WO1999017176A1

WO1999017176A1 - Module for diagnosing electrically controlled systems and total system diagnosing device

Info

Publication number: WO1999017176A1
Application number: PCT/EP1998/005324
Authority: WO
Inventors: Rainer Burkhardt; Andreas Heinzelmann; Bernd Rehfus
Original assignee: Daimlerchrysler Ag
Priority date: 1997-09-26
Filing date: 1998-08-21
Publication date: 1999-04-08
Also published as: DE19742448C1

Abstract

A module and device are disclosed for establishing a diagnosis. The diagnosis module diagnoses electrically controlled systems arranged within a total system on the basis of models of the total system. The module contains means for generating a reduction graphic comprising a head graphic with nodes having only fixed or measurable values, means for sensing the measurable variables required for the diagnosis, for interpreting the measurable values in the head graphic, for reproducing the interpretation carried out in the head graphic on the plane of the building elements and on the plane of the components and for signalling components which are defective or suspected of being defective. The diagnosis device gives a defect assessment of the whole system. It contains the diagnosis module, means for establishing a diagnosis of the whole system based on the functional relationships between the individual electrically controlled systems, and means for generating defect assessments for the whole system and assessments of the way in which the functions of the whole system are affected by said defects.

Description

Title: Diagnostic module for creating a diagnosis for electrically controllable systems and diagnostic device for creating an overall system diagnosis

DESCRIPTION

^• The invention relates to a diagnostic module for creating a diagnosis for electrically controllable systems and arranged within an overall system. Furthermore, the invention relates to a diagnostic device for an overall system with electrically controllable systems arranged within the overall system.

For a multitude of applications of complex networked complete systems consisting of electrically controllable components such as B. from the field of automation technology, process control technology, circuit development or vehicle electronics, it is necessary for fault diagnosis to know the normal and faulty behavior of the individual components, the effects of this behavior on other components and in particular on the overall system.

Through data material about the overall systems, e.g. B. from their development phase, it is possible to describe the behavior of the components using models and to generate computer-controlled knowledge data about the behavior of the individual components by simulation. Known simulation methods allow the simulation of individual components and subsystems, but very often fail if the simulations are to be extended to the entire system. This is essential if the effects of the individual component behavior or the behavior of groups of parts on the overall system are to be examined. Further declarative description procedures take the diagnostic knowledge from libraries or use targeted knowledge. DE 4124 542 C2 describes a fault diagnosis device for determining the cause of a fault in a device under test with a detection device and a memory device, the detection device detecting the parameters of the device under test. In the storage device, a search tree with nodes that correspond to the respective subunits of the device under test, as well as the test tables assigned to the nodes, in each of which at least one parameter to be detected by the detection device and a test operation in this regard are specified, an error probability table corresponding to the results of tests the at least one test condition and names of daughter nodes are stored beforehand, wherein at least two parameters and test conditions to be detected are additionally specified in a test table which is assigned to a node with at least three daughter nodes. In addition, a search / interference device is stored in advance in the memory device, which selects nodes along the search tree and evaluates the associated test tables, the node selection being carried out based on the result of the evaluation of the test tables.

US Pat. No. 5,099,436 describes a method and a device for carrying out a system fault diagnosis which are based on a hybrid representation of the knowledge of the system to be diagnosed. Data captured during system runtime is compared to an event-based system representation that includes a variety of predefined events. An event is recognized when the captured data matches the critical parameters of the event. The detected event and an associated set of ambiguity group effects, which identify components that are to be re-sorted according to an assigned sorting effect in an ambiguity group, are analyzed. A symptom failure model and a non-function model can also be analyzed to determine the symptom failure relationships and nature of the non-functions that are applicable to the system run. Each applicable symptom error relationship and each type of non-function is also assigned a set of ambiguity group effects that reorder the ambiguity group, starting with those components in the ambiguity group that are most likely to fail, a structural model is analyzed, and as a result of the analysis, repair suggestions are included in the system tests to be performed.

Structural parts of a computer-aided fault diagnosis device for a motor vehicle are in the publications N. Waleschkowski et al., "A knowledge-based vehicle diagnosis system for use in the motor vehicle workshop" Fundamentals and Applications of Artificial Intelligence, Springer-Veriag, 1993, page 277, and N. Waleschkowski et al., "Knowledge modeling and knowledge acquisition using the example of vehicle diagnosis", magazine Artificial Intelligence Kl 1/95, page 55. This facility contains a diagnostic process stage with a knowledge base that contains a structural model of the hierarchical structure of the technical system from individual subsystems, an effect model of the relationship between the individual subsystems, and a diagnostic model. process-defining fault model that shows the relationships between the causes of the fault and their effects, as well as suitable test sequences and repairs.

These known methods for generating diagnostic knowledge are declarative methods, i.e. Errors are only determined on the basis of previously known and previously defined or set events. Certain previously known effect chains are also stored and allow a description of the system behavior in that the symptoms found during diagnosis can be assigned to certain errors by comparison with the predetermined effect chains. Errors can thus be recognized by the known pattern recognition methods applied to the symptoms identified.

For example, the specification of normal behavior in a specific operating state and a comparison with reality, the detection and localization of faults. In practice, limit values are often used to indicate an error if the value is exceeded or not reached.

Declarative procedures are not suitable where the sub-systems or entire systems to be diagnosed change continuously and sometimes at very short notice and develop further and several different variants are used at the same time. The declarative diagnostic systems then require a lot of preparation and maintenance.

The object of the invention is to provide a diagnostic module and a diagnostic device for creating a diagnosis, which do not have the disadvantages mentioned and which, when making a statement about errors, include the information about the entire system available up to the time of diagnosis.

According to the invention, this object is achieved by the diagnostic module with the features of claim 1 and the diagnostic device according to claim 7.

The diagnostic module creates a diagnosis based on models of an overall system for electrically controllable systems which are arranged within the overall system. It contains means for setting up a reduction graph which has a head graph with nodes which are only of fixed or measurable sizes. The means for setting up the reduction graph comprise means for disassembling the electrically controllable systems into components which comprise one or more basic components, means for detecting the electrical connections between the basic components and between the components, means for assigning discrete electrical status values to the basic components, means for defining the operating states and the possible component states of the individual components in relation to the electrical state values of the basic components belonging to a component, means for determining the quantities necessary for the diagnosis and measurable on the basic components and components, means for determining basic components that refer to the specified measurable sizes have no influence and means for summarizing and eliminating the found basic building blocks.

The diagnostic module contains further means for recording the measurable quantities necessary for the diagnosis, for interpreting the measurable quantities in the header graph, for mapping the interpretation carried out in the header graph on the level of the basic building blocks, for mapping the interpretation depicted on the basic building blocks on the level of the components and to indicate suspect and / or defective components.

The diagnostic module takes into account the structure of the system components as well as the relationships between the components, the components and the basic building blocks as well as the component behavior. The entire system with its electrically controllable systems is thus represented by the direct use of its circuit diagram data and the huge amounts of data are meaningfully compressed.

The error message is analyzed using a simple model, the header graph reduced to nodes with fixed or measurable sizes. The error analysis therefore only takes place at the time of diagnosis. This offers the advantage of incorporating knowledge about the entire system or subsystems into the error analysis up to the time of diagnosis.

So z. B. during the life of the system to be diagnosed, aging probabilities, specific user behavior or environmental influences can also change the probability of failure.

The knowledge generation required for diagnosis is carried out directly from the design documents or circuit diagrams of the overall system to be examined using the reduction process. For this purpose, the design documents must contain the component information with the electrical status values as well as information on the measurable quantities. This type of diagnostic knowledge generation eliminates the otherwise time-consuming simulations that transfer the normal and incorrect behavior of individual components to the overall system.

The diagnostic device provides an error statement about the overall system. In addition to the diagnostic module for creating the diagnosis based on models of an overall system for the electrically controllable systems arranged within the overall system, it contains means for creating a diagnosis for the overall system based on the functional relationships between the individual electrically controllable systems. With these means, a graph consisting of two subgraphs can be created, the first subgraph representing the structure of the components and the second subgraph representing the relationships between the overall system functions and the functions of the individual components. The diagnostic device contains further means for generating error statements about the overall system and statements about the impairment of use of overall system functions. The diagnostic device not only delivers faulty components but also makes error statements about the overall system, it being particularly advantageous for the user of the overall system to recognize whether and where there is a faulty usage function.

A further advantage for the user is if the diagnostic device additionally comprises means for proposing suitable remedial measures, as stated in subclaim 12.

Further advantageous refinements are specified in the subclaims.

Preferred embodiments of the invention are described with reference to drawings, in which:

1 shows a reduction graph

2 shows the procedure for creating the system models

diagnosis

The essential components that appear in the periphery of automation devices are resistors, switches, lines, plugs, lamps, cable bundles, sensors, relays, fuses, solenoid valves, motors and signal transmitters. Some of the components can be summarized in aggregate A. The aggregate information with the structural relationships are usually available in a data processing form. This information is already available during the product development phases. B. available in CAD systems or can be derived from model-based programs.

The aggregate, the representation of the components, their behavior and the possible component states are described in detail in the patent application filed simultaneously with the application name Daim 27753 at the DPA. This description and the explanations for the representation of the reduction graph and for the custody are to be part of the present application.

As stated in the patent application Daim 27753, the periphery of the overall system can be modeled as a graph, in which the basic building blocks represent the edges. To connect the basic building blocks, connection points are required that appear as nodes in the graph.

The entire behavior of the electrical periphery is characterized by the elementary peripheral graph G ^ in an elementary peripheral graph G _ep , the node potentials P _s or the currents I can be measured by a basic module. The measurability of the node potential P _j or the branch current I; depend on the local position and the structure of the entire system and can assume the values measurable or not measurable. Starting from the elementary peripheral graph G _ep , the basic operations known from network analysis can be used to reduce the graph, which contains the essential behavioral properties with regard to the observable physical quantities.

As shown in FIG. 1, a reduction graph G _{r can} consist of several subgraphs. These include the elementary peripheral graph G _ep , several intermediate graphs G, and a head graph G _h . The nodes and edges of the individual subgraphs are connected to each other and represent the reduction rule.

The electrical properties contained and specified in the elementary peripheral graph are simplified by the reduction process with regard to the measurable quantities. The behavioral properties of the electrical components are clearly mapped to the elements in the head graph by the reduction. On the other hand, a clear mapping of the elements to the component level cannot be guaranteed based on the header graph. This property is due to the limited information in relation to the components causing it. In practice, this means that a certain fault symptom characterizes several causes of the fault.

The reduction takes place with the basic operations. After each reduction run, the number of nodes N ₍ . The elementary basic graph G _ep is processed until no node or edge can be removed from the graphs.

At the end of the reduction process, the head graph consists of nodes with fixed or measurable potentials. All nodes with unknown potentials are removed in the header graph G _h . The header graph now contains all the discrete states specified in the elementary peripheral graph.

Especially for complex overall systems with a large number of components, the time required for the creation of the reduction graphs is considerable. It can therefore be advantageous to carry out a clustering before the actual reduction process, which takes into account the knowledge that not all operating and faulty states of components affect all system sizes. Rather, there are spatially independent areas in which certain states of components have an effect, and areas which are not influenced by all components.

Clustering also allows very complex networked overall systems to be examined, since the subsequent diagnosis is reduced to a local problem.

2 schematically shows the procedure for creating the diagnosis based on system model.

First, with the steps described above, a reduction graph has to be drawn up, which forms the starting point for the model-based diagnosis. For this purpose, the electrically controllable systems must be disassembled into components with one or more basic components. The electrical connections between the basic components and between the components are then detected and discrete electrical status values are assigned to the basic components. The operating states and the possible component states of the individual components are defined in relation to the electrical state values of the basic components belonging to a component, and the quantities necessary for diagnosis and measurable on the basic components and components are specified. Finally, basic building blocks are determined that have no influence on the specified measurable quantities. The determined non-influential basic building blocks are summarized and eliminated. The reduction graph set up in this way has a head graph, the nodes of which either have a fixed potential or whose potential can be measured at the node.

In addition, means for detecting error messages can be provided. An incoming error message is recorded by monitoring the electrically measurable quantities and making them available as input information for the diagnosis.

Known error detection methods can be used to differentiate between normal and incorrect behavior of a subsystem. The resulting symptoms or error messages are interpreted for the subsequent diagnosis. The integration takes place in the head graph G _h taking into account all information necessary for the diagnosis in the form of measurable physical quantities.

The explanations for the symptom that occurred in the header graph G _{h represent} a further processing step on the basic building blocks of the elementary peripheral graph G _ep . This enables conclusions to be drawn in the reduction graph at the basic building block level.

The interpretation carried out in the head graph G _h is then mapped to the component level. The transfer of calculated discrete values of the elementary peripheral graph G _ep to the component level makes it possible to determine possible faulty and / or suspicious components.

In systems that can only be described by two measurable quantities or potentials, possible error messages can represent the state of an interruption of the control circuit, short circuit to ground or short circuit to \ l _∞ .

In the case of a large number of different potentials in the system, any error message due to a potential or current error, caused by the individual stem values at the measurable node, must be described. The following prerequisites must be observed so that the faulty component can be localized:

• There is a reduction graph with node N _: its potentials P, all of which are known in the diagnostic cluster. • There are possible discrete values for the individual basic modules Uj.

• For the localization of voltage errors, all node potentials of the node to be analyzed and the corresponding neighboring nodes must be known.

• For the localization of current errors, at least one current I must be known for the node N _a to be analyzed, through a basic building block U _{i (} the potential of the node to be analyzed and the corresponding neighboring nodes).

Only one measurable node a is taken into account in the analysis in the head graph. The external nodes i have a fixed potential for this consideration or are assumed to be fixed, so that the following initial situation results:

V.

8 a

The following applies to a potential deviation:

The measured potential is to be explained by the discrete resistance values and all possible combinations of the individual values from the head graph form the candidate set. The resulting node potentials are determined for all possible combinations. If the hypothetically determined potential agrees with the measured potential, the discrete values present in the combination form an explanation for the current observations.

The following applies to a current fault:

The calculation of the current error requires knowledge of the voltage potential at measuring node a and at outer node i. All measurable currents 1 ^ are included in the calculation. The arrows are all directed to node a.

Here too, all combinations of the individual values of the header graph form the possible explanations for the current error. If the potential between an outer node i and the inner node a is the same, the equation provides no information about the corresponding star resistance value Ri.

If there is an undefined potential at the measuring node, the elements can be determined directly without knowledge of the possible discrete values of the header graph.

Vic \ = 00

In the event of measurement inaccuracies, the values determined for the fault localization are already assigned predefined values. For example, a relatively high current is equated to a short circuit with l = ∞ compared to normal current. Measurement inaccuracies can be eliminated in this way.

Another possibility is to subject the hypothetically calculated value to an assessment of the extent to which it corresponds to reality. This allows the individual results to be weighted and prioritized in further processing.

The result of the analysis in the header graph is a set of discrete values of the individual elements, which explains the observed symptom.

The analysis carried out in the header graph is now mapped to the level of the basic building blocks. To do this, superordinate discrete values must be explained by subordinate values in the reduction graph. Some subordinate discrete values can be specified directly. For all graph elements that contain only one discrete value or several identical values, the subordinate value is already known regardless of the superordinate value. The subordinate values can be determined directly from the superordinate substitute values without a subordinate value being available for the calculation. Generally speaking, the result is always ambiguous, since the reduction is an abstraction of the system and a combination of several subordinate values into one superordinate value. For certain special cases, despite this general procedure, the subordinate values result directly from the abstraction operation and the superordinate values.

A conclusion in the reduction graph is necessary for all graph elements in the lower level with alternative discrete values. One possibility for this is to calculate the values from the graph below to the next higher one and to check whether they match the known values. It is also possible to deduce the discrete values of the subordinate graph elements directly from the known superordinate values and taking into account the type of the reduction operation.

A search with a specific calculation assumes a systematic search assuming a known value, and conditions resulting from the reduction operation can be included in the search. With the conditions specified below, impossible combinations of values can be excluded from the outset.

Conclusion about series reduction:

The following boundary conditions apply to the series reduction with the basic modules of the type resistors R _k and voltage sources Ci ,,:

for resistors V = R for voltage source with the same counting direction V = U

^{v '} + l) - ^r se - ^r ; (M)

3 V _{k M)} \ V _K = V _{k {M)} => V, _(W) = 0 Λ * ≠ i

Conclusion on parallel reduction:

The calculation of the individual causal values for the basic modules of the type resistors R _k and current sources Q; to simplify as follows:

for resistors R = 1 / V for current sources with the same counting direction Ql = V

V F ι. . -7 p ≥ - V:.

3 V _k \ V = V _t ^ VV ^ OA k ≠ i In contrast to known reduction processes, the series and parallel reduction are not only carried out with two components, but can have any order. This results in advantages with regard to a more compact and efficient representation of the overall system.

Conclusion about a Stern red uction:

In order to calculate the corresponding values of the star from the values of the polygon, the relationship between the two Schenkei resistors R ^, R _μ0 and the equivalent resistance R _{vμ applies} :

R ^ ≥ R ^ + R μO

The individual discrete values obtained from the elementary peripheral graphs can now still be transferred to the corresponding states of the components.

Due to the ambiguous inference process, contradictions can arise in the discrete states. The component states can be divided into the three categories "defective", "O.K." and classify "possibly defective". Contradictory statements as a result of the conclusion between "O.K." and any fault condition result in the "possibly defective" condition.

There are a lot of possible explanations for each observation. The individual explanations can be prioritized based on a simple assumption. Provided that the failure probability P of a component e.g. an exponential distribution of the shape

P (t) = \ -e- ^{> J}

and the failure probability of all components for all arbitrary times t _{x has} the same size, the probability of an n-fold error results from the following relationship: ^P n-faA ^t ) = Y [P _Ci t) = p (t) "do under the condition (t nff

This applies on the assumption that there are no probability dependencies between the components. From this it can be deduced that the higher the order of the error, the less likely it is to occur, or formally expressed:

P _n -.- f _c ) <P _n . _subject (t)

^P n -.- facH (0] θ, l [

The diagnostic result is output with a prioritization of the number of defective components per diagnostic result. If the probability of failure is present or determined during the runtime of the overall system, the diagnosis result can also be output after a corresponding prioritization.

During the conclusion in the reduction graph or when the diagnostic results are mapped onto the individual components, it is possible to identify inconsistencies between the knowledge gained from the measurements and the model. For example, if the calculated discrete values are not contained in a subordinate graph, this is an inconsistency:

The cause of the inconsistency can occur on the one hand in the measurements carried out or in the diagnostic knowledge, the model. Errors in the measurements caused by external influences or a faulty measuring device result in faulty measurement results, which in turn cause the inconsistency.

With the described procedures, the information necessary for the diagnosis is compiled from the complex overall system, taking into account the component structure and using suitable clustering methods. The model generation is limited to a necessary and thus limited number of components of the overall system, which leads to a considerable acceleration of the diagnostic process.

An inadequate model, ie a model with insufficient level of detail to describe the fault, is the second cause of the inconsistency. The second case is to be eliminated by error models with different levels of detail. This means that the underlying error model is expanded in the event of an inconsistency by specifying a new level of detail and starting the diagnostic process again. The extension is made according to the probability of occurrence of the errors per component. The diagnostic device according to the invention contains a diagnostic module, which creates a diagnosis for each electrically controllable system according to the method described in detail. The results of the individual system diagnoses are made available to a hierarchical structure that takes into account the functional relationships between the electrically controllable systems and the overall system structure

The hierarchical structure can be represented by a directional graph consisting of two subgraphs. The first subgraph describes the physical structure of the individual components. The second subgraph characterizes the graded relationships between the functions of the overall system and the individual components. Both graphs consist of nodes and edges E _j

The edges N _: represent the relations between the individual nodes. The nodes of the graph include components c ,, functions 1j. and connection operations _ok . Each node has additional information that is stored in attributes.

The first subgraph contains the components installed in the system as nodes and is therefore referred to as the compositional hierarchy. The attributes of the nodes include the information as to whether the respective component represents a smallest exchangeable unit. An attribute also indicates whether a node is a surplus. The excess quantities allow the individual components to be structured better. For example, the excess of automation devices includes all individual automation devices. The individual components are connected to each other with a conjunctive link, since the individual systems can be clearly assigned to an overcomponent. The direction of the edges points to the supercomponent above.

The second subgraph describes the function hierarchy. In this subgraph, the individual functions are listed as nodes. Each node that represents a function contains the level of output and that of criticality as an attribute. The output level is used to differentiate between error messages for different user groups of the overall system. B. for the user or for the service technician.

In addition, the second subgraph contains connection operations. The connecting operations consist of conjunctive and disjunctive connecting elements as well as negations. To connect the individual nodes, the hierarchical function graph contains directed edges, each of which refers to the next higher-level function. There can be one or more connection operations between two functions, which represent the logical dependencies between the over and under functions. Individual functions, each relating to the normal function of the individual subsystems, form the basis for several overfunctions. Connection operations are therefore included in the subgraph for multiple use. To connect the two subgraphs, the overall graph contains edges in which the respective elementary basic components are mapped to the elementary subfunctions.

The hierarchical structure comprises the smallest interchangeable unit belonging to a faulty component, which can be determined for a given function or component.

The information contained in this hierarchical system representation can be used in different ways. One possibility is to determine the respectively impaired functions in the corresponding level from a defective component. Another possibility, based on a defective function, determines the components causing it. A mixture of the two methods, in which a certain number of defective components and failed functions are known and from which all unknown states of the components and functions are determined, is also conceivable.

In order to draw a conclusion in the hierarchical representation, dynamic data are required that represent the current status of the individual components and functions. These dynamic data include the three states "faultless component / function", "unknown state" and "faulty component / function".

In order to deduce higher-level states from known subordinate states by means of the connection operations, relations are established with the three states, each of which relates an output value of a connection operation to one or two input states.

According to this scheme, a conclusion can also be made from a higher level to a lower level.

In order to determine the smallest exchangeable unit of a component with the "faulty component / function" state, the subgraph containing the component is run in the direction of the edges until the search method encounters a component that has the attribute smallest exchangeable unit and the state Contains "faulty component / function". The states of the higher-level components result from relations.

If the states of the individual components are known, a search method first determines the states of the functions in the lowest level of the second subgraph. Then all functions are found in this graph in the direction of the directed edges, which have the "faulty component / function" state with the previously defined output level. The state of the individual elements in the graph results from the relations established during the search process. The function nodes determined provide the information about the failed functions with the appropriate criticality. If the failure of a function is known, the components causing it can be determined. Based on the function with the "faulty component / function" state, the method determines all subordinate function states down to the lowest level with the help of the relations established. The states of the basic functions represent the component states and can be mapped to the basic components determine the smallest exchangeable units. Only those functions and components that have the status "faulty component / function" or "unknown status" are of interest in the analysis. An occurring disjunctive connection operation forms an alternative path and thus an alternative set of causing components.

For complex, networked overall systems that consist of a large number of components, the functional and component hierarchy that is as automatic as possible is useful for economic reasons, especially if a large number of versions and variants of the systems also occur. An analysis of the underlying knowledge base shows that the function hierarchy is subject to ever increasing specialization due to the actual implementation as the level of accuracy increases. Functions that are at a very high level of abstraction are hardly or not at all characterized by the implementation. Functions at the lowest level, on the other hand, have a very strong implementation dependency that is shaped by the individual components. The top part of the functions can be assumed to be fixed with regard to the version and variants of a system, whereas the bottom part of the function hierarchy is characterized by the individual implementation.

The component hierarchy can essentially be found in a library. The individual attributes of the components can be stored in the component library. Data from the development process chain or from the production documents provide the exact number of units installed in the overall system. The component library can also group the individual components. The basic functions can also be stored in the component library.

The creation effort for the fixed part of the function hierarchy with the individual attributes is a one-time process, since there are only minor changes, for example with a further development of the overall system. Therefore, the fixed part of the function hierarchy can also be taken from an overall system library.

The implementation-dependent part of the function hierarchy is filled with data about the entire system, e.g. B. with data from the development process of the system or with construction drawings or circuit diagrams. The occurring signal, energy and information flows serve as input data for this. Each individual Signa!, Energy or information flow contains a certain fixed function in the overall system. For example, there is a signal flow in the overall system of a vehicle available with the task of displaying the speed in the display instrument. All components with the corresponding component functions enable the speed display. The individual connection operations of the second Teiigraphen result directly from the signal flow.

For electrical components in the periphery of automation devices, the normal states of the elements occurring in the header graph can also be used to determine the components belonging to a function. For this purpose, the electrical components present in the periphery with their normal state are depicted in a header graph in accordance with the reduction method described above. Fixed functions from the specification of the automation device are known for each input / output of the automation device. These functions can be assigned to the individual elements in the header graph, which are located at the corresponding input / output of the automation device. The components involved must now be determined from the individual elements of the header graph with the corresponding functions. There is a conjunctive connection between the components. This makes it possible to automatically determine the functions of the individual electrical components present in the periphery and finally to enter them in a function hierarchy.

Compared to known and z. B. in Wiedmann, Hans: "Object-oriented knowledge representation for model-based diagnosis at manufacturing facilities", Berlin, Springer-Veriag 1993 detailed functional system descriptions, the hierarchical structure can be represented in any logical dependencies with the present approach. The connection of the functional and components here - chie takes place only at the lowest levels and thus considerably simplifies the automatic creation and consistency of the two viewing levels, functions and components.

The functions involved can be determined from the components in different levels of abstraction. The components involved can also emerge from the faulty functions.

Using the overall system structure thus constructed, error statements about the overall system can be made taking into account the suspect components. The described structure of the overall system structure also offers the possibility of making statements about the impairment of functions of the overall system.

When diagnosing highly complex overall systems, it can be advantageous to localize errors in cross-subsystem networks. This requires a higher-level model based on a functional description of the effects. This can be used to describe the signal, energy and information flows that occur. A method is used to localize faults between the subsystems, which is based on the AAAI Workshop Reasoning in J. Stickten et a., "Troubleshooting based on a Functional Device Representation: Diagnosing Faust in the External Active Thermal Control System of Space Station FRREDOM" About Function, Washington DC, July 11, 1993, pp.149-156, builds the FR-Dx algorithm described and develops it further.

With the functional description of effects, a directed function graph is created for the real system in which the individual functions of the components are shown with the system variables to be observed. The edges of the function graph reproduce the individual functions of the components and the nodes characterize the system sizes. Binary statements about their status, normal status or error status are possible for edges and nodes.

Connection elements between the individual information flows are required for an adequate representation of the real system. If effects with normal behavior as well as effects with faulty behavior are to be shown, besides conjunctive also disjunctive system dependencies are shown. In addition to the connection elements, the function graph requires branches that the individual system sizes can provide for various other dependencies.

The individual dependencies between the system sizes and the functions result from a graph transfer process. Starting from each individual system size, the graph is run against the directed edges and the individual relations are determined from this.

The overall system description is made with a disjunctive normal form, in which a system size is represented by several disjunctively and conjunctively linked functions.

The effect relationships can be represented in a three-dimensional function matrix, the columns in each case representing the system sizes and the rows each representing the function of the components on one level. The elements within a level are conjunctively linked and the connection between the individual levels is disjunctive.

The creation of the function graph and the function matrix depends on the structure of the individual information sources. In addition to e.g. Declarative relationships between the system variables and the components can also be stored as a model from the development process chain of the overall system, since the description elements provided allow any dependencies between system variables and components to be recorded. Declarative relationships can also be included in the diagnosis. The functional effect descriptions can be created without any problems because the method does not require exact mathematical relationships, but only records the relationships between the input and output variables.

The method takes into account not only faulty functions but also system variables that have no errors. Information about defective components from subsystem diagnostics can also be taken into account as an input variable and included in the processing. This extended FR-Dx method allows the display and diagnosis of complex networked systems with different types of components, such as B. of electronic, mechanical and / or hydraulic components. As with all model-based diagnostic methods, in contrast to the declarative methods, the extended FR-Dx method can also detect multiple errors.

It is essential in this method that it is primarily not the defective components but the faulty functions that make up the result of the diagnosis.

The diagnostic device advantageously contains a data memory so that the determined diagnostic results and other data relevant to the diagnosis can be stored.

A dialog control unit that controls the dialog with the users of the overall system is advantageous. If the overall system is, for example, a vehicle, the driver or the workshop personnel must be specified as users. The dialogue is made easier for them if the error statements and the statements on the impairment of use are prepared audiovisually. Appropriate means can be provided on the diagnostic device for this.

A further relief for the user is if he not only receives error statements about the overall system, but if additional measures are suggested based on the existing knowledge of the overall system structure.

The structure and mode of operation of the diagnostic module used in the diagnostic device allow the diagnostic device to be used not only for diagnostic purposes if an error message is received. Rather, the diagnostic device can be used during the normal use of the overall system, in the example of the vehicle while driving, without impairing the usage.

Claims

Patent claims

1.

Diagnostic module for creating a diagnosis for electrically controllable systems arranged within an overall system, the diagnosis being based on models of the overall system, with the features:

- Means for setting up a reduction graph with a head graph that only has fixed or measurable variables, the means comprising: means for breaking down the electrically controllable systems into components that include one or more basic building blocks, means for detecting the electrical connections between the basic building blocks and between the components, means for assigning discrete electrical state values to the basic building blocks, means for defining the operating states and the possible component states of the individual components in relation to the electrical state values of the basic building blocks belonging to a component, means for determining the and necessary for the diagnosis the basic building blocks and components measurable quantities, means for determining basic building blocks that have no influence on the defined measurable quantities and means for combining and eliminating the determined basic building blocks without any influence;

- Means for recording the measurable quantities necessary for diagnosis;

- Means for interpreting the measurable quantities in the head graph;

- Means for mapping the interpretation carried out in the header graph to the level of the basic building blocks and

- Means for mapping the interpretation mapped to the basic modules to the level of the components, these means indicating suspected and/or faulty components as a diagnostic result.

2.

Diagnostic module according to claim 1, characterized in that different models of the overall system can be used for the diagnosis and the models differ in the accuracy of the overall system description.

3.

Diagnostic module according to claim 1 or 2, characterized in that the model of the entire system can be taken directly from the construction documents.

4.

Diagnostic module according to one of claims 1 to 3, characterized in that means for detecting an error message are present in the diagnostic module and an error message can be interpreted by the means for interpreting the measurable variables in the header graph.

5.

Diagnostic module according to one of claims 1 to 4, characterized in that the suspected and/or faulty components can be sorted according to their probability of failure.

6.

Diagnostic module according to claim 5, characterized in that all information about the entire system that is available up to the generation of error statements about the entire system is included in the calculation of the probability of failure.

7.

Diagnostic device for an entire system with electrically controllable systems arranged within the overall system, with the features:

- Diagnostic module for creating a diagnosis based on system models for each electrically controllable system that can be dismantled into components according to one of claims 1 to 6, wherein the diagnostic module indicates suspected and/or faulty components as a diagnostic result;

- Means for creating a diagnosis for the overall system based on the functional relationships between the individual electrically controllable systems, with the means being able to create a graph consisting of two sub-graphs and the first sub-graph representing the structure of the components and the second sub-graph representing the relationships between the represents overall system functions and the functions of the individual components; and

- Means for generating error statements about the entire system and statements about impairment of use of overall system functions.

8th.

Diagnostic device according to claim 7, characterized in that means for creating diagnoses for subsystems are present in the diagnostic device and with these means a function graph can be created which depicts the function of the components as well as system variables describing the entire system.

9.

Diagnostic device according to claim 7 or 8, characterized in that means for controlling the dialogue with the users of the diagnostic device are present in the diagnostic device.

10.

Diagnostic device according to one of claims 7 to 9, characterized in that means for storing data, in particular diagnostic results and error statements, are present in the diagnostic device.

11.

Diagnostic device according to one of claims 7 to 10, characterized in that the means for creating a diagnosis for the entire system are designed in such a way that information about system variables that describe the normal function of the overall system, information that indicates malfunction and information that is in the Means for storing data are stored, are recordable.

12.

Diagnostic device according to one of claims 7 to 11, characterized in that the means generating the error statements comprise further means for suggesting remedial measures and the suggestions are accessible via the means for controlling the dialogue using the diagnostic device.

13.

Diagnostic device according to one of claims 7 to 12, characterized in that means for audiovisual input and output of information are present in the diagnostic device.

14.

Diagnostic device according to one of claims 7 to 13, characterized in that the diagnostic device can be used during normal operation of the entire system and continuous monitoring can be ensured over the entire service life of the entire system.