Fast evaluation of operating state of technical system

Abstract

The invention relates to a method for evaluating the operating state of a technical system, characterized by at least one total variable Q, the value of which is obtained by a plurality of elementary variables e1,..., en corresponding to the configuration interaction of the technical system, comprising the steps of: providing an effect function indicating the dependency of Q on the elementary variables; factorizing the effect function into a product of a plurality of contributions, the plurality of contributions being dependent on different subsets Ei of the elementary variables; creating a graph G, the nodes of which correspond to the contributions fi and the edges of which correspond to the elementary variables, in which # imgabs0 # issues, from each node corresponding to the contribution fi, an edge relating to each elementary variable on which the contribution fi depends; and # imgabs1 # connecting two nodes corresponding to such contributions fi by at least one side corresponding to the elementary variables on which the two or more contributions fi depend; determining a Laplacian matrix L of G; and determining the extreme value characteristic value alpha of L as an evaluation of the searched operating state.

Description

Technical Field

[0001] The invention relates to the evaluation of the operating state of a technical system, which evaluation can have been carried out in advance, for example, in a simulation, with the aim of operating the system optimally. Background Art

[0002] When operating many technical systems, the quality of operation is measured by specific parameters. Thus, for example, when operating electronic circuits, it is desirable that the temperature of the circuits or specific components does not exceed a predetermined value in order to avoid excessively shortening the service life of capacitors or semiconductor components in particular.

[0003] The parameters can usually be measured during operation so that a reaction to the measured values ​​can be made. However, it is desirable to be able to predict the values ​​of the parameters for a specific configuration of the technical system in order to be able to take countermeasures in good time if necessary, or even to be able to optimize the structural design of the technical system with respect to the parameters even before the physical realization of the technical system. The more complex the relationships within the technical system that lead to the final value of the parameter, the greater the computational effort required for the prediction. This in turn affects how far into the future it is possible to predict given a given computational effort budget. Summary of the invention

[0004] The present invention provides a method for evaluating the operating state of a technical system. The operating state is characterized by at least one total variable Q. The value of the total variable Q is determined by a large number of basic variables e 1 , ..., e n The interaction results from the configuration of the technical system. In the example mentioned at the beginning, for example, the temperature of an electronic circuit as a total variable Q is derived from where in the circuit heat is generated and how it is dissipated. This interaction is represented by the effect function f(e 1 , ..., e n ) describes the effect function of the total variable Q on the basic variable e 1 , ..., e n dependency.

[0005] Therefore, the method provides the effect function f(e 1 , ..., e n )start.

[0006] Then, the effect function f(e 1 , ..., e n ) is factorized into multiple contributions f i (E i ) Here, these contributions i (E i ) depends on the basic variable e 1 , ..., en Different subsets of E i Thus, for example, the effect function f(u, w, x, y, z) can be rewritten as the product f(u, w, x, y, z) = f 1 (u, w, x)·f 2 (x, y, z)·f 3 (z), so the subset of basic variables E i For E 1 ={u,w,x},E 2 =(x, y, z} and E 3 = {z}. As shown in this example, multiple subsets E i Of course they can overlap.

[0007] Based on this factorization, a graph G is created. The nodes of this graph G correspond to contributions f i The edges of the graph G correspond to the elementary variables e 1 , ..., e n . From each corresponding contribution f i The nodes will send out the contribution f i Each basic variable e depends on 1 , ..., e n Here, it is not necessary for each edge to connect two nodes. Instead, it is possible for an edge to start from one node and leave the other end open. However, if two or more contributions f i Depends on the same basic variable e 1 , ..., e n , then it corresponds to such a basic variable e 1 , ..., e n At least one edge connects two corresponding contributions f i Then, for example, you can use the same basic variable e 1 , ..., e n Assign to one or more links connecting two contributions f i The same basic variable e 1 , ..., e n Assigned to one or more edges with open ends. In this case, the dependency does not have to be direct (explicit), but can also be implicit, for example via an intermediate result, in which other related elementary variables e then appear. 1 , .. , e n .

[0008] Therefore, the graph G can be used in particular to reflect, for example, the basic variable e 1 , ..., e nThe sensor system may be at least partially redundant. Thus, for example, the detection areas covered by the multiple sensors may partially overlap. Thus, the values ​​measured by the multiple sensors may vary in a time-dependent manner and / or there may be time-dependent correlations between the values.

[0009] The Laplace matrix L of the graph G is determined. The Laplace matrix L contains the degrees of the nodes on its diagonal, which indicate the number of connections of the corresponding node to other nodes. The off-diagonal elements of the Laplace matrix L indicate the adjacency of the nodes to each other. The adjacency can in particular contain a statement about whether a connection exists between two given nodes, and can also optionally weight the connection. Therefore, the Laplace matrix L is uniquely determined for each graph G.

[0010] The extreme (ie, largest or smallest) eigenvalue α of the Laplace matrix L is determined as an estimate of the operating state being sought.

[0011] It has been recognized that this extreme eigenvalue α is proportional to the total variable Q being sought and in particular assumes an extreme value precisely when Q also assumes an extreme value. Therefore, if the exact value of Q is not required to evaluate the operating state, but only the point at which Q assumes an extreme value, information about the extreme value of the eigenvalue α can be obtained much faster than by calculating the course of Q and then determining the extreme value from this course of change. The understanding behind this is that when from the original relationship Q=f(e 1 , ..., e n ) to the extreme values ​​of the eigenvalue α, only linear transformations are used. This simplification does not affect the accuracy with which the extreme value positions can be determined for any variable. At the same time, this simplification introduces a level of abstraction, which eliminates dependencies on irrelevant details of the technical system from the outset.

[0012] If not only the extreme value position but also the maximum or minimum value of the total variable Q is required, the conventional method can be used to directly obtain the maximum or minimum value of the basic variable e. 1 , ..., e n The value at the extremum location is calculated. This is relatively large in terms of computational effort, but this effort only happens once at the extremum location, whereas it has happened thousands of times so far in the search for extrema. Therefore, the computational time savings from a one-time explicit evaluation of Q are barely reduced.

[0013] A significantly faster evaluation of the extreme characteristic value α is particularly advantageous when the question is how the total variable Q changes depending on the configuration of the technical system. In particular, in many applications of technical systems the question is how the configuration of the technical system is to be designed in order to achieve an optimal value for the total variable Q. The search for the corresponding optimal value requires a large number of evaluations of the direction in which the total variable Q moves in response to a specific change in the configuration. If this is necessary each time based on the effect function Q=f(e 1, ..., e n ) explicitly evaluates Q, this would take too long.

[0014] In a particularly advantageous embodiment, a plurality of candidate configurations of the technical system are created, which differ in the basic variable e 1 , ..., e n The values ​​of these basic variables e 1 , ..., e n to form the overall variable Q. The resulting operating states are then evaluated for each candidate configuration.

[0015] In this way, the candidate configuration leading to the best operating state can be determined. In particular, the rapid evaluation of the operating state via the extreme eigenvalue α can be used as a proxy for the basic variable e 1 , ..., e n Feedback of optimization of at least one value of and / or at least two basic variables e 1 , ..., e n Feedback for the optimization of the interaction of the two. By aligning the optimization with the goal of improving the evaluation of the operating state resulting from the thus changed configuration of the technical system, the evaluation must be calculated very frequently. This applies in particular in the context of gradient-based optimization methods that cannot be used. A significant advantage of gradient-based methods is that the evaluation of the objective function (here: effect function f) can be planned in a targeted manner, so that the total number of these evaluations can be advantageously reduced. However, the optimization of the candidate configurations discussed here is a mixed integer problem, since the candidate configurations cannot be optimized continuously at many points, but only discretely. Thus, for example, two or three fans can be installed, but 2.5 fans cannot be installed.

[0016] In order to produce a new configuration of the technical system, in a further advantageous embodiment, at least one component of the technical system is

[0017] - Replace with another component, and / or

[0018] - placed in a location, and / or

[0019] -Manipulate in other ways.

[0020] For example, a component that releases heat can be replaced by a component that releases less heat. Conversely, for example, a heat-sensitive component can be replaced by a less heat-sensitive component, thereby allowing higher temperatures in its environment.

[0021] By placing the heat-generating component at another location in the technical system, the goal can be achieved, for example, that the heat from the component at this other location can be dissipated better so that it does not accumulate. Conversely, for example, a fan or other cooling element (e.g. a vapor chamber) can be positioned so that it can dissipate the most heat.

[0022] This problem can already occur, for example, when configuring a PC. If the configuration already contains many components and a powerful graphics card, for example, adding another SSD mass storage device at a distance that seems to have nothing to do with the graphics card can have the effect that the heat from the graphics card accumulates at the SSD mass storage device and eventually overheats both components.

[0023] For example, a component can also be controlled in another way with the goal of generating less heat. Thus, for example, the heat output of a processor increases in a highly nonlinear manner as the clock frequency at which the processor is operated increases. In the case of frequency converters using pulse width modulation, the power loss and the resulting heat are strongly dependent on the switching frequency and on the phase control factor of the pulse width modulation.

[0024] It is particularly advantageous if the candidate configuration with the best evaluation of the operating state resulting therefrom and / or the optimal configuration found within the scope of the optimization is implemented in the hardware of the technical system. In this way, it is ensured that the advantages of each selected configuration are actually also reflected in the operating state of the physical system. The implementation in hardware can, for example, in particular include that the technical system is composed of hardware components based on the selected configuration and that these hardware components are spatially arranged according to the selected configuration.

[0025] According to what has been said above, in a particularly advantageous embodiment, the total variable Q is a measure of at least one heat flow to be transferred into or out of the technical system and / or a measure of at least one temperature at a predetermined location in the technical system. It is precisely for these thermal variables that a fast simulation analysis with the aid of extreme characteristic values ​​α is advantageous, since experimental analysis in the laboratory is extremely slow due to the inactive cooling and heating processes at the heat capacities located in the system.

[0026] Thus, in particular, for example, at least one elementary variable e 1 ,...,e n A measure of the heat conductivity and / or transport capability of heat within a technical system may be expressed as heat flow from a heat source, heat flow into a heat sink, and / or heat flow.

[0027] In a further particularly advantageous embodiment, the total variable Q is a measure of the probability of at least one undesirable event occurring in the technical system, for example a "system failure". The method can therefore also be used to evaluate the fault tree leading to the undesirable event significantly faster. This in turn can be used in the above-described manner, for example, to optimize the configuration of the technical system in such a way that the probability of the undesirable event occurring is minimized. Similar to what has been described previously, the evaluation of the extreme value via the extreme value characteristic value α does not yet directly provide the probability as the total variable Q. However, after having previously saved several thousand such evaluations, it is no longer necessary to explicitly evaluate the probability again for this case according to conventional methods.

[0028] Therefore, for these and similar applications, it can be particularly advantageous to provide an effect tree W, the nodes of which correspond to the elementary variables and the connections between which correspond to the logical links of the elementary variables to form the overall variable within the scope of their interaction. This effect tree W can then be used to determine the effect function f(e 1 ,...,e n ). This effect tree W should not be confused with the Lambert-Omega function, which is usually denoted by the same letter.

[0029] In another particularly advantageous embodiment, the basic variable e 1 ,...,e n The basic probabilities of the following elementary events are included, which can be the cause of an unexpected event in the technical system individually or in combination. The logical interaction of the elementary events until the unexpected event may occur is set by the configuration of the technical system. The configuration can, in particular, set logical dependencies and action mechanisms, i.e., to what extent the interaction of multiple elementary events (possibly also via one or more intermediate stages) ultimately leads to the occurrence of the unexpected event. The effect tree W can then be understood as a fault tree. Therefore, the method can then be used to provide a fast approximation for the known fault tree analysis. Fault tree analysis using traditional methods is becoming more and more accurate, but the calculation speed is significantly slower and slower. The approximate nature lies in the simplified assumptions on the basis of which the effect function is factored: in fact, the fault tree also contains contributions that are not detected by the model of the factor graph. When calculating using the factor graph, these contributions are ignored, for example, perturbation theory or linearization also ignores specific contributions. The speed gain depends on the wise choice of the specific factorization.

[0030] In a particularly advantageous embodiment, at least one basic probability is set as a membership function of at least one state variable of the technical system and / or its environment, and / or as a function of at least one other basic probability. These basic probabilities are also referred to as fuzzy probabilities. This converts the fault tree into a fuzzy fault tree. If a certain proportion of the basic probabilities consists of fuzzy probabilities, this does not necessarily affect the basic execution of the method described here. However, the practicality of the method is still significantly improved, because the method saves the evaluation of basic probabilities, which is particularly "expensive" in terms of computing effort. Therefore, the potential speed gain in the case of approximating fuzzy fault trees by factor graphs is even greater than in the case of approximating conventional fault trees.

[0031] Furthermore, the evaluation of the sought probability of the occurrence of an undesirable event is still significantly faster than the traditional evaluation according to the "minimal cut set" method, even in the case of a fault tree with fuzzy probabilities. Even if the technical system is only of moderate size and has about 100 different components, there may already be thousands of "cut sets" of elementary events that, in combination, could be the cause of the undesirable event. These "cut sets" must all be checked to determine whether they are minimal. This analysis must take into account both the technical propagation of faults from one component to the next, as well as the statistical dependencies of the components and the common causes of multiple seemingly independent failures.

[0032] Basic events can be redundant, for example. On the one hand, redundancy can be achieved, for example, by using multiple devices or components that can complement or replace each other to provide a specific function. This redundancy can be achieved, for example, as "hot redundancy", in which multiple devices or components are always active at the same time. This is the situation that is mainly considered here, especially for the purpose of at least partially automated driving. In addition, there is also "cold redundancy", in which other devices or other components become active only as needed when an active device or active component fails or fails. An example of this is an emergency diesel engine that is only activated when the power is off.

[0033] However, redundancy can also be achieved, for example, in that one and the same device or one and the same component is relevant at multiple stages of the causal chain leading to the undesired event.

[0034] All types of redundancy listed here can also be mixed in the same technical system.

[0035] The evaluation of the operating state based on variables different from the probability of an undesired event can also be understood based on the fuzzy fault tree. Here, each entity corresponding to a node in the fault tree has a state vector. The membership function in the fuzzy fault tree indicates for each entity (e.g., a device or component) how important this entity is for the overall state of the system. In particular, the interaction of multiple influences on the entity can be modeled in the form of a membership function. Therefore, within the scope of optimizing the configuration of the technical system, the membership function can also be optimized, so as to ultimately improve the physical interaction.

[0036] In this way, for example, it can be determined that the lack of heat resistance of a particular electrolytic capacitor is a significant factor in the overall failure probability of a technical system. By replacing this electrolytic capacitor with a higher quality electrolytic capacitor, the reliability of the system can then be significantly improved.

[0037] Thus, for example, a fuzzy fault tree can be created for the reliability of thermal management in a device having a heat source (e.g. a CPU or GPU) and monitored by a large number of temperature sensors. This reliability can then be defined, for example, via the probability of an undesirable event "system failure", which in turn can be the cause of an undesirable event "system failure", such as a basic event such as exceeding the maximum permissible temperature at a specific component. The basic probability of the occurrence of a basic event can then depend, for example, on the measured values ​​of a temperature sensor. The method proposed here can then be used particularly quickly to check which changes in the system configuration reduce the probability of a system failure.

[0038] In a particularly advantageous embodiment, a control device for a vehicle is selected as the technical system. This device is a safety-relevant component of the vehicle and must therefore be certified in a relatively complex process. It is therefore advantageous to optimize the control device in a computer using the method described here before the control device is physically implemented and subsequently certified. If it is later found that the performance with respect to the total variable Q is insufficient and the device has to be reworked, the certification may be invalidated. The human and financial effort required for the certification then occurs again.

[0039] Furthermore, the problem of heat dissipation, which is an important application of the method proposed here, is a difficult problem precisely for control devices, since the installation space and thus the possibilities for heat dissipation are very limited.

[0040] This applies in particular to another particularly advantageous embodiment in which the control device is designed to control at least partially automated driving functions of a vehicle. This application usually requires the use of a hardware accelerator (e.g. a GPU) for evaluating the neural network. Such hardware accelerators generate particularly high amounts of heat.

[0041] The method can in particular be implemented completely or partially by a computer. Therefore, the invention also relates to a computer program with machine-readable instructions, which, when executed on one or more computers and / or computing instances, cause the one or more computers and / or computing instances to carry out the described method. In this sense, control devices for vehicles and embedded systems for technical devices that are likewise capable of executing machine-readable instructions should also be considered computers. Computing instances can be, for example, virtual machines, containers or serverless execution environments, which can in particular be provided in the cloud.

[0042] The invention also relates to a machine-readable data carrier with a computer program and / or a download product. A download product is a digital product transferable via a data network, ie downloadable by a user of a data network, which may be sold for immediate downloading, for example in an online store.

[0043] Furthermore, one or more computers and / or computing instances can be equipped with the computer program, the machine-readable data carrier or the download product.

[0044] Further measures for improving the invention are shown in more detail below together with the description of preferred exemplary embodiments of the invention based on the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 An exemplary embodiment of a method 100 for evaluating 2 an operating state of a technical system 1 is shown;

[0046] Figure 2 An exemplary path starting from a fuzzy fault tree for evaluating an operating state via an extreme characteristic value α is shown;

[0047] Figure 3 A control device for a vehicle is shown as an exemplary technical system 1 in which the method 100 is applied. DETAILED DESCRIPTION

[0048] Figure 1 2 is a schematic flow chart of an exemplary embodiment of a method 100 for evaluating 2 an operating state of a technical system 1. The operating state is characterized by at least one overall variable Q, the value of which is determined by a plurality of basic variables e 1 , ..., e n The interaction corresponding to the configuration (3) of the technical system is obtained.

[0049] According to block 106, the total variable Q can be a measure of at least one heat flow to be transferred into or out of the technical system 1 and / or a measure of at least one temperature at a predetermined location in the technical system 1. Then, for example, according to block 106a, at least one basic variable e 1 , ..., en A measure of the heat conductivity and / or transport capability of heat flow from a heat source, into a heat sink and / or within a technical system.

[0050] According to box 107, the total variable Q can be a measure of the probability of at least one undesirable event occurring in the technical system 1. Then, according to box 107a, the basic variables e 1 , ..., e n The basic probabilities of the following basic events may be included, which alone or in combination may be the cause of an undesirable event in the technical system 1. At least one basic probability may then be set according to block 107b as a membership function of at least one state variable of the technical system 1 and / or its environment and / or as a function of at least one other basic probability.

[0051] According to block 108 , a control device for a vehicle may be selected as technical system 1 . According to block 108 a , the control device may then be configured to control at least partially automated driving functions of the vehicle.

[0052] In step 110, an effect function f(e 1 , ..., e n ), the effect function describes the total variable Q on the basic variable e 1 , ..., e n dependency.

[0053] According to block 111, an effect tree W may be provided, the nodes of which correspond to elementary variables and the connections between which correspond to logical links of elementary variables forming overall variables within the scope of their interaction. In this case, the overall variable Q sought is located at the root of the effect tree W. According to block 112, the effect tree W may then be used to determine the effect function f(e 1 , ..., e n ).

[0054] In step 120, the effect function f(e 1 , ..., e n ) is factorized into multiple contributions f i (E i ) The multiple contributions depend on the basis variable e 1 , ..., e n Different subsets of E i .

[0055] In step 130, a graph G is created whose nodes correspond to contributions f i and its edges correspond to the elementary variables e 1 , ..., e nIn this case, from each corresponding contribution f i The nodes all emit i Each basic variable e depends on 1 , ..., e n In addition, with two or more contributions f i The basic variable e that depends on 1 , ..., e n There is at least one edge connecting two corresponding contributions f i The corresponding node.

[0056] In step 140, the Laplacian matrix L of the graph G is determined. The Laplacian matrix L is uniquely determined.

[0057] In step 150 , the extreme eigenvalue α of the Laplacian matrix L is determined as an estimate 2 of the sought operating state.

[0058] According to block 105, a large number of candidate configurations 3* of the technical system 1 can be created. These candidate configurations 3* differ in that the basic variables e 1 , ..., e n The values ​​of these basic variables e 1 , ..., e n to form the overall variable Q. Then, according to block 160, the operating state resulting therefrom can be evaluated for each of these candidate configurations 3*.

[0059] exist Figure 1 In the example shown, in step 170 the basic variable e is optimized. 1 , ..., e n at least one value and / or at least two elementary variables e 1 , ..., e n The aim of the interaction is to improve the evaluation 2 of the operating state resulting from such a changed configuration 3 of the technical system.

[0060] In this case, according to block 105a or 171, in particular at least one new configuration 3 of the technical system 1 can be generated, for example, by converting at least one component of the technical system 1 into a new configuration 3.

[0061] - Replace with another component, and / or

[0062] - placed in a location, and / or

[0063] -Manipulate in other ways.

[0064] exist Figure 1In the example shown, in step 180 , the candidate configuration 3 * with the best evaluation of the operating state resulting therefrom and / or the optimal configuration 3 found within the scope of the optimization are implemented in the hardware of the technical system 1 .

[0065] Figure 2 The path for evaluating the operating state via the extreme characteristic value α starting from the fuzzy fault tree as effect tree W is shown. The effect tree W contains the basic events e 1 , e 2 , e 3 and e 4 , each event has a basic probability of occurrence. 1 , e 2 , e 3 and e 4 By generating an intermediate result g 1 , g 2 , g 3 and g 4 The logic gates are linked to each other. These intermediate results g 1 , g 2 , g 3 and g 4 can be considered as a combination of events that can have a specific technical significance in the context of a technical system 1. Thus, for example 3 can be the probability of a fan problem in general, and can be determined by the basic probability of failure of a specific fan, e 2 and e 3 At the top of the effects tree W are undesirable events, whose probability should be assessed or reduced.

[0066] When converting the effect tree W into a graph G, on the one hand, the basic event e 1 , e 2 , e 3 and e 4 And link these basic events 1 , e 2 , e 3 and e 4 The logic gates of the graph G exchange roles to some extent. The nodes of the graph G correspond to the gate functions f 1 , f 2 , f 3 and f 4 , these gate functions calculate the intermediate result g 1 , g 2 , g 3 and g 4 For the sake of clarity, Figure 2 It is not shown that the effect function f(e) is first determined from the effect tree W. 1 , ..., e n ), and then factorize the effect function into contribution fi (E i ) is an intermediate step.

[0067] exist Figure 2 In the example shown, the function f 1 Directly dependent on e 1 , and through g 3 Indirectly dependent on e 2 and e 3 The function f 2 By g 3 and g 4 Indirectly dependent on e 2 、e 3 and e 4 . This also covers the following facts: g 3 It depends on e 2 , and g 4 It depends on e 2 Only through f 2 and f 3 Note that the edge between 2 and f 3 All depend on e 3 In the same way, through f 2 and f 4 Note that the edge between 2 and f 4 All depend on e 4 Therefore, f 1 and f 2 Also by corresponding to e 2 The edge connection.

[0068] The graph G uniquely yields a Laplace matrix L, whose extreme eigenvalues ​​α can be used as an estimate of the operating state of the technical system 1 .

[0069] Figure 3 A control unit for an automated driving function of a vehicle is shown by way of example as technical system 1 , the operating state of which can be evaluated in a particularly advantageous manner using method 100 . Figure 3 The side of the circuit board equipped with active devices is shown. The circuit includes, among other things, hardware accelerators A, B and C for specific functions (e.g. inference of a neural network). Numbered measurement points ("MP") are distributed on the circuit board, at which the local temperature is measured. On the back side of the circuit board ( Figure 3 A "heat sink" (not shown) is provided as a passive cooling element.

Claims

1. A method (100) for evaluating (2) an operating state of a technical system (1), wherein the operating state is characterized by at least one overall variable Q, the value of which is determined by a plurality of elementary variables e1, ..., e n The method is derived in accordance with the interaction of the configuration (3) of the technical system (1), comprising the steps of: Provide (110) effect function f(e1,…,e n ), the effect function describes the total variable Q on the basic variables e1,…,e n Dependence; · The effect function f(e1,…,e n ) factorizes (120) into multiple contributions f i (E i ) The multiple contributions f i (E i ) depends on the basic variables e1,…,e n Different subsets of E i ; Create (130) a graph G whose nodes correspond to contributions f i And the edges of the graph G correspond to the elementary variables e1,…,e n ,in οFrom each corresponding contribution f i The nodes all emit i Each basic variable e1,…,e n The edge of ο with two or more contributions f i The basic variables e1,…,e that it depends on n There is at least one edge connecting two corresponding contributions f i The corresponding node; Determine (140) the Laplacian matrix L of the graph G; The extreme eigenvalue α of the Laplacian matrix L is determined ( 150 ) as the evaluation of the operating state sought ( 2 ).

2. The method (100) according to claim 1, wherein · Creating (105) a large number of candidate configurations (3*) of the technical system (1), the candidate configurations differing in that the basic variables e1, ..., e n The value of and / or the elementary variables e1,…,e n The interactions that form the total variable Q; and The operating state resulting from this is evaluated ( 160 ) for each of the candidate configurations ( 3*).

3. The method (100) according to any one of claims 1 to 2, wherein the basic variables e1, ..., e n at least one value and / or at least two elementary variables e1,…,e n The aim of the interaction is to improve the evaluation (2) of the operating state resulting from such a changed configuration (3) of the technical system.

4. The method (100) according to any one of claims 2 to 3, wherein at least one new configuration (3) of the technical system (1) is generated (105a, 171) by replacing at least one component of the technical system (1) with a new configuration (3). Replace with another component, and / or Placed in a location, and / or Manipulate in other ways.

5. A method (100) according to any one of claims 2 to 3 and optionally also according to claim 4, wherein a candidate configuration (3*) having an optimal evaluation of the operating state derived therefrom and / or an optimal configuration (3) found within the scope of the optimization is implemented (180) in the hardware of the technical system (1).

6. A method (100) according to any one of claims 1 to 5, wherein the total variable Q is (106) a measure of at least one heat flow to be transmitted within or from the technical system (1), and / or a measure of at least one temperature at a predetermined position within the technical system (1).

7. The method (100) according to claim 6, wherein at least one elementary variable e1, ..., e n A measure representing (106a) the heat flow from a heat source, the heat flow into a heat sink and / or the heat conductivity and / or transport capacity of the heat flow within the technical system.

8. The method (100) according to any one of claims 1 to 7, wherein the total variable Q is (107) a measure of the probability of at least one undesirable event occurring in the technical system (1).

9. The method (100) according to claim 8, wherein the basic variables e1, ..., e n The basic probability of the following basic events is included (107a), which can be the cause of the occurrence of the undesirable event in the technical system (1) individually or in combination.

10. A method (100) according to claim 9, wherein at least one basic probability is set (107b) as a membership function of at least one state variable of the technical system (1) and / or its environment, and / or is set (107b) as a function of at least one other basic probability.

11. The method (100) according to any one of claims 9 to 10, wherein redundant elementary events are selected.

12. The method (100) according to claim 11, wherein redundancy of basic events (3) in the technical system (1) is achieved by using a plurality of devices or components that can complement or replace each other to provide a specific function.

13. The method (100) according to claim 12, wherein the redundancy in the technical system (1) is implemented as hot redundancy, so that the plurality of devices or components providing the same function are always active at the same time, and / or Cold redundancy, so that when an active device or active component fails or malfunctions, another device or another component becomes active as needed.

14. A method (100) according to any one of claims 11 to 13, wherein redundancy of basic events in the technical system (1) is achieved in that the same device or the same component is relevant at multiple stages of a causal chain leading to an undesirable event.

15. The method (100) according to any one of claims 1 to 14, wherein a control device for a vehicle is selected (108) as the technical system (1).

16. The method (100) of claim 15, wherein the control device is configured (108a) to control at least partially automated driving functions of the vehicle.

17. The method (100) according to any one of claims 1 to 16, wherein providing (111) an effect tree W, the nodes of which correspond to elementary variables and the connections between which correspond to the logical links of the elementary variables forming the overall variable within the scope of their interaction, wherein the overall variable Q sought is located at the root of the effect tree W, and Use the effect tree W to determine (112) the effect function f(e1, ..., e n ).

18. A computer program comprising machine-readable instructions which, when executed by one or more computers and / or computing instances, cause the one or more computers and / or computing instances to perform a method (100) according to any one of claims 1 to 17.

19. A machine-readable data carrier and / or download product having a computer program according to claim 18.

Images