous forme de schémas blocs, ou des automates à états finis pour représenter les composants logiciels.

The causal modeling of the control part and causal behavior patterns distinguishes mainly four types of elements. These types of elements are either:

• input sensors, which come from acausal components, or configuration parameters,
• output commands, which go to acausal components, such as software-controlled switches,
• input / output buses that group causal signals from one point to another, for example from one Electronic Control Unit to another, or
• simple signal processing elements in the form of block diagrams, or finite state machines for representing the software components.

This causal modeling is an abstraction of the functioning of the Electronic Control Units. It is however useless to consider malfunctions on the software part, on which the car mechanic will not be able to intervene. Advantageously, only the input / output malfunctions of the causal signals are taken into account.

The observables of this causal modeling will be variables accessible by the garage mechanic via the diagnostic socket of the vehicle.

Thanks to these models one can simulate the whole of the function in mode of good functioning and in the various modes of malfunction anticipated. It should be noted that the simulation of a given function makes it possible to study a large number of variables, both in the control system and in the controlled system.

The implementation of the obtained model is performed on a hybrid and multi-physics simulator. This simulation can be based on the Modelica language, which can be found in the article P. Fritzson, "Principles of Object-Oriented Modeling and Simulation with Modelica" (Wiley-IEEE Computer Society Pr, 2003, ISBN 0471471631 ). Such a simulator may be for example the Dymola environment (Dynasim company, Dassault Systèmes).

The construction of the signature table for a function involves simulating this function in the operating mode, and in the various malfunction modes anticipated, for all the possible configurations of the observable N _o of the function.

If, for the function, N _f is the number of anticipated malfunctions in the physical components and the control part, and N _c is the number of configurations different, and knowing that for each configuration it is also necessary to simulate the good functioning, it will be carried out a number of different simulations equal to (N _f +1) x N _c .

An individual signature time signal is generated for each given observable Oi in each of its given configurations Ci and each of its anticipated operating modes Fi, namely a mode of operation and one or more modes of malfunction. For a simulation of a given test Ti of a given observable Oi in a given configuration Ci, the collected individual signature time signal can thus take the form of one of N _f + 1 different curves.

Each simulation must be performed for a certain duration. This duration must be of the order of magnitude of twice the maximum period of periodic phenomena that exist in the system, and greater than the duration of all the transients. In the example of wiping a vehicle rear window whose modeling will be described below, the duration is of the order of 20 seconds.

The individual signature time signals are compared with one another, advantageously by a signal resembling algorithm, and then these individual signature signals are grouped, according to their resemblance, into a plurality of equivalence classes or modalities each of which is represented by a reference signal. To do this, a signal comparison algorithm determines the resemblance between two signals, and groups them together if the similarity is proven.

For example, the DTW algorithm (Dynamic Time Warping) can be used. For information on this algorithm, for example, in the article of H. Sakoe and S. Chiba, "Dynamic Programming Algorithm Optimization for Spoken Word Recognition" (IEEE Trans. Acoust., Speech, Signal Processing, Vol 26, No. 1, pp 43-49, Feb 1978 ), or in the article of L. Rabiner and BH Juang, "Fundamentals of Speech Recognition" (Prentice Hall, Upper Saddle River, New Jersey 07458, 1993 ).

So, as shown in figure 1 for each test Ti of a given observable Oi in a given configuration Ci, a reference signal or equivalence class Mij is associated with each of the anticipated modes of operation F _i .

A reference time signal representative of a first equivalence class must be distinct from the reference temporal signal representative of a second equivalence class, in order to be visually distinguishable by the mechanic. It is the role of the comparison algorithm, which determines the equivalence classes translated into reference signals, to manage this notion of visual tolerance.

Preferably, the modeling and simulation of a given function focuses on observables, that is to say variables that the garage owner or other operator can observe.

The signature table is constructed by repeating the simulation operations described above for each of the mechatronics system functions that are likely to be diagnosed.

An example of modeling and simulation of a function for obtaining a table of signatures will be described below for the function of wiping the rear window of a motor vehicle.

The operating principle of this function is such that the user makes a request for rear wiping by means of a control lever, this request is transmitted to a computer associated with the passenger compartment of the vehicle via a network of CAN type information. The vehicle computer manages the intermittent wiping behavior by applying tension to the wiper motor until information is obtained that the wiper has returned to its idle position. It then suspends the application of the voltage for a few seconds before starting a cycle again.

The modeling of this function is done by hierarchization. On the set of figures that follow, it should be noted that the ports are squares for acausal signals and arrows for causal signals. A causal signal goes from one point to another point. There is a sense of information flow from where the arrow. An acausal signal corresponds to a variable in an algebraic equation. There is no clearly defined meaning (even for current intensities, for example, meanings are conventions, intensities can be positive or negative depending on the value of the components).

1. 1. des composants directement issus des librairies Modelica / Dymola
2. 2. des composants créés pour la modélisation des différents modes de fonctionnement. Ces composants dits de « diagnostic » sont regroupés dans une bibliothèque de composants de diagnostic. Ces composants sont des automates hybrides comme décrit précédemment et ont des modes de fautes.

The Figures 3, 4 , 5, 6 and 7 are graphic models made in the Dymola environment which is a Modelica development environment (Dynasim Corporation, Dassault Systèmes). The modeling components used in these models are of two types:

1. 1. Components directly from Modelica / Dymola bookstores
2. 2. Components created for modeling different modes of operation. These so-called "diagnostic" components are grouped into a library of diagnostic components. These components are hybrid automata as previously described and have fault modes.

When we say that the system is malfunctioning, Fi for example, this means that one of the diagnostic components is in a malfunction mode. A diagnostic component may have several modes of malfunction: for example a resistor may have 1 / R = 0 or R = 0 as a malfunction mode.

A first level of modeling is illustrated at the figure 3 . This figure shows the computer 31 associated with the control lever and the computer of the cockpit 32, interconnected by a CAN bus 33. The computer dedicated to the passenger compartment 32 is connected to a power source 34, beam connectors 35, wires 36 and the rear wiper means 37 consisting of the engine and the mechanical part of the wiping of the rear window.

The figure 4 illustrates a second level of modeling, namely the model of the cabin calculator 32 mentioned above. This figure shows the control software 41 of the rear wiping function, the power electronics elements 42, and the software 43 for testing the motor of the wiping means. The power electronics elements include among others a fuse 44, a protection resistor 45 and a switch 46.

The modeling of the purely software control part 41 of the cabin computer 32 is illustrated in the form of a finite state machine at the figure 5 , on which the squares represent states and the bars of the transitions between these states.

The first state E0 corresponds to a state where the rear window wiper function can not be activated because the vehicle contact has not been closed by means of the ignition key. The first transition T1 corresponds to the information that the ignition key has been rotated and the ignition of the vehicle closed. The next state E1 corresponds to the state of closure of the contact by the ignition key. The transition following T2 corresponds to a request for scanning the rear window at low speed made by the user of the vehicle. The state E2 is the one in which there is scanning of the rear window at low speed. The following transition T3 is representative of a "fixed stop" information provided by a sensor for detecting the angular position of the blade, to stop it when it has returned to its initial position.

The state E3 corresponds to the rest position of the blade after one go and a return of it on the window. The transition T4 represents the information provided by a timer to go to the scanning state after the lapse of a predetermined time, for example 12 seconds. The transition T4 leads to the input of the state E2 which is the scanning state.

The transition T5 connected to the output of the state E2 corresponds to a command provided by a timer in order to cut off the power supply of the electric scanning motor if the wiper blade has been in a sweeping state for more than 7 seconds. The state E4 is a state of protection of the motor in which the latter is no longer electrically powered. The transition T6 between the output of E4 and the input of E2 corresponds to a command delivered by a timer to trigger the scanning again after a determined time, for example 30 seconds, after the scanning engine has been found put in the state of protection E4. The transition T7 between the output of the state E4 and the input of the state E1 corresponds to a command from the user to end the scanning at low speed. Similarly, the transition T8 between the output of E2 and the input of E1 and the transition T9 between the output of E3 and the input of E1 also correspond to a command by the user to interrupt the scanning at low speed. Finally, the transition T10 between the output of E1 and the input of E0 corresponds to the breaking of the contact of the vehicle by the user by means of the ignition key.

The figure 6 is an example of global multi-acausal model of the electric motor part and linkage of the wiping function of the rear window, which incorporates electronic, electromechanical and mechanical components.

The electrical part of the motor is represented by an inductor 60 in parallel with a resistor 61 which represents the ohmic value of this inductance. The inductor 60 is shown connected in series by one of its terminals with a resistor 62 and a wire 63 connected to an acausal port P3 representing a pole of a power supply. The other terminal of the inductor 60 is connected to the port P2, which constitutes the other pole of the power source, via a component 64 constituting an "electrical-mechanical transformer", and a a fixed switch 65 controlled by a sensor (modeled by the component 68) which detects the return of the brush in the vicinity of its starting position.

When the wiper blade returns to its starting position, the sensor closes the switch 65 whose terminal opposite to the port P2 is connected to a port P1 via a wire 66. The current flow in the port loop P1, switch 65, wire 66, port P2 is detected by the cockpit computer 32 to interrupt the scanning as illustrated by the state E3 of the figure 5 .

The component 64, called "electrical-mechanical transformer" from the point of view of modeling, represents the transformation of a current that passes through a mechanical movement of continuous rotation. The component 67 connected to the component 64 illustrates the reciprocating transformation of the continuous rotational movement at the output of the component 64. This component 67 can be represented in the form of a two-state hybrid automaton and two transitions. Finally, the component 68 corresponds to a modeling of the aforementioned mechanical sensor which detects the angle of rotation of the wiper blade and closes the switch 65 when its angular position relative to the rest position becomes lower than a given threshold. .

Finally, the component 69 represents a model of the insulator between the wires leading to the ports P2 and P3 and corresponds in normal operation to an open circuit, and to a closed circuit in the event of a short circuit.

The model of the rear screen wiper function is thus carried out in a hierarchical manner, the function is successively decomposed down to the elementary components. For the sake of clarity, an illustration of the modeling in graphical form, not in the form of equations. But, of course, it is these equations available to those skilled in the art using modeling tools (Modelica) that make it possible to generate by simulation the temporal signals of individual signature of the observable variables.

Thus, the results matrix of the figure 8 illustrates the temporary signals of individual signature that are obtained by simulation, thanks to the model which has just been described above with regard to Figures 3 to 6 , for the rear windscreen wiper function.

The lines V1, V2 and V3 of the matrix correspond to signals which are present at the terminals V1, V2 and V3 at the input of the wiping motor ( figure 3 ), for different modes of operation which will be detailed below.

The "stop-fixed" line corresponds to a parameter of the passenger compartment computer relating to the circulation or non-circulation of a current in the loop P1, 66, 65, P2 as described above with respect to the figure 6 . During a test on a mechatronic system, this parameter will be read by the diagnostic tool using the diagnostic socket of the vehicle.

Similarly, the line "CMB_PV_AR" is also a parameter that reflects the fact that the cockpit computer 31 controls or not the electric motor 37 of the windscreen wiper system.

Finally, the line "brush speed" shows as a time signal the speed of the brush or brooms in a number of operating modes. During a diagnosis, this observable corresponds to a functional observation made by the mechanic and returned by him in the diagnostic tool in the form of information.

Column BF corresponds to the mode of operation.

The column I46.RC corresponds to a malfunction according to which the cockpit calculator sees the switch 46 of the figure 4 permanently closed, that is to say it remains glued.

The FIL 36 ₁ .CO column means that the wire 36 ₁ of the figure 3 is in open circuit. Similarly, the following column FIL 36 ₂ .CO corresponds to the malfunction according to which the wire 36 ₂ of the figure 3 is in open circuit.

The column 31.VR Cal corresponds to the situation in which the outgoing command of the computer 31 always remains in the state "true", that is to say a control state.

The matrix of the figure 8 shows at each intersection of a row and a column the individual signature time signal, generated by simulation using the appropriate equations, which is obtained for the relevant variable in the relevant operating mode.

It can be seen from this matrix that a number of time signals of individual signatures have kinships. As previously described, these signals are compared with each other and grouped into equivalence classes if these relationships are sufficient.

Thus, for example, for line V1, the signals of columns I46.RC and Cal 31.VR are sufficiently close to the point that they can be grouped together in a mod12 modality.

The process of comparison and classification in terms or classes of equivalences is carried out for the whole of the results matrix of the figure 8 and gives rise, at the end of the process, to obtaining the signature matrix of the faults of the figure 9 .

In this fault signature matrix, a modality mo_ii is associated with each of the rows and columns of the matrix, the modalities being identical in the same row when the temporal signals of individual signatures obtained in the result matrix of the matrix. figure 8 have been considered sufficiently similar.

It must be understood that at each of the designations mo_ii of the cells of the figure 9 of the matrix corresponds a reference signal which is the signal of the corresponding cell of the matrix of the figure 8 or a signal approaching in case of grouping by resemblance of several signals in the same equivalence class.

It is this signature matrix of the faults which makes it possible to build the table of signatures of the figure 1 , which is used for diagnosis as previously described.

Of course, the realism of the reference signals generated by modeling and simulation depends on the degree of abstraction of the modeling of the function. In other words, the higher this abstraction, the more the shape of these signals is moving away from that of the real signals that can be found on the mechatronic system. Thus, for example, the "brush speed" signal in the BF mode of the matrix of the figure 8 corresponds to a relatively idealized modeling of the function. In order to generate a simulated signal closer to physical reality, it would be appropriate to introduce into the model, as is well known to those skilled in the art, a certain number of components that make it possible to take into account additional phenomena of an electrical nature, mechanical or other such as, for example, inertia, friction, etc. of the set of brooms and linkage. The choice of the degree of abstraction is a question of compromise between the complexity of the model and the minimum of realism of the signals considered necessary to allow the operator to implement the diagnosis.

Claims (21)

1. Method for diagnosis of a malfunction in a mechatronic system, according to which:

   a) in response to the observation of a malfunction of a function of the said system, a first of a plurality of predetermined tests (Ti) of the said function is selected,

   b) the said first selected test (Ti) is implemented to gather a test time signal representative of a first observable variable of the said function,

   c) the gathered test time signal is compared with a group of reference time signals, each associated to one of a plurality of known malfunction modes (Fi) of the said function,

   d) the gathered test time signal is selected by similarity between the said test time signal and the said reference signals to which the said reference signals correspond,

   e) the said observed malfunction is identified if the selected reference signal is specific to one of the said known malfunction modes (Fi), and

   f) if this is not the case, steps a) to d) are repeated with a new test (Ti) until a sequence of selected reference signals is obtained which, for the tests implemented successively, is specific to one of the said known malfunction modes.
2. Method for diagnosis according to claim 1, according to which the said test time signal is also compared with a reference signal representative of a good operating mode of the said function.
3. Method for diagnosis according to one of claims 1 and 2, characterised in that the said observable variable (Oi) is a parameter of the said function.
4. Method for diagnosis according to one of claims 1 and 2, characterised in that the said observable variable (Oi) is a physical quantity measured on a physical component of the said function.
5. Method for diagnosis according to claim 1, characterised in that it also comprises a step for selection of one of the said reference signals in response to a direct observation by an operator of the said observable variable (Oi).
6. Method for diagnosis according to one of the preceding claims, in which each test of the said plurality of tests (Ti) consists of gathering the said test signal in a given configuration of the said observable variable (Oi).
7. Method for diagnosis according to claim 6, characterised in that the said new test concerns the said first observable variable (Oi) in a different configuration from that of the said first test.
8. Method for diagnosis according to claim 6, characterised in that the said new test concerns a different observable variable (Oi) from the said first observable variable.
9. Method for diagnosis according to one of the preceding claims, according to which each of the said reference signals is representative of an equivalence class (M_ij) of individual-signature signals grouped by similarity and each of which is the signature of one of the said observable variables in a given configuration and a given operating mode.
10. Method for diagnosis according to one of claims 6 to 9, according to which the said reference signals are stored in a table in which one of the said tests is associated with each of the said observable variables in each of its configurations, in its different anticipated operating modes.
11. Diagnostic tool (2) for implementing the method for diagnosis of a malfunction in a mechatronic system according to one of claims 1 to 10, comprising a computer (4), at least one memory (6), man-machine interface means including at least display means (10) and variables acquisition means (14, 16) of the said system linked to the computer, characterised in that:

    - the said reference time signals are stored in the said memory (6);

    - the said computer (4) is adjusted to implement the said test and acquire the said test time signal from the said acquisition means (14, 16); and

    - the said computer (4) is adjusted to fulfil at least one of the two functions comprising:

    - display of the said first test time signal and the said group of reference signals associated with the said test;

    - comparison of the said first test time signal with the said group of reference signals associated with the said test and determination of the said equivalence class of the said first test time signal.
12. Diagnostic tool (2) according to claim 11, characterised in that it is adjusted to display on the said display means the said first test time signal and the said group of reference signals associated with the said test and to visually allow the said comparison by an operator.
13. Diagnostic tool (2) according to claim 11, characterised in that the computer (4) is programmed to operate the comparison of the said first test time signal with the said group of reference signals associated with the said test.
14. Diagnostic tool (2) according to one of claims 11 to 13, characterised in that the variables acquisition means include an electronic diagnostic circuit (14) provided with a diagnostic connector.
15. Diagnostic tool (2) according to one of claims 11 to 13, characterised in that the variables acquisition means include an electronic circuit for measuring physical quantities (16).
16. Method for obtaining a signatures table (M) for implementing the method according to claim 10, characterised in that:

    i) the group of resources needed to produce a function of the mechatronic system is modelled,

    ii) a group of tests of the said function is simulated in good operating mode (F0) and in the various anticipated malfunction modes (F1, F2, F3 etc.) for all the possible configurations (Ci) of the function resources,

    iii) an individual-signature signal is generated for each of the simulations,

    iv) these individual-signature signals are compared with one another,

    v) the said individual-signature signals are grouped, as a function of their similarity, in a plurality of equivalence classes (M_ij), each of which is represented by a reference signal, and

    vi) the said table (M) is constructed by associating one of the said reference signals with each of the said operating modes (Fi), for each of the said tests (Ti).
17. Method for obtaining a signatures table according to claim 16, characterised in that the modelling in step i) is performed by hierarchy and broken down into the elementary components of the function.
18. Method for obtaining a signatures table according to one of claims 16 or 17, characterised in that the modelling in step i) is a causal modelling of the control systems and an acausal modelling of the control systems.
19. Method for obtaining a signatures table according to claim 16, characterised in that the simulation in step ii) is performed solely for the variables observable by the operator.
20. Method for obtaining a signatures table according to claim 16, characterised in that the grouping of the individual-signature signals in step v) is operated by a signal similarity algorithm.
21. Method for obtaining a signatures table according to the preceding claim, characterised in that a reference time signal representative of a first equivalence class must be distinct from the reference time signal representative of a second equivalence class in order to be able to be visually distinguished by the mechanic.

Images