DE102007010978A1 - Electrical system's diagnosis supporting device for use in motor vehicle, has evaluation unit to produce list of incorrect components that are sorted based on dependence value, and output/supply unit to display or provide list - Google Patents

Abstract

The device has a node assigned to components and error events. A bayes network defines a cause-effect relationship between the components and the error events. A determining unit determines an indication of one of the error events. An evaluation unit evaluates the bayes network and determines a dependence value for defectiveness of each component. The evaluation unit produces a list of incorrect components that are sorted based on the dependence value. An output/supply unit displays or provides the list. An independent claim is also included for a method for supporting the diagnosis of an electrical system of a motor vehicle.

Description

The The invention relates to a method and apparatus for assisting a diagnosis of an electrical system, in particular an electrical System of a motor vehicle, of interconnected components.

electrical Systems include a variety of interconnected components. These components include earth points, wires, lamps, actuators, Sensors, complex control devices, etc. Especially complex Components are often self-diagnostic today. This is understood below that these components interference recognize and communicate independently. In the field of automotive engineering, the term on-board diagnostic interfaces and communication protocols defined over those on complex components, also referred to as controllers can be accessed to store event entries so-called error event entries (DTC) Trouble Code) include readings and other information. This information is provided by diagnostic routines within the complex Components determined. There are different ones for this Method. So usually all read by sensors Analog signals are monitored to see if they are within a permissible range. In many cases also plausibility of sensor signals against each other. For more complex mechatronic systems, as z. B. for controlling a combustion process In addition, model-based engines are used in an engine Diagnostic method used.

Under A term off-board diagnostic system will be a class of systems if necessary, for example in a workshop, be connected to an electrical system and from outside via a data connection to the components of the electrical system access. These off-board diagnostic systems have the option on, the event store entries of all existing components of the electrical system and in one for one Mechanic to edit and output readable form. Often such systems have additional diagnostic functions, which allow for further diagnostic routines of individual To access components. This will be the opportunity created to interrogate individual readings from sensors, actuators (Actuators) to target and / or software in one Component is implemented, update.

The Reading out error event entries is common first clues where localized a source of error of the electrical system could be. A unique mapping of error event entries however, only a few faulty components are defective Cases possible. Often one can Variety of components of the electrical system for the Occurrence of an error event entry in the sense responsible Be that error event entry in one of the complex components is set if one of these components is defective, d. H. Not okay. To locate a fault as quickly as possible and efficiently, it is desirable to locate those components most likely due to their defectiveness for the occurrence of the corresponding Error event leading to the error event entry Has.

Out DE 10 2005 015 664 A1 a diagnostic system is known for determining a weighted list of potentially defective components from vehicle data and customer information. The invention relates to a computer-aided diagnosis system that generates a weighted list of possibly defective motor vehicle components with the aid of a diagnostic program from vehicle data and customer information. The identification of the possible error candidates takes place via an evaluation of a rule table reflecting the diagnostic knowledge. The additional evaluation of vehicle functions possibly also affected by the error candidates extends the search space. A service technician can restrict troubleshooting to selected fault codes or functions by setting a focus within the determined search space. Then only the possible candidates relevant to the selected error codes or functions are considered. The error candidates belonging to this focus set are weighted by offsetting several error probabilities for error codes, components and affected functions. Conditional probabilities that indicate the certainty with which a defective component or a candidate causes an error code, a malfunction, or a fault image must be provided for this purpose. In addition, relative error weights are needed, which are selected, for example, between 1 and 100 and represent a relative failure parameter. Alternatively, still known error images, which are coupled error codes that always occur together, can be used for the billing. To be able to use this known system efficiently, it is necessary that a very extensive diagnostic knowledge is determined for each individual vehicle. Especially in modern motor vehicles, which are produced with a very large number of variants in series, the creation of the necessary rule base for each motor vehicle is individually almost impossible, as far as errors in individual simple components such as cables, plugs, mass points, Splices, etc. should be determined next to the errors in more complex components such as ECUs.

From the DE 102 22 072 A1 An automatic diagnostic method for a dynamic technical system is known, which responds to an excitation with a dynamic system output behavior. The known method comprises the method steps of determining coefficients characterizing the system output behavior from at least one measurement curve representing the system output behavior, determining discrete system states from the coefficients, assigning the discrete system states as input data of a probability network and processing the input data in the probability network for calculating a diagnosis result, wherein at least a, the excitation of the system characterizing parameter is assigned as input to the probability network and enters into the calculation of the diagnosis result. Furthermore, a corresponding device for carrying out the method in the DE 102 22 072 A1 specified. The proposed probability network is a Bayesian network. The proposed solution assumes that for the diagnosis of the electrical system all causalities causing a fault are at least probabilistically completely known and the complex diagnostic knowledge is thus integrated into the probability network. Furthermore, the known method is only applicable to such error events associated with generating a trace.

A automated support of the diagnosis of electrical Systems made in a large number of variants are currently not satisfactorily possible. Desirable It is to provide a method and a device which has a weighted list of possible error candidates. Based This list is then a service technician or another Automated method or device possible to initiate or propose suitable further diagnostic steps or carry out a limitation of the faulty Enable component.

Of the The invention is therefore based on the technical problem of a method and a device for assisting the diagnosis of electrical system, in particular an electrical system of a Motor vehicle, to create a list with possibly faulty components sorted according to determined trustworthiness spend in a flaw and automate easily to a variety of variants generated in variants of the electrical system are applicable.

The technical problem is achieved by a device having the features of claim 1 and a Method solved with the features of claim 11. Advantageous embodiments of the invention will become apparent from the Dependent claims.

The Invention is designed so that only in production of the electrical system arising knowledge about the individual Variants is used to make it a probability-based Network to generate, with the help of which observed error events are evaluable to possibly a desired list faulty components weighted according to trustworthiness for a Incorrectness of the individual components.

• – eine Netzerzeugungseinheit zum Erzeugen eines Bayesschen Netzes
• – aus einem Strukturdatensatz, der Informationen über die Komponenten und deren elektrische Verschaltung umfasst;
• – aus Ausfallwahrscheinlichkeitsinformationen, die es ermöglichen, jeder der Komponenten eine Ausfallwahrscheinlichkeit zuzuordnen und
• – aus Fehlerereignisinformationen, die mögliche Fehlerereignisse und den Fehlerereignissen zugeordnete Zuweisungsangaben umfassen, wobei für jedes der Fehlerereignisse zumindest eine Zuweisungsangabe existiert und eine Zuweisungsangabe ein Fehlerereignis einer der Komponenten zuordnet, wobei jeder Komponente und jedem Fehlerereignis ein Knoten zugeordnet ist und das Bayessche Netz eine Ursache-Wirkungs-Beziehung zwischen den Komponenten und den Fehlerereignissen abbildet;
• – eine Erfassungseinheit zum Erfassen mindestens einer Fehlerereignisangabe, die ein Auftreten eines entsprechenden der Fehlereignisse angibt;
• – eine Auswerteeinheit zum Auswerten des Bayesschen Netzes und Ermitteln eines Vertrauenswertes für eine Fehlerhaftigkeit jeder der Komponenten und Erzeugen einer Liste von möglichen fehlerhaften Komponenten sortiert nach den bei der Auswertung ermittelten Vertrauenswerten; und
• – eine Ausgabe- und Bereitstellungseinheit zum Ausgeben oder Bereitstellen der Liste.

In particular, a device is proposed for supporting a diagnosis of an electrical system, in particular of an electrical system of a motor vehicle, of interconnected components, which comprises:

• A network generation unit for generating a Bayesian network
• - From a structural data set that includes information about the components and their electrical wiring;
• - probability of default information that allows each of the components to be assigned a probability of default and
• Assignment information associated with possible error events and error events, wherein at least one assignment indication exists for each of the error events, and an assignment indication assigns an error event to one of the components, each node and each error event being assigned a node and the Bayesian network having a cause Depicts the effect relationship between the components and the error events;
• A detection unit for detecting at least one error event indication indicating an occurrence of a corresponding one of the failure events;
• - An evaluation unit for evaluating the Bayesian network and determining a confidence value for a defectiveness of each of the components and generating a list of possible faulty components sorted according to the confidence values determined in the evaluation; and
• An output and provision unit for outputting or providing the list.

It is therefore sufficient to know those structural data which indicates an interconnection of the components in a simple circuit diagram. In addition, failure probabilities are needed that allow the individual components each to assign a default probability. This information is known from the quality assurance of the installed components or the production and may optionally be replaced and / or supplemented by experience gained from the operation of the electrical system or other variants of the electrical system. Furthermore, the determination of the weighted list of error candidates requires error event information indicating the possible error events and indicating at which component the error events may occur. In this case, components are considered only on a basis in order (OK) and not in order (NOK). An internal structure of a complex component, such as a control unit, is not considered in detail. However, if a control unit has several connections, these are regarded as separate components or their connection points as separate components. Observable states of the electrical system or of individual components are regarded as error events. Observable in this sense are readable error memory entries (DTC entries). For these error event entries determined by an on-board diagnosis, it is specified in the production of the complex component which or which electrical inputs / outputs may or may be responsible for the occurrence of the corresponding error event. This means that the inputs and outputs of the complex component are known, which are related to a fault event when connected to another faulty component. If an error event is independent of all outputs and inputs of the corresponding component, an occurrence of this error event is an indication that this component is faulty. In these rarely occurring cases, error localization to a component is thus easily possible anyway. In order to be able to determine confidence values for an error for the individual components of the electrical system, it is proposed to create a Bayesian network from the information just mentioned, all of which are known from the production of the electrical system. Using simple electrical engineering knowledge, the structural data set can be used to determine how failure of one component affects other components in a cause-and-effect relationship. Here, the components are considered as two-state systems only, that is, they are either OK (OK) or not OK (NOK). Each component and each error event is assigned a node of a Bayesian network. It is crucial that each error event has a component associated with the occurrence of the error event. If, for example, a complex component designed as a control unit generates an error event entry, if an out-of-normal voltage value is measured at an input, then the error event identified by the error event entry is assigned to this output of the control unit. When creating the Bayesian network, one node is created for each error event and for each component. The Bayesian network maps a cause-and-effect relationship between the components and the error events: If, in the example given above, another component is connected to the input of the component, then the input, the component and the further component are assigned to the error event. This means that the error event is assigned to the component input of the control unit.

The default probability information provided provides individual failure probabilities for the individual components. It is possible to provide only a single value in the probability of failure information for the same or similar components. This means that, for example, connection lines can each be assigned a uniform default probability value. The default probabilities are assigned as a priori probabilities to the individual nodes of the Bayesian network. For the error events, a maximum uncertainty can be assumed when creating the Bayesian network, ie the probability that the error event has occurred (an error event entry is set) can be regarded as 50% and is just as likely as the non-occurrence. On the basis of the a priori probabilities it is possible by means of the Bayesian network to calculate a confidence value for each of the nodes and thus also for each of the components that the respective component is faulty. This information is complementary to the information that indicates confidence that the corresponding component is not faulty, ie, okay. For the node associated with an error event, the failure to acknowledge the error is okay with the condition and the occurrence of an error event is not properly associated with the condition. If at least one error event is found, it can be "entered" as evidence in the Bayesian network, which means setting the trustworthiness of the node in question to 100%, advantageously analyzing all error events, ie all error event memories of all complex ones Components of the electrical system are read out and thus all confidence values in the error of the corresponding nodes of the error events are set to 0 or 100%, depending on whether a corresponding error event entry is not present or present by feeding this evidence into the Bayesian network the confidence values of the individual components for recalculate their defectiveness. Thus one obtains for each of the components a confidence value for the respective defectiveness. Based on these trust values, it is now possible to create a sorted list of potentially faulty components sorted according to these trusted values. Preferably, the list is thus sorted such that the components that are most likely to be faulty are placed first in the list. It is understood that the output list does not have to include all possibly faulty components. According to a preferred embodiment of the invention, only a certain number of components is output at most or only all components which have the same, preferably maximum, confidence in the defectiveness. The list thus determined is output or made available so that it can be used by a service technician or another method or another device in order to determine the actually defective component via further diagnostic steps and to replace or repair it. The proposed solution is characterized by the fact that only a limited amount of information is necessary in order to allow for any variant of the electrical system reliable determination of those components which are most likely responsible for an error event that has occurred. Apart from the information given above, which in any case arises in the production process, apart from fundamental rules of electrical engineering, which are necessary for the analysis of the structural data set, ie a circuit diagram, no further information is required to create the corresponding Bayesian network. Thus, no special diagnostic knowledge is required about the dependencies of the individual components on each other beyond the diagnostic knowledge contained in the above-described structure data and the error event information. It should be noted that the error events do not necessarily have to be accompanied by the setting of an error code in an error event memory. Likewise, it is possible to use other easily observable or measurable events that allow a fair / out of order classification as error events. For example, lighting a glove box interior light with the glove box lid open may be used as a fault event that occurs when the glove box light is not lit when the glove box lid is open. The indication of whether a possible error event has occurred and, if so, which is, is recorded via an error event indication. This may also include information about the defectiveness of individual components or their perfect condition.

A improved statement is obtained with an embodiment, in which the Bayesian network is simplified when creating so that at least for each of the components, via assignment instructions more than one error event is assigned to the Bayesian network a so-called knitting knot is added over the cause-and-effect relationship with the corresponding error events mapped node is mapped. This embodiment is in each case advantageous if one and the same component several error events are assigned via assignment information. For example, the occurrence of a reading outside a standard range at an input of a component, if necessary with further findings, on two different error events to lead. In a determination of the confidence values of the individual Components would now be all components, who are with the one input in a cause and effect relationship, burdened with the occurrence of both error events "double". With burdened in this context is common that one each conditional probability of the components that the trust worth indicates the defectiveness is increased. However, such modeling is often undesirable.

The Propagation of injected evidence into the Bayesian network, d. h., recalculating the confidence values of the individual components after entering a new error event indication is very time consuming and complex if a node has a large number of parent nodes having. A preferred embodiment of the invention therefore, provides that the network generation unit is a simplification unit to simplify the Bayesian network, so that each node has only a maximum of two parent nodes, with simplification be added to the Bayesian network switching node and wherein the components are parent parent nodes and the error event nodes End child nodes are. The insertion of the switching nodes is done so that the cause-and-effect relationship does not change becomes. This means in particular that the probabilities the switching nodes are initialized such that these probabilities do not affect the results of the cause-and-effect relationship.

Since a defectiveness of individual components can have different effects on a chain of action (or effect) of interconnected components, it is provided in a preferred embodiment of the invention that the network generation unit is designed to include classification information in the generation of the Bayesian network in such a way the probability values associated with the direct child nodes of the components are influenced, the classification information influencing Components for classes of components and the influencing factors a weighting indicate how strongly a failure of a component of the corresponding class enters into effects dependent on this component, whereby each component is assigned to one of the classes. In this way, regardless of the probability of failure of the individual components, a possibility is created of treating them differently according to a classification scheme.

As well is it possible, alternatively or additionally, to provide information to include the influence of a chain of action on the occurrence of an error event. For example, it is off In the design process of a complex component known that the occurrence of an error event at 40% with a faulty one Measurement signal at a first input and in 60% of cases with correlated to a faulty signal at a second input, so this information can be included in the evaluation. This information is also already out of the design process known or can be derived from information in the Maintenance and repair of electrical systems are obtained which also include the corresponding component. A preferred Embodiment of the invention therefore provides that the Network generation unit is formed, impact information, indicate how strong an impact chain is on the occurrence of a Error event affects the generation of the Bayesian network to include that associated with the direct child node of the knitting knots Probabilities are influenced, with an action chain one the knitting knot and the hereby in a cause and effect relationship standing, components associated nodes and, where appropriate between these nodes associated with the components and the knitting node arranged switching node includes. An action chain includes thus in a cause-and-effect relationship all those nodes of Bayesian Net, which act causally on the knitting knot. event nodes or switching nodes that connect to it can be direct Child node of a knot to be its probability through the impact information is affected.

Often it is desirable to use the statements of other diagnostic methods to involve. After an independently inventive Proposal is therefore provided a further embodiment, in which at least one further diagnostic unit is provided, based on at least one further diagnostic method for at least one of the components has further confidence in determines the defectiveness and additionally a fusing unit to create a union list of possibly faulty components sorted according to a total trust worth is provided for the defectiveness, whereby the of The evaluation unit determined confidence values a reliability value and the at least one further confidence value another reliability value and where the reliability value and the more reliability value a weighting for one Significance of the respective method for determining the trustworthiness or the at least one further confidence value, and the merger unit determines the total confidence levels by: a confidence weighted merger of the for a respective confidence value determined by the components is performed. This makes it possible, the statements obtained by different diagnostic methods Overall statement to merge. It avoids that contradictory Information to a service technician or other procedure or a device. By the reliability values is it possible for the different methods to have a different influence value on the total trust worth for a particular component set.

at a preferred embodiment of the invention is provided that the detection unit comprises at least one interface, at least when detecting the at least one error event indication a fault memory of one of the components of the electrical system read. This creates the opportunity the error event entries are automated and swift read directly from the diagnosed electrical system.

at a further embodiment of the invention is provided that provided the interface or another interface are the structural data set and / or the default probability information and / or the error event information when detecting one Database server, preferably a database server of the manufacturer of the electrical system, to retrieve. An identification of the structure data record For example, in an electrical system of a motor vehicle over a chassis number of the motor vehicle. Further will be Structural data sets that uniquely identify the motor vehicle, Already today with motor vehicle manufacturers in one Database kept for the life of the motor vehicle, to be able to understand at any time which components the respective vehicle comprises. The reading out of the probability of default information from a central database also offers the Advantage that the default probability information due to Updated diagnostic information of other comparable systems and can be adjusted.

When particularly advantageous, it has been found, in addition to the stored in control units as components Error event entries additional easy to Observing events as error events in the diagnosis with too consider. Therefore, a preferred embodiment includes a user interface for detecting the at least one error event indication via an input from a user.

The Features of the method according to the invention have the same advantages as the corresponding features of the invention Contraption.

following the invention is based on a preferred embodiment explained in more detail with reference to a drawing. Hereby show:

1 a schematic flow diagram of a method for supporting a diagnosis of an electrical system;

2 a section of an electrical circuit diagram of an electrical system;

3 a table indicating a result of the analysis of a structural data record and error event information;

4 a Bayesian network;

5 a section of a simplified Bayesian network;

6 another section of a randomly simplified Bayesian network;

7 a table that recreates a matrix of cause pairs for the section of Bayesian mesh 5 indicates;

8th a section of the Bayesian network after 5 after a first optimized simplification step;

9 a table that recreates a matrix of cause pairs for the section of Bayesian mesh 8th indicates;

10 a section of the Bayesian network after 5 after completing optimized simplification;

11 a table that recreates a matrix of cause pairs for the section of Bayesian mesh 10 indicates;

12 the Bayesian network after 4 in which the chains of effects are optimally simplified;

13 this will simplify the Bayesian network 4

14 a table that recalculates a number of probability entries of the nodes of the chains of effects for the differently simplified sections of the Bayesian network 4 indicates;

15 a probability table for an exemplary switching node;

16 a probability table of a switching node whose parent nodes are assigned influencing factors;

17 a schematic representation of a device for supporting a diagnosis of an electrical system and the electrical system; and

18 a schematic flow diagram of a method for merging lists of possibly faulty components, which are determined by means of different diagnostic methods.

Based on 1 should a procedure 1 to determine a list of possible faulty components. As the system to be diagnosed, an in 2 illustrated detail of an electrical system 51 considered. The section of the electrical system comprises a control unit of an air conditioning ECU. To the control unit ECU four lines L1, L2, L3 and L4 are connected, which are connected to a recirculation damper U. Here, the line 2 connect L2 and line 3 L3 a motor M of the recirculation damper U. The line 1 L1 and line 4 L4 connect a potentiometer P, which is operated by the recirculation damper and thus makes a current angle of the recirculation damper measurable. On the line 1, a splice S is also attached. This section of the electrical system 51 will be used in the following for illustrative purposes as an example circuit diagram. It should be noted that the recirculation damper U is considered as a complex component. The control unit ECU, the lines 1 to 4 L1-L4, the splice S and the recirculation damper U thus constitute interconnected components of the electrical system 51 which is to be diagnosed.

The ECU ECU is a component that performs a self-diagnosis. Here, two error event entries DTC A and DTC B can be entered into an error event memory (not shown). The error event DTC A may occur if a deviation from the self-diagnostic routine is detected at the input to which line 1 L1 is connected is provided. The error event DTC B may occur when a deviation in the standard is detected at one of the inputs of the control unit ECU to which one of the lines 1 to 4 L1-L4 is connected. The circuit diagram, which includes information about the individual components included as well as their circuits, is summarized in a structural data record. Such a structural data set is obtained, for example, in the production process of a modern motor vehicle. For example, information is collected for each vehicle variant, which include the harnesses installed and the other components connected thereto. In a further data record so-called error event information is stored. This is information that indicates possible observable error events. In the in 2 As shown, the error events are predefined by the error event codes DTC A and DTC B. Each error event is assigned to one of the components of the electrical system via at least one assignment statement. In the above in 2 The error event which triggers the error entry DTC A is assigned to the line 1 L1 via an assignment specification. There are four assignments for the error event associated with triggering the DTC B error code. These each assign the error event to the components line 1 to line 4 L1-L4. Hereinafter, the error events are abbreviated with their error event codes DTC A and DTC B respectively.

When however, possible error events may be others observable events as error event entries in error event stores be used.

Next the structure data and the error event information are additional Probability of default information necessary, preferably summarized in another record. These give default probabilities for individual components or groups of components. Preferably the components in the structural data set are specified in such a way that similar Components are recognizable as such and in the probability of failure information for all these similar components only one default probability is deposited. This can be done, for example, that the individual components are assigned a numeric code. For example, you can all lines 1 to 4 L1-L4 in the example above the same Have default probability.

In a preferred embodiment, classification information and impact information in the form of one or more data sets are additionally available. The use and content of this information is discussed below. All the information and data records mentioned so far are preferably stored on a database server of the manufacturer of the electrical system. If required, they can be retrieved from this database server in order to use the diagnostic procedure described in 1 is shown execute. In the process steps 3 to 11 become the structure data 3 , the default probability information 5 , the error event information 7 and, where appropriate, the classification information 9 and the impact information 11 each recorded. As already mentioned, the detection preferably takes place by means of retrieving the information from a database of the manufacturer of the electrical system or another database. It is also possible to save the information on a data carrier. An assignment of the information to the individual electrical system to be diagnosed can take place via its serial number.

With the help of the available information, a Bayesian network will be created 13 , For this purpose, the structural data set is analyzed with the aid of basic electrotechnical knowledge. This analysis can be automated. Furthermore, the error event information is analyzed. It looks for cause-and-effect relationships between the individual components and error events. Each component can assume two states, in order and not in order. Likewise, an error event may have occurred or may not have occurred.

In Table 1 after 3 the result of the analysis of the structural data record and the error event information is listed in tabular form. It is indicated which components and in which combinations or chains of action can lead to an error event which triggers one of DTC A or DTC B error codes.

If one maps this cause-and-effect relationship between the components and the error events, one obtains a Bayesian network 61 as it is in 4 is shown.

For each of the components of the electrical system 2 a component node K1-K7 is created. Also, an error event node F1 and F2 are created for the error events. Over edges 71 the cause-and-effect relationships are shown. On the basis of the probability of default information and the rules that apply to Bayesian networks, the probability tables associated with the component node K1-K7 and the error event node F1-F2 can now be filled in, as in the method step 23 to 1 is indicated.

Afterwards it is possible an observing capture event information 25 detecting at least one observable error event indication 27 includes. This is preferably done by reading all error memory of components of the electrical system 29 , Alternatively or additionally, observable events may be detected via a user interface 31 , The events thus observed may be used to determine the probabilities of the error events F1 and / or F2. If the error entry DTC B is stored in the error memory, the entry Error event entered in the error event node F2 is set to 1 and the corresponding entry Error event not occurred is set to 0. This introduction of observed information, referred to as evidence, results in confidence values being able to be calculated for each node using the Bayesian network, which indicates a probability of a defectiveness of the corresponding component. The calculation of these confidence values is in the method step 37 to 1 specified. The individual components are sorted based on the calculated confidence values to create one of the potentially erroneous components, weighted by the confidence values 39 , This list is then output or provided 41 , A service technician or other method or device may use this information to perform further diagnostic steps on the electrical system. Events observed or errors that have occurred can be recorded via a user interface 33 and be entered as new evidence in the Bayesian network. As events, all observations can be used that allow a statement about one of the components or one of the error events. That is, indicate whether the corresponding component is OK or not, or state that an error event has occurred or not. Then, the confidence values of each component are recalculated and a new list of potentially erroneous components weighted with their new trusted values created and output. This process may be iteratively continued until one or more faulty components are uniquely identified and / or no fault events on the electrical system are observable.

Looking at the structure of the Bayesian network 61 to 4 with the circuit diagram of the electrical system 51 to 2 Thus, it will be noted that in the event of a failure of the line 1 causing a tripping of both the DTC A and DTC B event, components such as lines 2 and 3 will also be "negatively stressed" in their confidence, although they will can not be the cause of the occurrence of this error event. Therefore, in a preferred embodiment it is provided that so-called knit knots are introduced which "duplicate" those components to which the error events are assigned via allocation indications. This ensures that independent chains of effects can also be treated as independent.

In 5 is a Bayesian network extended by corresponding active nodes WK1 to WK4 61 ^I shown. Since the creation of the list of possible error candidates is simplified as a whole by the insertion of the knitting nodes, the Bayesian network becomes smaller 5 also referred to as a simplified Bayesian network.

The size of the probability tables associated with each of the nodes depends, on the one hand, on a number of states that this node can assume, and on the other hand on the number of parent nodes and their number of ingestible states. Since all of the nodes considered here can only enter or fail in the condition of okay or out of order, the probability tables of the component nodes which only occur as parent-nodes with respect to the cause-and-effect relationship comprise only two probability entries. The nodes that occur as child nodes in the cause-and-effect relationship contain
2 × 2 ^{number of parent nodes} .

In order to accelerate the calculation of the confidence values, it is provided in a preferred embodiment of the invention that the Bayesian network is extended by additional so-called switching nodes, which leads to a simplification with regard to the determination of the list of faulty components. For this, new nodes are inserted into the cause-and-effect relationship such that each node has a maximum of two parent nodes. In 6 is such an expanded or simplified Bayesian network 61 ^II shown in detail, with only the chains of action are included. In this case, six switching nodes VK1-VK6 were inserted. The fragmentary in 6 illustrated Bayesian network 61 ^II is not yet optimally simplified in the sense that it comprises a minimum number of switching nodes. Rather, the switching nodes VK1-VK6 were quasi "arbitrary" in the Bayesian network 61 ^II inserted.

In a preferred embodiment of the method or a preferred embodiment of a device for generating a list of possible defective components, the simplification of the Bayesian network 61 to 4 however, optimized performed. For this purpose, first for the chains of action of the kneading knots WK1-WK4 erwei Bayesian network 61 ^I to 5 Determines how often the individual components K1-K7 together with another of the components K1-K7 enter into one of the chains of action in pairs. This result is preferably in a so-called matrix of the cause pairs, as shown in Table 2 below 7 is shown, registered. Subsequently, a maximum of the registered values is searched for, ie that component pair which together most frequently enters one of the effect chains. In the illustrated case, these are the components air conditioning ECU and recirculation damper U, which are represented by the component nodes K4 and K5. These are pairwise in all four chains of action, which are represented by the active accounts WK1-WK4, a. A switching node VK1 is now inserted into these chains of action. The result of the simplified Bayesian network 61 ^III is in 8th shown. The matrix of the cause pairs is now extended by the switching node VK1 and table 3 is obtained 9 , The maximum value 2 occurring in Table 3 occurs, for example, for the component pair line 1 L1 and splice S. Thus, a switching node VK2 whose parents are the component nodes for the line 1 L1 and the splice S is inserted into the active connection. The process can thus be continued step by step and, after completion of the process, the in 10 shown section of the Bayesian network 61 ^IV that depicts the chains of action. In total, five switching nodes VK1-VK5 had to be inserted in this case. At the in 10 shown section of the simplified Bayesian network 61 ^III the condition is already met that each node has a maximum of two parent nodes. This need not always be the case after completion of the optimization procedure as just described. However, the remaining switching nodes to be inserted can be chosen arbitrarily. In Table 4 after 11 the matrix of the causal pairs is shown, which belongs to the 10 corresponds.

In 12 is the entire Bayesian network 61 ^IV represented by insertion of switching nodes VK1-VK5 simplified chains of action, the fragmentary in 10 is shown. Analogously to the effect chains, the cause-and-effect relationships of the active nodes WK1-WK4 to the error event nodes F1-F2 can also be simplified, as described above, by inserting further switching nodes VK6-VK7 optimally.

In 13 becomes the entire simplified Bayesian network 61 ^v shown. For the chains of action, each comprising the component nodes K1-K7 and the active nodes Wk1-Wk4, as well as the switching nodes VK1-VK5, which may be arranged here between, is shown in Table 5 14 in each case indicated how many probability entries of the chains of action comprising the chains of action each of the Bayesian network. It can be clearly seen that the total number of required probability entries decreases significantly. Furthermore, the calculation of confidence values is greatly simplified and accelerated.

In Table 6 below 15 For example, a probability table for the switching node 5 VK5 according to the example above is shown. If both parent nodes, the switching node 1 VK1 and the switching node 3 VK3, both OK (OK), so is the switching node 5 VK5 okay (OK), so that this state has a probability 1. Accordingly, the complementary state not in the order of the switching node 5 VK5 in this case has the probability 0. If one of the two switching nodes 1 or 3 VK1 or VK3 in the state is not in order (niO), then the switching node 5 VK5 with a probability 1 in the state is not in order (niO). Accordingly, the complementary state OK (OK) is assumed with a probability of 0 in such a case. Such a determination has the effect that the switching node does not influence the chains of action that run through it.

In a preferred embodiment of the method or the device, it is provided that the individual components are assigned to classes and the classes are each assigned an influence factor which indicates how strongly the respective components influence an action chain. Suppose that lines are grouped together and that they should each have an impact factor of 0.9 in the event of a failure. These influencing factors each affect the probability values of direct child nodes of the component nodes. In this case, for example, to the switching node 3, which has the line 2 L2 and the line 3 L3 as the parent node. The resulting probability table is shown in Table 7 16 shown. Likewise, one can use impact information to weight the impact of an impact chain on an error event. In this case, the probability factors of the direct child nodes of the knitting knots are influenced or determined by the corresponding factors.

In 17 schematically is a device for supporting a diagnosis 101 shown. The device 101 is preferably designed as a tester. This includes a detection unit 103 , The detection unit comprises an interface 105 , via which an exchange of data via a motor vehicle interface 107 of a motor vehicle 109 with an electrical system to be diagnosed 111 of the motor vehicle 109 is possible. The electrical system 111 includes component 113 to 119 , From these components 113 to 119 are composites components 113 to 117 complex components that are self-diagnostic capable. The registration unit 103 also includes another interface 121 via which communicate with a database 123 one of the manufacturer of the motor vehicle 109 is possible. In the database 123 are a structural data set 125 and probability of failure information 127 , Error event information 129 and optionally classification information 131 and impact information 133 for the electrical system 111 saved. The in the database 123 stored information can be through the interface 121 the registration unit 103 be captured and sent to a grid-generating unit 135 be handed over. The network production unit 135 is preferably designed in hardware and software. That means the grid generation unit 135 a computer unit 137 and a memory unit coupled thereto 139 includes. In the storage unit 139 is a program code deposited on the computer unit 137 is executable. In interaction of the program code with the computer unit 137 becomes the functionality of the network generation unit 135 provided. The network generation unit 135 creates a Bayesian network, as explained above. The network generation unit has this feature 135 also via a simplification unit 141 , This is also by means of the computer unit 137 and in the storage unit 139 stored program codes realized.

About the interface 105 the registration unit 103 or via an input device 143 a user interface 145 may be an error event, ie an error event indication, of the electrical system 111 be recorded. Preferably, via the interface 105 and the motor vehicle interface 107 all error event entries of the self-diagnosable complex components 113 - 117 read. Alternatively or additionally, via the input device 143 the registration unit 103 Event information is captured, which is preferably entered manually by a service technician and describes observable events. By means of an evaluation unit 147 , which is preferably also formed by hardware and software, become confidence values of the individual components 113 - 119 calculates a list of potentially bad components based on the calculated confidence values 113 - 119 sorted according to the confidence values generated for a defectiveness. The list thus generated is issued. The provisioning can be done, for example, via the interface 121 or via an output device designed, for example, as a monitor 149 the user interface 145 respectively.

The device 101 includes another diagnostic unit 151 which, by means of another diagnostic method, for example a model-based diagnostic method, can also make a list of possibly erroneous components weighted according to otherwise established trustworthiness. In order to be able to merge the two lists produced in different ways into one overall list, there is a merging unit 153 intended. For this, the confidence values of the different diagnostic methods are assigned reliability values. The fusion unit 153 Determines an overall list based on total trusted values for each component. This total list is also provided or via the output unit 149 output.

The method just described for merging lists of possibly faulty components sorted according to confidence values for their defectiveness is shown in a schematic flow diagram 18 shown graphically again. First of all, lists of possibly faulty components are sorted on the basis of different methods, according to the reliability of their defectiveness 161 - 165 , Subsequently, a reliability value is assigned to each diagnostic procedure 167 , The reliability values indicate a weighting for a meaningfulness of the confidence values determined by the respective methods. Subsequently, total confidence values are determined for the individual components 169 by fusing the confidence values weighted with the reliability values. The components 113 - 119 are sorted based on the total confidence values in the overall list 173 and then issued or provided.

The described method and its different embodiments as well as the device described are merely exemplary embodiments. In particular, the network production unit, the simplification unit, the evaluation unit and the merging unit can the same hardware in conjunction with a possibly integrated one Use software.

1

method to create a list of potentially malformed components

3

To capture of structural data

5

To capture of probability of default information

7

To capture of error event information

9

To capture of classification information

11

To capture of impact information

13

Produce a Bayesian network

15

Analyze of the structure data record Error event information

17

Simplify the Bayesian network

19

Insert of knitting knots

21

Insert of switching nodes

23

Fill out the probability tables

25

To capture observable event information

27

To capture at least one error event indication

29

select of error event entries

31

To capture an error event via a user interface

33

To capture of observable events of the components

37

To calculate the confidence values of the components

39

Produce the list of potentially malformed components

41

Outputting / providing the list

51

electrical system

61, 61 ^I -61 ^V

Bayesian Network (or part of it)

71

edge

101

contraption to determine a list of potentially malformed components

103

acquisition unit

105

interface

107

Motor vehicle interface

109

motor vehicle

111

electrical system

113-119

components

121

Further interface

123

Database

125

Structure record

127

Failure probability information

129

Error Event Information

131

classification information

133

impact information

135

Network generating unit

137

computer unit

139

storage unit

141

simplifying unit

143

input unit

145

User interface

147

evaluation

149

output unit

151

diagnostic unit

153

Fusionierungseinheit

161-165

Produce lists of potentially malformed components

167

To assign of reliability values

169

Merge the lists

171

To calculate of total trustworthiness

173

sort by of components according to total trustworthiness

175

Outputting / providing a total list

ECU

control unit air conditioning

L1-L4

cables 1 to 4

U1

recirculation damper

M

engine

P

potentiometer

S

splice

K1-K7

component node

WK1-WK4

active node

F1-F2

Error Event node

CC1-VK7

switching node

QUOTES INCLUDE IN THE DESCRIPTION

This list The documents listed by the applicant have been automated generated and is solely for better information recorded by the reader. The list is not part of the German Patent or utility model application. The DPMA takes over no liability for any errors or omissions.

Cited patent literature

• DE 102005015664 A1 [0005]
• - DE 10222072 A1 [0006, 0006]

Claims (19)

Contraption ( 101 ) for supporting a diagnosis of an electrical system ( 51 . 111 ), in particular an electrical system ( 51 . 111 ) of a motor vehicle ( 109 ), of interconnected components ( 113 - 119 ) comprising: a network production unit ( 135 ) for generating a Bayesian network - from a structural data set ( 125 ), which includes information about the components and their electrical interconnection; - from probability of default information ( 127 ), which allow each of the components to be assigned a probability of failure, and - from error event information ( 129 ), which comprise possible error events and assignment information assigned to the error events, wherein at least one assignment statement exists for each of the error events and an assignment statement exists for an error event of one of the components ( 113 - 119 ), each component ( 113 - 119 ) and each error event is assigned a node (K1-K7, F1-F2) and the Bayesian network ( 61 ) a cause and effect relationship between the components ( 113 - 119 ) and the error events; a registration unit ( 103 ) for detecting at least one error event indication indicating an occurrence of a corresponding one of the error events; an evaluation unit ( 147 ) for evaluating the Bayesian network ( 61 ) and determining a confidence in a defectiveness of each of the components ( 113 - 119 ) and generate a list of possible defective components ( 113 - 119 ) sorted according to the confidence values determined during the evaluation; and an output / provision unit for outputting or providing the list. Contraption ( 101 ) according to claim 1, characterized in that the Bayesian network ( 61 ) is simplified so that at least for each component (L1), which is assigned more than one error event via assignment statements, the Bayesian network ( 61 ) In addition, a so-called knot (WK1-WK4) is added, via which the cause-effect relationship is mapped to the corresponding error events to ordered nodes (F1, F2). Contraption ( 101 ) according to one of the preceding claims, characterized in that the network production unit ( 135 ) a simplification unit ( 141 ) to simplify the Bayesian network ( 61 ), so that each node has only a maximum of two parent nodes, and in simplifying the Bayesian network ( 61 ), And wherein the nodes (K1-K7) associated with the components are parent parent nodes and the error event nodes (F1) are the end child nodes. Contraption ( 101 ) according to one of the preceding claims, characterized in that the network production unit ( 135 ), classification information ( 131 ) in the generation of the Bayesian network ( 61 ^v ) in such a way that the direct child nodes of the components ( 113 - 119 ) are influenced, the classification information ( 131 ) Influencing factors for classes of components ( 113 - 119 ) and the influencing factors indicate a weighting, how strongly a failure of a component ( 113 - 119 ) of the corresponding class in of this component ( 113 - 119 ) receives dependent effects, each component being assigned to one of the classes. Contraption ( 101 ) according to one of claims 2 to 4, characterized in that the network production unit ( 175 ), impact information ( 133 ), which indicate how strongly an action chain affects the occurrence of an error event in the generation of the Bayesian network ( 61 ^v ) in such a way that the direct child nodes of the action nodes (WK1-WK4) are assigned to the probabilities, an action chain comprising the active node (WK1-WK4) and the components (CW1-WK4) 113 - 119 ) associated nodes (K1-K7) and the optionally between these the components ( 113 - 119 ) associated nodes (K1-K7) and the active node (WK1-WK4) arranged switching nodes (VK1-VK5). Contraption ( 101 ) according to any one of the preceding claims, characterized in that at least one further diagnostic unit, based on at least one further diagnostic method for at least one of the components ( 113 - 119 ) determines another confidence value for the defectiveness, and a fusion unit ( 153 ) for generating a union list of possibly faulty components ( 113 - 119 ) are arranged according to an overall confidence value for the defectiveness, whereby the information provided by the evaluation unit ( 147 ), wherein the reliability value and the further reliability value indicate a weighting for a validity of the respective method for determining the confidence value or the at least one further confidence value, and the fusing unit ( 153 ) determines the total confidence values by using a reliability weighted merger for each of the components ( 113 - 119 ) is carried out. Contraption ( 101 ) according to one of the preceding claims, characterized in that the detection unit has at least one interface ( 105 ) in order to detect at least one error memory of one of the components ( 113 - 119 ) of the electrical system ( 111 ) read out. Contraption ( 101 ) according to one of the preceding claims, characterized in that the interface ( 105 ) or another interface ( 121 ) are provided to the structural data set ( 125 ) and / or the probability of default information ( 127 ) and / or the error event information ( 129 ) when capturing from a database server ( 123 ), preferably a database server of the manufacturer of the electrical system ( 111 ), to retrieve. Contraption ( 101 ) according to one of the preceding claims, characterized by a user interface ( 145 ) for detecting the at least one error event indication via an input of a user. Contraption ( 101 ) according to one of the preceding claims, characterized in that the user interface ( 145 ) for detecting inputs in order to change the confidence values of nodes, the checked components ( 113 - 119 ) assigned. Procedure ( 1 ) in support of a diagnosis of an electrical system ( 111 ), in particular an electrical system ( 111 ) of a motor vehicle ( 109 ), of interconnected components ( 113 - 119 ) comprising: acquiring a structural data set ( 125 ), information about the components ( 113 - 119 ) and their electrical interconnection comprises; Detecting default probability information ( 127 ), which allow each of the components ( 113 - 119 ) assign a default probability; Capture error event information ( 129 ), which comprise possible error events and assignment information assigned to the error events, wherein at least one assignment statement exists for each of the error events and an assignment statement exists for an error event of one of the components ( 113 - 119 ) assigns; automatic creation of a Bayesian network ( 61 ) based on the structural data set ( 125 ), the probability of default information ( 127 ) and the error event information ( 129 ), each component ( 113 - 119 ) and each error event is assigned a node (K1-K7, F1, F2) and the Bayesian network ( 61 ) a cause and effect relationship between the components ( 113 - 119 ) and the error events; Detecting at least one error event indication indicating an occurrence of a corresponding one of the error events; Evaluation of the Bayesian network ( 61 ) and determining trustworthiness of the components ( 113 - 119 ) and outputting a list of possible defective components ( 113 - 119 ) sorted according to the confidence values determined during the evaluation. Procedure ( 1 ) according to claim 11, characterized in that the Bayesian network ( 61 ) is simplified so that, at least for each assignment instruction associated with one of the error events to which more than one of these assignment instructions is assigned, the Bayesian network ( 61 ) a so-called knitting knot (WK1-WK4) is added, which is inserted into the depicted cause-effect relationship. Procedure ( 1 ) according to one of claims 11 or 12, characterized in that the Bayesian network ( 61 ) is simplified so that each node has only a maximum of two parent nodes by adding switch nodes (VK1-VK7), the nodes (K1-K7) associated with the components being the parent parent nodes and the error event nodes (F1, F2) being the parent node End child nodes are. Procedure ( 1 ) according to one of claims 11 to 13, characterized in that classification information ( 131 ) and each component ( 113 - 119 ) is assigned to a class, the classification information for each class comprising an influencing factor that indicates a weighting of how severely a failure of a component ( 113 - 119 ) of the corresponding class in of this component ( 113 - 119 ), and the factors influencing the direct child node of the components ( 113 - 119 ) to influence assigned probability values. Procedure ( 1 ) according to one of claims 12 to 14, characterized in that impact information ( 133 ), which indicate how strongly an action chain affects the occurrence of an error event, and this impact information ( 133 ) are used to influence the probabilities associated with the direct child nodes of the knitting nodes (WK1-WK4). Procedure ( 1 ) according to any one of claims 11 to 15, characterized in that based on at least one further diagnostic method for at least one of the components ( 113 - 119 ) another trustworthiness for the defectiveness is determined, and the trustworthiness of the list and the at least one further trust worth each a reliability value is assigned, wherein the reliability value indicates a weighting for a validity of the respective method for determining the confidence value or the at least one further confidence value, and a union list of possibly faulty components ( 113 - 119 ), the total confidence value being determined by a merit weighted merger for each component (s). 113 - 119 ) is determined. Procedure ( 1 ) according to any one of claims 11 to 16, characterized in that when detecting the at least one error event indication via a user interface ( 145 ) are entered by a user entered event information. Procedure ( 1 ) according to any one of claims 11 to 17, characterized in that when detecting the at least one error event indication error memory of components of the electrical system ( 51 . 111 ). Procedure ( 1 ) according to one of claims 11 to 18, characterized in that the structural data record ( 125 ) and / or the probability of default information ( 127 ) and / or the error event information ( 129 ) when capturing from a database server ( 123 ), preferably a database server ( 123 ) of the manufacturer of the electrical system ( 51 . 111 ).