DE102017110620A1 - Fault diagnosis in an electrical network - Google Patents

Abstract

The invention relates to a method for fault diagnosis in an electrical network, wherein the electrical network has a plurality of electrical network components, with the following steps:
- Providing a simulation model for the electrical network, which contains at least for a part of the electrical network components in each case a network component model and for at least a part of the possible with respect to the network components electrical errors in each case a fault model, wherein
- Each fault model of a respective electrical network component as possible electrical states of the respective electrical network components each having a normal electrical state of the respective electrical network component and at least one electrical fault condition of the respective electrical network component, and
a plurality of mutually different possible network states is defined via the permutation of at least part of the respective possible electrical states of the respective electrical network components,
Determining the respective current electrical state of a plurality of the electrical network components of the electrical network, and
Determining the current network state by determining which defined network state defined in the simulation model has the individual currently determined electrical states of the electrical network components. Thus, such a method for diagnosing faults in an electrical network with a plurality of electrical network components, with which it is possible in a simple and reliable way to detect a fault relating to an electrical network component and to locate and identify.

Description

The invention relates to methods for fault diagnosis in an electrical network, wherein the electrical network has a plurality of electrical network components. In particular, the invention relates to a method with which it should be possible to detect, locate and identify all relevant faults in a DC power supply line network, in particular in a 48 V vehicle electrical system.

The errors that can occur in particular in a 48 V vehicle electrical system and are to be detected, i.a. Short circuits to ground, broken lines or unintentional opening of plug contacts, loose contacts, contact aging or deterioration of contact resistance by e.g. Corrosion and creeping short circuits that can cause fault currents. In addition to fault conditions, normal operating conditions should also be specifically detected.

The requirements for low-voltage power supply systems in motor vehicles are steadily increasing. By increasing the power requirement by new consumers, but also by greater power requirements of already known consumers, the established 12 V voltage level reaches its limits. In order to be able to reduce the cable cross-sections, a new low-voltage voltage level for motor vehicles (motor vehicles) was introduced at 48 V. 48 V electrical systems are also available for deviations from the rated voltage, by e.g. a high state of charge, even in the field of safety extra-low voltage.

In 48 V vehicle electrical systems, in contrast to conventional 12 V networks, in addition to the above-mentioned errors, the risk of arcing. For arcing faults in power supply networks, a distinction is made between serial and parallel arcs. Serial arcs are error conditions caused by e.g. Line breaks, poor contact of connectors due to corrosion or separation of current-carrying connectors may occur. They come in series to the load. Parallel arcs occur when a voltage carrying component (e.g., 48V wiring with damaged insulation) comes into contact with lower potential components (e.g., 12V or 0V / ground), and are typically significantly more current intensive. If the low potential contact is short enough and the resulting arc resistance is high enough that the inertia and current carrying capacity of the protective device (e.g., blade fuse) prevents it from tripping, a stable arc can result which, in extreme cases, can cause a vehicle fire. These arcs occur parallel to the load.

Some of the mentioned new ECUs bring significantly more safety-critical functions into the vehicle than previously available. Especially in the context of energy supply systems for highly automated or autonomous driving, the demands on reliability and availability are extremely high. Therefore, it is essential to integrate sufficient reliability rating for individual on-board electrical systems and a possibility of fault detection, localization and identification. Not only the consideration of electric arcs, but also diagnostic methods for any electrical system errors, are therefore necessary.

Some methods are already known from practice, with which errors, in particular serial and / or parallel arcs, can be detected in said environment. Some of these are based on identifying typical frequencies emitted by electric arcs in the on-board network or on specially coupled auxiliary signals in order to obtain a sufficient indication of the presence of an electric arc.

The DE 10 2014 206 925 A1 describes a method which is to ensure the arc detection by monitoring a self-coupled test signal. In this case, a test signal is coupled to a line by a transmitting unit and evaluated by a receiving unit. If the test signal does not reach the receiving unit, which can be checked, for example, by a handshake procedure, the presence of an arc or the interruption of the route can be concluded under certain conditions. An assessment of the changes in the received test signal may also provide evidence of an arc.

In the DE 103 59 532 describes a method that identifies the arc as a wideband noise source in a distribution network. For this purpose, a specific noise detector is constructed, which can distinguish the naturally occurring signals from those generated by an arc. The setup includes a bandpass filter controlled by a microprocessor that adjusts the corresponding frequencies of the filter. Thus, the spectrum can be examined in sections. In the DE 10 2004 037 193 A1 a similar approach is followed.

In the DE 101 32 952 B4 or A1, finally, a method is described which offers arc detection on the basis of the known arc voltage. A measuring device in front of a load measures the voltage applied to the load. As soon as a voltage drop occurs around the known arc voltage, the load should be switched off or, if possible, increased, in order to lower the arc current until the minimum arc current has been undershot and the arc extinguishes. However, since even in normal operating conditions high voltage dips, eg in the starting torque, may occur, the method provides a reconnection of the dropped loads. Assuming that a serial arc occurs, for example, due to cable breakage or bad / broken contact within connectors, there will be no supply to the monitored load after the serial arc is extinguished. The attempt to reconnect thus confirms or disproves the previous identification of a serial arcing fault.

Such a method is not appropriate depending on the connected control unit. Particularly safety-critical functions and control units must not be switched off without the presence of an error. Also, the recurrent normal state after triggering and re-switching to the sample is not sufficient indication that no serial arc has occurred. There is a possibility that, in the case of serial arcs, a reconnection by e.g. Welding the loose contacts or broken cable ends is made. Nevertheless, the fact that an arc has arisen must be detected in order to detect possible late effects or resulting errors, e.g. Short circuits due to damaged insulation can be estimated. In addition, the voltage supply of the measuring control unit must be ensured and must not be done via the cable affected by the arc fault, otherwise the functionality can not be ensured.

The solutions known from the state of the art are thus common in that a device-intensive effort must be made to diagnose the fault of an electrical network without the diagnosis leading to satisfactory and reliable results. In particular, fault conditions which concern a supply of the measuring means used have the effect that the recognition according to the patents presented can no longer be carried out.

It is therefore the object of the invention to provide such a method for fault diagnosis in an electrical network with a plurality of electrical network components, with which it is possible in a simple and reliable manner to detect, locate and identify a fault with respect to an electrical network component.

This object is solved by the subject matter of claim 1. Preferred embodiments of the invention are described in the dependent claims on claim 1.

• - Bereitstellen eines Simulationsmodells für das elektrische Netzwerk, das wenigstens für einen Teil der elektrischen Netzwerkkomponenten, vorzugsweise für alle elektrischen Netzwerkkomponenten, jeweils ein Netzwerkkomponentenmodell sowie für wenigstens einen Teil der bezüglich der Netzwerkkomponenten möglichen elektrischen Fehler jeweils ein Fehlermodell enthält, wobei
  • - jedes Fehlermodell einer jeweiligen elektrischen Netzwerkkomponente als mögliche elektrische Zustände der jeweiligen elektrischen Netzwerkkomponenten jeweils einen elektrischen Normalzustand der jeweiligen elektrischen Netzwerkkomponente und wenigstens einen elektrischen Fehlerzustand der jeweiligen elektrischen Netzwerkkomponente aufweist, und
  • - eine Mehrzahl von voneinander verschiedener möglicher Netzwerkzustände über die Permutation wenigstens eines Teils der jeweils möglichen elektrischen Zustände der jeweiligen elektrischen Netzwerkkomponenten definiert ist,
• - Bestimmen des jeweiligen aktuellen elektrischen Zustands einer Mehrzahl der elektrischen Netzwerkkomponenten des elektrischen Netzwerks, und
• - Ermitteln des aktuellen Netzwerkzustands durch Feststellen, welcher im Simulationsmodell definierte mögliche Netzwerkzustand die einzelnen aktuell bestimmten elektrischen Zustände der elektrischen Netzwerkkomponenten aufweist.

In this respect, the invention is therefore in a method for fault diagnosis in an electrical network, wherein the electrical network has a plurality of electrical network components, with the following steps:

• - Providing a simulation model for the electrical network, which contains at least for a portion of the electrical network components, preferably for all electrical network components, each a network component model and for at least a portion of the possible with respect to the network components electrical faults each a fault model, wherein
  • - Each fault model of a respective electrical network component as possible electrical states of the respective electrical network components each having a normal electrical state of the respective electrical network component and at least one electrical fault condition of the respective electrical network component, and
  • a plurality of mutually different possible network states is defined via the permutation of at least part of the respective possible electrical states of the respective electrical network components,
• Determining the respective current electrical state of a plurality of the electrical network components of the electrical network, and
• Determining the current network state by determining which defined network state defined in the simulation model has the individual currently determined electrical states of the electrical network components.

In the method according to the invention, a special feature is that it is possible to dispense with introducing new control devices or measuring devices into the electrical network. Rather, it is possible, for example in an electrical system with electrical control units, to use existing sensors in the control units in order to realize state detection and fault diagnosis by the method according to the invention. A further special feature of the method according to the invention is that the sensor system does not have to be in the immediate vicinity of the electrical network or subnetwork to be observed, as is the case with a majority of the known concepts, but also that measured values recorded in other branches are due to their sensitivity against the changes in the network to be detected.

The electrical states of the network components may in principle be different properties that characterize a network component in electrical terms. According to a preferred embodiment of the invention, however, it is provided that the electrical state of a network component is a current value and / or a voltage value and / or a resistance value. This means that a respective current value in amps, voltage value in volts and / or resistance value in ohms for this network component is assigned to both the normal state and each fault state of an electrical network component.

Preferably, the method of the invention serves to determine what type of fault (e.g., short circuit, open circuit, etc.) exists in exactly which network component. For this purpose, according to a preferred development of the invention, it is provided that the possible network states defined in the simulation model are each assigned a description of which type of fault currently exists on which electrical network component.

According to a preferred embodiment of the invention, however, it is also provided that in the simulation program possible network states, whose associated electrical states of the electrical network components each deviate from one another by a respective predetermined value, form a common class. In this way, network conditions may be combined with similar electrical states of the network components to at least limit the nature of the fault or faults in the electrical network.

In principle, it is naturally helpful if the simulation model models as many as possible or all the electrical network components of the electrical network. Preferably, the simulation model takes into account as network components at least on the one hand a battery and / or a generator and on the other hand a plurality of consumers and a plurality of lines between the consumers themselves and between the consumers and the battery and / or the generator.

In the permutation of at least part of the respective possible electrical states of the respective electrical network components, basically all states of each electrical network component can be combined with all states of all other electrical network components. However, the permutation can be simplified, e.g. by making the assumption that two errors never occur at the same time. In this respect, it is provided according to a preferred development of the invention that when permuting at least part of the respective possible electrical states of the respective electrical network components only those permutations are used which have an electrical fault state for only a single electrical network component.

Furthermore, it is provided according to a preferred embodiment of the invention that the respective current electrical states of a plurality of the electrical network components of the electrical network are at least partially transmitted via a bus from the respective electrical network components to a diagnostic device in which the determination of the current network state is performed. Such buses are used in typical motor vehicle electrical systems.

Moreover, the invention also relates to the use of a method, as described above, for an electrical system of a vehicle as an electrical network, preferably for the electrical system of a motor vehicle, very particularly preferably for a 48 V vehicle electrical system.

In this case, the network components preferably include control devices with current and / or voltage sensors, wherein the determination of the respective current electrical state of a plurality of the electrical network components of the electrical network, the detection of current and / or voltage values by means of these current and / or voltage sensors. Very particularly preferred is in this Connection, that the detection of current and / or voltage values takes place exclusively by means of these current and / or voltage sensors. In this way, eliminates the need to install own sensors for the inventive method, which further simplifies matters.

• 1 ein Schema einer modellbasierten Überwachung und Fehlerdiagnose in einem Kfz-Bordnetz,
• 2 schematisch ein Bordnetz gemäß einem bevorzugten Ausführungsbeispiel der Erfindung mit der Besonderheit einer fehlenden Messung in einem fehlerbehafteten Strang,
• 3 verschiedene Netzwerkkomponentenmodelle und Fehlermodelle gemäß dem bevorzugten Ausführungsbeispiel der Erfindung,
• 4 ein Gesamtmodell des Bordnetzes gemäß dem bevorzugten Ausführungsbeispiel der Erfindung,
• 5 Variationen der Batteriespannung in Abhängigkeit der berücksichtigten Gesamtsystemzustände gemäß dem bevorzugten Ausführungsbeispiel der Erfindung,
• 6 Berechnung der Fehlernormen für einen bestimmten Netzwerkzustand und eine Darstellung der Fehlernormen als Histogramm gemäß dem bevorzugten Ausführungsbeispiel der Erfindung,
• 7 der Verlauf der Netzwerkzustände gemäß dem bevorzugten Ausführungsbeispiel der Erfindung und
• 8 ein Ergebnis der Klassifikation und einen Vergleich der klassifizierten Netzwerkzustände gemäß dem bevorzugten Ausführungsbeispiel der Erfindung.

The invention will be further described by means of preferred embodiments with reference to the drawings. In the drawings show:

• 1 a schematic of a model-based monitoring and fault diagnosis in a motor vehicle electrical system,
• 2 1 schematically shows a vehicle electrical system according to a preferred exemplary embodiment of the invention with the special feature of a missing measurement in a faulty line;
• 3 various network component models and error models according to the preferred embodiment of the invention,
• 4 an overall model of the electrical system according to the preferred embodiment of the invention,
• 5 Variations of the battery voltage as a function of the considered overall system states according to the preferred embodiment of the invention,
• 6 Calculation of the error standards for a specific network state and a representation of the error standards as a histogram according to the preferred embodiment of the invention,
• 7 the course of the network states according to the preferred embodiment of the invention and
• 8th a result of the classification and a comparison of the classified network states according to the preferred embodiment of the invention.

Basis for the described invention is that in today's automotive environments already a very large number of voltage and current sensors is present. Control units are increasingly equipped with these because they have to monitor their own electrical behavior. Thus, each individual control unit ensures a certain diagnostic capability. In addition, versatile communication via different bus systems is basically possible for almost every ECU.

On this basis, the existing sensor information can be communicated to a central control unit. There, the monitoring and diagnostic concept presented below realizes the monitoring, fault detection, localization and identification. 1 FIG. 12 shows a schematic of a model-based monitoring and fault diagnosis in a motor vehicle electrical system in order to implement the model-based monitoring and the model-based fault diagnosis according to the present exemplary embodiment.

On the one hand, the real vehicle electrical system exists 1 which can supply current and / or voltage values at various locations or in the control units. These values are summarized in the vector x = (v ₁ , ..., i ₁ , ...) ^T. Including in 1 you find the electrical system model 2 with all considered error models. By using a simulation model 3 can all associated error vectors x _i determined and a clustering 4 be subjected. By comparing the measured state with the possible calculated states in the context of a classification, it is possible to determine whether an error exists and this can be limited and output 6 be issued.

To use this model-based method, it is necessary to have an overall model for the real vehicle electrical system to be monitored 1 to create. In the simplest case, a model is sought, which supports a static analysis. This means that only sources and resistances have to be used to model. With these submodels can now the real electrical system 1 in the electrical system model 2 be transferred.

Behavioral variations of the real components are understood as discrete states. These discrete states may be passed to the models in the form of a parameter variation. To the mentioned errors in the electrical system model 2 To be able to take into account, fault models must be created and at the points to be monitored in the electrical system model 2 be inserted. A distinction between error-free state (normal state) and error state can also be made by a parameter variation in the error models. It must be ensured that there is a state of the error models that has no influence on the rest of the network and thus represents a normal state. If, for example, a modified node potential analysis (MNA) is used for the simulation, the modeling for a static simulation can yield an algebraic equation system of the following form: $$G\cdot \widehat{x}=b$$

Here G describes the admittance matrix and b the excitation vector. The solution vector x = (v ₁ ..., î ₁ , ...) describes the predicted current and voltage values at different points in the network. The solution vector x may have more components than the measurement vector x.

Based on the overall system model, the component states can be permuted. This permutation produces a permutation matrix P, which may have the following form, for example: $$P=\left[\begin{array}{c}1\\ 1\\ 1\\ 1\\ 1\end{array}|\begin{array}{ccc}1& & 1\\ 1& & 2\\ 1& \cdots & 2\\ 1& & 3\\ 2& & 1\end{array}\right]$$

The example shows a permutation matrix for a wiring system model with five electrical elements. In this case, the lines represent variations of the component states and the columns total system states of the electrical system. In the environment of classification and clustering methods described later, the current and voltage measurements are taken as characteristics of the respective state and assigned to the overall system states defined in the permutation matrix. A theoretical number of system states can be specified with the following rule: $${\text{\#}}_{states}={\displaystyle \underset{n=1}{\overset{N}{\Pi }}{n}^{{\text{\#LD}}_n}}$$

Here, the number of states of the total system # _{states is} calculated. n represents the number of component states, while #Id _{n represents} the number of n-state controllers. N describes the number of states that the component with the most states has.

After a complete permutation, various simplifying assumptions can reduce the number of system states. One assumption may be that two errors never occur simultaneously, so that two or more error models with a state describing an error do not occur simultaneously in one column of the permutation matrix.

After the definition of the overall system states in the permutation matrix P, the method states according to the present preferred exemplary embodiment of the invention can be used to simulate the considered system states and thus create solution vectors x. The totality of all solution vectors to the various system states can be summarized in the solution matrix X, as described above 1 is described.

In principle, each predicted solution vector x can be understood as a separate class to which the measurement vector x is assigned in the classification (and thus also the states / columns in the permutation matrix). However, as a preprocessing can be a clustering 4 , as in 1 shown, are provided. This area is used for a closer examination of the considered states. Here an evaluation of the observability can be made. While different states may have large differences, some states may also have high similarities. These similarities can be identified and evaluated through grouping or class formation. Grouping can be done on the basis of different cluster algorithms. A special feature is that the reasonable number of classes can not be specified as a whole. Therefore, a combination of unsupervised and supervised procedures should be used.

One possibility is the k-nearest-neighbor algorithm in conjunction with the Elbow method. In this combination of methods, the k-nearest-neighbor algorithm with values for k starting at 1 is used successively. The quality of each variant is evaluated by calculating the deviation of the assigned states to the class center of gravity. A minimum quality is achieved with k = 1, a 100% accuracy with k = number of columns of the permutation matrix. As soon as the improvement of the quality in the comparison of the kth and (k + 1) iteration falls below a limit value, the number of classes regarded as optimal is set to this value. The limit to be set depends i.a. together with the available measurement resolution.

In addition to this method combination, there are a large number of other possibilities that can also be used in principle.

After class formation, individual observability of specific points in the electrical system can be assessed individually. If there is no class that contains both normal and error states, the error can be assumed to be completely detectable. If the different error states each form their own classes, a complete localizability and identifiability can additionally be assumed. If mixed classes have formed, a statement can be made as to which states can not be distinguished from one another.

Both component modeling and overall system modeling, permutation and simulation of states, as well as class formation can form the basis of classification and need not be done online. Only the result of these processing steps is implemented in a control unit (observer control unit). This considerably reduces the required computing power. The task to be performed online is solely to assign the measurement vector x to the previously defined classes. This classification is determined by the classification 5 in 1 represents.

In principle, many different classification algorithms can be applied and lead to the goal. For example, the simplest way is to create error norms (eg, f ₁ = || x ₁ -x ||) between all simulated states and the measurement vector without prior class formation or summary of various states. The least deviation between the measurement vector and the variables created in the simulation forms the most probable state of the overall system. In addition to on, localization and identification, the method also offers the possibility of identifying a normal state.

The information taken into account need not be limited to current and voltage values. Some controllers lack sensors, so that the current or voltage information is not immediately available. However, many of these special control units have the option of communicating their own status via a bus system. Assuming that this state information is error-free, it can be used in the observer controller in which the data fusion takes place in order to exclude certain states or classes from the classification. If more precise component models of these special loads or control units are available, the corresponding current and voltage values in the classification can also be taken into account by means of ancillary simulations. This increases the number of properties and at the same time reduces the number of classes, which considerably increases the probability of a successful classification.

To implement a measured value communication, different strategies can be pursued independently of the bus actually used. A preferred case is the constant supply of the evaluating control unit with all available measured values. The temporal resolution should be as high as possible in order to be able to recognize even briefly occurring changes in state. The recorded measured values can be stored in an array by the respective control devices with a specific memory depth. This array can be organized like a shift register so that the oldest metric must go astray once the array is completely filled with metric. The central control unit inquires successively the required data. The new data packet of the first control unit, which repeatedly sends data must, in order to ensure seamless monitoring, still contain at least one value already known to the central control unit. Another possibility is that the sending ECUs instead of waiting for a query, independently send the data one after the other. The transmission permission or the transmission request can be forwarded cyclically by the control units.

Important here is the sending of the data packets, consisting of a number of measured values depending on the selected memory depth, together with time stamps derived from a common time base. A prior clock synchronization is therefore necessary, and a subsequent measurement value synchronization in the central control unit thus possible. Different resolutions can be adjusted by interpolation.

For the method according to the presently described preferred embodiment of the invention, it is initially only necessary for all available data to be available at a fixed point in time. Therefore, another possibility of the communication organization is to send a broadcast message of the central control unit to all measuring units, consisting of a start signal and a common time base. From this point, the ECUs start recording readings with a defined resolution (if provided) and a defined memory depth (if provided) complete filling of the defined memory space, as described above, send back to the central control unit. This procedure leads to a significant reduction of the bus load.

Another possibility, which in turn results in a further reduction of the bus load, provides that the measuring control units independently then send their measured values to the central control unit as soon as their condition or their electrical behavior deviates from the previous one by a certain size. This method has the disadvantage that a failure of the communication could not be detected and a classification would continue to use a larger number of possible classes.

• - Es werden nur Komponenten verwendet, die bereits in einem Bordnetz vorhanden sind:
  • - Messtechnik, die aufgrund von der geforderten Eigendiagnosefähigkeit der Steuergeräte, bereits vorhanden ist;
  • - Implementierung des Klassifikationsalgorithmus kann auf einem bereits vorhandenen Steuergerät, wie z.B. dem Gateway, Batteriemanager oder Energiemanager, geschehen;
  • - Nutzung der vorhandenen Bus-Systeme;
• - Auf Basis der zur Verfügung stehenden Sensoren kann eine Bewertung der Beobachtbarkeit einzelner Stellen im Bordnetz gewährleistet werden; so kann die Methode im Vorhinein eine Aussage über die eigene Fähigkeit, einzelne Fehler- und spezifische Normalzustände zu identifizieren, treffen.
• - Es wird nicht für einen bestimmten Fehler an einem bestimmten Ort ein bestimmter Sensor verwendet oder sogar gefordert, sondern die Fähigkeit aller Sensoren wird darauf geprüft, Fehler an verschiedenen Stellen zu erkennen; so wird die Empfindlichkeit verschiedener bzw. aller Sensoren auf Änderungen im gesamten Bordnetz genutzt.
• - Das Ergebnis einer Klassifikation ist immer ein spezifischer Zustand. Es werden keine binären Aussagen über bestimmte Zustände getroffen (z.B. Fehlerzustand vorhanden/Fehlerzustand nicht vorhanden).

The presently described monitoring and diagnostic method offers several important advantages compared to the concepts discussed above from the known patent applications:

• - Only components that are already present in a vehicle electrical system are used:
  • - Measurement technology that already exists due to the required self-diagnostic capability of the ECUs;
  • - Implementation of the classification algorithm can be done on an existing control unit, such as the gateway, battery manager or energy manager;
  • - use of existing bus systems;
• - On the basis of the available sensors, an assessment of the observability of individual points in the electrical system can be ensured; In this way, the method can make a statement in advance about its own ability to identify individual fault and specific normal states.
• - A particular sensor is not used or even required for a particular fault in a particular location, but the ability of all sensors is checked to detect faults in different locations; The sensitivity of different or all sensors is used for changes in the entire vehicle electrical system.
• - The result of a classification is always a specific condition. No binary statements about certain states are made (eg error state present / error state not present).

The preferred embodiment for a static simulation described here represents the simplest form of the application. Extensions to dynamic analyzes are easily possible by impressing a test signal at a central or suitable point in the vehicle electrical system or by using an existing time-varying signal. Current and voltage curves are then used as characteristics of the states. The ability to assess the observability of specific points in the on-board network with regard to defined possible errors can be a basis for harnessing the method in its best form. Measuring points can be identified in which supplementing the entire system with different measurements can provide considerably better observability and thus more robust error detection.

Hereinafter, a preferred embodiment of the invention will be described in detail. In this case, the method is reduced to the observation of a special, critical line for better understanding, with two possible errors being taken into account. The presented example is scalable and the complexity of the network can be extended as required. The number of possible errors can also be extended depending on existing models. Likewise, the number of measuring points can be adapted to the respective individual requirements.

The present example deals with the detection of line breaks or triggered fuses and serial arcs. The described exemplary embodiment is based on a motor vehicle electrical system 1 with an energy storage 7 , a generator 8th and a first consumer 9 and a second consumer 10 out. The network components are connected via a neutral point with lines 11 . 12 . 13 and 14 interconnected.

To clarify the key advantage of the present method, all network components have power and voltage monitoring, except for the controller which is from the line affected by a fault 14 is supplied (redundant information is neglected). These values can be sent via bus systems to a central location. The scenario considered first sees a fault in the line 14 in front. If measurements were available at this point, a communication abort indicates an error. Thus, possible states can still be limited. The fundamental lack of measurements at this point thus has an additional difficulty. The procedure shows an example of the targeted investigation of a line for possible sources of error.

$$v_B=\left\{40V,45V,50V\right\}$$

The models used for the DC simulation are described below with reference to FIG 3 described. The battery will, as in 3a ) shown in the context of an operating point analysis as a voltage source in conjunction with an internal resistance. Both the source voltage and the internal resistance vary, for example, based on the battery type, the SoC and the SoH. Therefore, as mentioned in the description of the method, parameter spaces are given for the rating. In the examples given, only a limited number of possible states are taken into account for reasons of simple traceability. In the present model, the states of the battery in terms of resistance and voltage are: $$R_B=\left\{10m\mathrm{\Omega }.100\text{m}\mathrm{\Omega }\right\}$$

$$v_B=\left\{40V.45V.50V\right\}$$

$$v_G=\left\{45V,50V,55V\right\}$$

The generator will, as in 3b ) modeled as a DC motor for a DC simulation with a voltage source and a resistor. The resistance represents the ohmic portion of the winding, the voltage source represents the induced voltage. Depending on the operating strategy and regulation in the vehicle, the voltage of the voltage source varies. The ohmic portion is assumed to be constant. In the present model, the states of the generator in terms of resistance and voltage are: $$R_G=120m\Omega$$

$$v_G=\left\{45V.50V.55V\right\}$$

• R = {10Ω, 20Ω) oder
• R=15Ω

Consumers will, as in 3c ), considered as pure power consumers in the DC simulation. The value of the equivalent resistance varies depending on the consumer characteristics. The following possible characteristics are used in the present model:

• R = {10Ω, 20Ω) or
• R = 15Ω

• R = 15mΩ oder
• R = 10mΩ oder
• R = 20mΩ

Fuses, contacts and line resistances are in the in 3d ) are summarized. The sum of the individual resistances then forms the total resistance. This can be calculated by considering eg material, contact surfaces or cross sections. Accepted are:

• R = 15mΩ or
• R = 10mΩ or
• R = 20mΩ

• R = {lmfl, 100MΩ}

The realization example 3e ) explicitly shows the consideration of wire breaks, unintended openings of plug contacts or triggered fuses. The resistance model considers a fault condition and a normal condition within the parameter variation with the states:

• R = {lmfl, 100MΩ}

• v_arc = {0V, 15V}

As a further cause of a ECU and thus communication failure is a serial arc in question. The typical arc voltage can be mapped in a DC simulation by a voltage source, as in 3f ), with the states:

• v _arc = {0V, 15V}

From the presented submodels it is possible to create an overall system model for the present vehicle electrical system, as in 4 shown. 4 shows the overall on-board network model based on the component models and error models for DC analysis with targeted fault investigation in a failed string with the consumer R33 and serial arc and line break or fuse blown. The available vector x includes the battery voltage, the generator voltage, the Consumer voltage at node 4 , the battery current and the generator current x = (v ₁ , v ₂ , v ₄ , i ₁ , i ₂ ) ^T - redundant information is not taken into account.

By way of example, the variation of a measured value, in this case the measured battery voltage, is shown as a function of the different states, see 5 , The large variance in the y-direction shows that a distinction is possible. In addition, the similarity of some state features shows that a complete distinction based solely on this measurement can not be fully guaranteed. Therefore, the consideration of further measuring points offers a great added value and a strong improvement.

In 6 By way of example, for the classification of a measured state, ie for a specific vector x, the calculation of the error norms is presented. For the sake of clarity and clarity, the classification about the error norms with all possible states, without reduction due to similarities between states that are difficult to distinguish, is shown.

The representation of the error norms shows that some simulated states have greater similarities with the measurement vector. The state marked on the bottom left has the smallest deviation and can therefore be considered the most probable state.

The scenario in the presented example on-board network provides that the simulation or test time 10 s is. At the time t = 1 s, a serial arc occurs in the affected line, and from t = 5 s, the corresponding load is no longer supplied due to a break in the line or a fused fuse. The remaining states remain constant over the entire simulation or test time. In 7 is the result of the classification, applied to the shown on-board network with the described scenario shown.

The course of the overall states indicates changes at the expected times. The classified states are 9, 10 and 23. A comparison of the states is given in the 8th shown in the table. While state 9 has been identified as a healthy condition, can change the states 10 and 23 the corresponding errors "serial arc" or "line break" are assigned.

The invention underlying this patent application was developed in a project which was funded under the grant number D-11-41883-001-081082 by the BMBF.

LIST OF REFERENCE NUMBERS

Real on-board electrical system 1 Wiring system model with fault models 2 simulation model 3 clustering 4 classification 5 output 6 energy storage 7 generator 8th first consumer 9 second consumer 10 cables 11, 12, 13, 14

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• DE 102014206925 A1 [0007]
• DE 10359532 [0008]
• DE 102004037193 A1 [0008]
• DE 10132952 B4 [0009]

Claims (10)

A method of fault diagnosis in an electrical network, wherein the electrical network comprises a plurality of electrical network components, comprising the following steps: - Providing a simulation model for the electrical network, which contains at least for a part of the electrical network components in each case a network component model and for at least a part of the possible with respect to the network components electrical errors in each case a fault model, wherein - Each fault model of a respective electrical network component as possible electrical states of the respective electrical network components each having a normal electrical state of the respective electrical network component and at least one electrical fault condition of the respective electrical network component, and a plurality of mutually different possible network states is defined via the permutation of at least a part of the respectively possible electrical states of the respective electrical network components, Determining the respective current electrical state of a plurality of the electrical network components of the electrical network, and Determining the current network state by determining which defined network state defined in the simulation model has the individual currently determined electrical states of the electrical network components. Method according to Claim 1 wherein the electrical state of a network component is a current value and / or a voltage value and / or a resistance value. Method according to Claim 1 or 2 , wherein the possible network conditions defined in the simulation model are each assigned a description of which type of fault currently exists on which electrical network component. Method according to one of the preceding claims, wherein in the simulation program possible network states whose respective electrical states of the electrical network components each deviate from one another by a respective predetermined value form a common class. Method according to one of the preceding claims, wherein the simulation model takes into account as network components a battery, a generator, a plurality of consumers and / or a plurality of lines. Method according to one of the preceding claims, wherein in the permutation of at least a portion of the respective possible electrical states of the respective electrical network components only those permutations are used which have an electrical fault state for only a single electrical network component. Method according to one of the preceding claims, wherein the respective current electrical states of a plurality of the electrical network components of the electrical network are at least partially transmitted via a bus from the respective electrical network components to a diagnostic device, in which the determination of the current network state is performed. Use of a method according to one of Claims 1 to 7 for a vehicle electrical system of a vehicle as an electrical network, preferably for the electrical system of a motor vehicle, very particularly preferably for a 48 V vehicle electrical system. Use after Claim 8 wherein the network components comprise control devices with current and / or voltage sensors, and determining the respective current electrical state of a plurality of the electrical network components of the electrical network, the detection of current and / or voltage values, preferably exclusively, by means of these current and / or Voltage sensors done. Use after Claim 9 , wherein the detection of the current and / or voltage values takes place exclusively by means of these current and / or voltage sensors of the control devices.

Images