METHOD FOR DETERMINING A STATE OF A TEST STAND, COMPUTER PROGRAM, CONTROL DEVICE AND TEST STAND

Abstract

A computer-implemented method for determining a state of a test bench (1) for testing control units (2) is provided. The method comprises obtaining at least one target measured value for a control unit (2) to be tested, obtaining at least one measured value for the control unit (2) to be tested, obtaining at least one calibration value of a standard control unit, and determining the state of the test bench (1) based on the target measured value, the measured value, and the calibration measured value. Furthermore, a computer program, a control device (3), and a test bench for testing control units (2) are provided.

Description

[0001] The present invention relates to a computer-implemented method for determining a state of a test bench for testing control units, a computer program, a control device, and a test bench for testing control units.

[0002] Today, a multitude of control units are installed in vehicles. To ensure trouble-free operation, the control units are tested at their respective locations before installation (for example, in vehicles as vehicle control units and / or transmission control units). Test benches are used to test the functionality of the control units. This allows faulty control units to be identified early and their installation prevented. This avoids the subsequent, more costly, replacement of faulty control units.

[0003] However, test benches also contain electronic circuits and electronic components that can fail or slowly deteriorate over time. This creates the risk that otherwise perfectly good control units will be identified as defective by a defective or faulty test bench and subsequently rejected. This can lead to otherwise perfectly good control units not being used. This can lead to ineffective quality management.

[0004] DE 10 2016 220 896 A1 discloses a method and device for testing a test bench designed to test high-voltage storage devices. US 2011 / 0 307 217 A1 discloses a method and a device for analyzing and evaluating measurement data from a measuring system.

[0005] Therefore, it is an object of the present invention to provide a method and a device which can monitor a test bench.

[0006] The present invention solves this problem with a computer-implemented method for determining a state of a test bench having the features of claim 1, with a computer program having the features of claim 10, with a control device for determining a state of a test bench having the features of claim 11 and with a test bench for testing control devices having the features of claim 12.

[0007] According to one aspect of the present invention, there is provided a computer-implemented method for determining a state of a test bench for testing control units, the method comprising: - Obtaining at least one target measurement value for a control unit to be tested, - Obtaining at least one measured value for the control unit to be tested, - Obtaining at least one calibration value of a standard control unit, and - Determine the condition of the test bench based on the target value, the measured value and the calibration measured value.

[0008] Compared to the state of the art, the test bench on which the control unit can be tested is also checked. Thus, the above procedure can detect both complete test bench failures (e.g., due to connection errors) and a gradual deterioration of test benches over time. As soon as failures or deterioration are detected, necessary countermeasures, such as recalibration of the respective test bench, can be initiated. This ensures high quality standards for the tested control units and prevents the unjustified rejection of otherwise flawless control units.

[0009] The status of the test bench can be defined by whether or not it recognizes faulty control units as faulty. Likewise, the status of the test bench can be defined by whether or not it recognizes faultless control units as faulty. To do this, a test bench can transmit input signals to a control unit under test and then record and evaluate the output signals issued by the control unit. If the output signals are within a predetermined range, the test bench can determine that the control unit is OK. If, on the other hand, the output signals are outside the predefined range, the test bench can determine that the control unit is defective. The test bench can have electronic circuits such as relays and the like for inputting and evaluating the signals.

[0010] The control units to be tested can, for example, be computer-like units that are designed to receive information or signals, process them, and generate and output commands as output signals based on the received signals or commands. The control units to be tested can, for example, be vehicle control units (ECUs), transmission control units (TCUs), or the like. Furthermore, the control units can be designed to control the operation of a vehicle (in particular an electric vehicle). The control unit can, for example, comprise a processor (e.g., CPU) and other circuits. Depending on the area of ​​application of the control unit, the control units tested on the test bench can differ from one another. This means that the test procedures performed by the test bench can also differ depending on the type of control unit.Accordingly, the results to be evaluated for a test of a specific control unit may differ from the results of other control units. The test bench can be designed to automatically detect the type of control unit being tested. Consequently, a variety of different control units can be tested on the same test bench.

[0011] Obtaining the measured values ​​can involve receiving the measured values. For example, the measured value(s) can be obtained from a database. This also enables subsequent evaluation or determination of the test bench's condition. This provides traceability in cases where the cause of a defect needs to be subsequently clarified. Additionally or alternatively, the measured values ​​can also be obtained directly during a measurement process of the respective control unit under test. The measured values ​​can be received or obtained via a wired or wireless connection.

[0012] The target measured value for a control unit under test can be an optimum value for a test result for that control unit. In other words, the target measured value can be the desired measured value for the control unit. The target measured value can therefore represent a measurement result at which the control unit would pass the test. If the control unit actually achieved the target measured value, the control unit would be released for later use. The target measured value can, for example, be obtained from a database. Thus, the target measured value can, for example, be derived from a table for the respective control unit under test. Furthermore, the target measured value can be determined individually for each control unit under test. For example, individual components that make up the control unit can influence the target measured value. Thus, determining the target measured value can involve obtaining information about the control unit under test.For example, an algorithm can be used to determine the target measurement value individually for each control unit under test. A linear machine learning model can also be used for this purpose (further details follow below). This makes it possible to obtain a multitude of different target measurement values. Consequently, a large number of different control units can be tested without having to manually determine a target measurement value beforehand. This can increase the efficiency of the quality management process. The target measurement value can be within a range at which a control unit passes the test on the test bench. In other words, a control unit under test does not have to reach exactly the target measurement value to pass the test.

[0013] The measured value for the control unit under test can be a current measured value. In other words, the measured value can be the result of the control unit under test, which is currently being tested on the test bench. The measured value can therefore be the test result of the control unit under test. More precisely, the measured value can be the data about the control unit under test obtained from the test bench.

[0014] The calibration value can be a measured value of a standard control unit. The standard control unit can also be called a test control unit or calibration control unit. The standard control unit has no errors. It can therefore be assumed that testing the standard control unit leads to a measurement result (i.e., a calibration value) according to which the standard control unit passes the test. It can be assumed that the standard control units have no errors. Consequently, the calibration value can match the target measured value. During operation of the test bench, a plurality of control units can be tested one after the other. The above procedure can be carried out continuously for each control unit. The standard control units can be fed into the test circuit at regular intervals to test the test bench. Thus, the condition of the test bench can be continuously determined during ongoing operation of the test bench.This enables ongoing quality assurance of both the control units to be tested and the test benches being tested.

[0015] To determine the condition of the test bench, the target measurement value, the measured value, and the calibration measurement value can be considered. In other words, these values ​​can be compared sequentially or simultaneously, and the condition of the test bench can be determined based on them. In other words, the condition of the test bench can be determined based on these values. Consequently, not only the condition of the control units under test can be monitored, but also the condition of the test bench. This ensures a high standard of quality control.

[0016] According to the invention, determining the state of the test bench comprises comparing the target value with the measured value. Thus, after obtaining the measured value (current measured value) of the control unit under test and the target value of the same control unit under test, the target value can be compared with the measured value. The result of such a comparison can indicate that the target value corresponds to the measured value or that the target value does not correspond to the measured value. Thus, for certain control units, discrepancies may occur between the target measured value and the measured value of the control unit under test. If the target value does not correspond to the measured value, an anomaly may exist. If, on the other hand, the target value corresponds to the measured value, an ideal situation may exist, and it can be determined that the state of the test bench is OK. Furthermore, a predetermined deviation between the target value and the measured value can still result in a determination of the ideal state.In other words, a certain deviation between the target value and the measured value can be tolerated. This allows certain manufacturing and / or measurement tolerances to be taken into account.

[0017] Preferably, determining the state of the test bench comprises comparing the target value with the calibration value. Immediately before comparing the target value with the calibration value, the calibration value can be obtained or received. In other words, the measured value and the target value can be obtained first, then the target value can be compared with the measured value. If the anomaly exists, i.e., the target value does not correspond to the measured value, the calibration value can be obtained. After that, the target value can be compared with the calibration value. As already described above, the calibration value can be a measured value of a perfectly functioning control unit (i.e., a standard control unit). Just as with the comparison of the target value with the measured value, the comparison of the target value with the calibration value can also be a check to see whether both values ​​correspond to one another. If the target value is equal to the calibration value, it can be defined that the tested control unit (i.e.,If the target value (i.e., the control unit that provided the current measured value) is defective and the test bench is OK, it can be determined that the test bench is defective. This allows countermeasures to be initiated to recalibrate the test bench or to repair or replace it in some other way.

[0018] The method can be carried out for one control unit at a time. The standard control unit can correspond to the type of control unit to be tested. In other words, the control unit to be tested corresponds to the standard control unit if the control unit to be tested is free of defects. For example, obtaining the calibration value can only be carried out in an ongoing test procedure if the step of comparing the target value with the measured value results in a discrepancy between the two values ​​(i.e., an anomaly exists). Thus, a standard control unit can only be run on the test bench and tested with it if the target value does not correspond to the measured value. This allows a particularly effective test sequence to be implemented.

[0019] The target measurement value is preferably obtained using a linear machine learning model. The linear machine learning model can, for example, be an algorithm that generates an output value based on input values. In the present case, such an algorithm can include, for example, information about the control unit to be tested as input values. This information can, for example, represent which components the control unit comprises. Furthermore, a configuration of the control unit can be input into the algorithm. The algorithm can then output the target measurement value. In other words, the algorithm can determine which target measurement value the control unit outputs when it is functioning properly. The linear machine learning model can therefore function flawlessly and always consistently output the same output value when it receives the same input values. The linear machine learning model can, for example, be an artificial neural network.Thus, an instance can be provided that consistently delivers an error-free target value. However, it should be noted that the target value only concerns the control unit under test (i.e., it defines the optimal result of a control unit under test). In contrast, the target value cannot include any properties of the test bench.

[0020] According to a further aspect of the present invention, a training method for training a linear machine learning model is provided. The training method comprises providing input training data to be input into the linear machine learning model and providing output training data to be output by the machine learning model. The input training data may be information about components and / or structure of a control unit. The output training data may be target values ​​(i.e., target measured values) that the control unit achieves when its state is normal. By training the linear machine learning model, the model can adapt internal parameters so that, upon receiving unknown information about components and / or structure of a control unit, it outputs a target value for this control unit.

[0021] Preferably, determining the state of the test bench comprises determining at least a first measurement difference between the at least one target measurement value and the at least one measured value, and comparing the at least one first measurement difference with a threshold value range. According to the present embodiment, the method can be used not only to determine obvious faults such as the anomaly described above, but can also be used to detect a slow deterioration of the test bench over time. For this purpose, a measured value of the control unit to be tested can be obtained or received. Furthermore, the target value associated with the respective control unit under test can be obtained or received. Subsequently, any deviation between the target value and the measured value can be determined. The deviation can be the first measurement difference. For this purpose, the target value can be subtracted from the measured value.In other words, when testing a control unit under test, the control unit can be deemed OK if the deviation between the measured values ​​and the target value lies within a predetermined range. For example, due to manufacturing tolerances and measurement inaccuracies, the target value may not always be precisely achieved. Nevertheless, the control unit may be suitable for its intended purpose and thus OK. Therefore, it may be useful to define a defined range with an upper and lower limit, within which there are a number of measurement differences that, if reached, can be deemed OK for the control unit under test. This range can be referred to as the threshold range. If the first measurement difference is within the threshold range, it can be determined that no anomaly exists.If, however, the first measurement difference determined in this way lies outside the threshold range, it can be determined that an anomaly exists. In the latter case, the procedure described above can then be followed. By considering a threshold range, manufacturing tolerances and / or measurement tolerances can also be taken into account. Furthermore, a requirement for the control unit in an intended area of ​​application can also be taken into account. This enables efficient and customized quality assurance of the control units.

[0022] Preferably, determining the state of the test bench comprises determining at least one second measurement difference between the at least one target measurement value and the at least one calibration value, and comparing the at least one first measurement difference with the at least one second measurement difference. For example, after comparing the at least one first measurement difference with the threshold range and determining that the at least one first measurement difference lies within the threshold range, obtaining or receiving the calibration value can be carried out. In other words, in a case where the control unit to be tested successfully completes the test, not simply the next control unit can be tested, but the specific result of the control unit to be tested can be further investigated. For this purpose, a deviation between the target measurement value and the calibration value can be determined. This deviation can be the second measurement difference.In other words, a measured value of the standard control unit can be determined using the test bench in order to establish whether the standard control unit (i.e. the control unit which reproduces the target measured value in the normal case of the test bench) also delivers the target measured value in this specific case. Comparing the at least one first measured difference with the at least one second measured difference can be a check to see whether the first measured difference is equal to the second measured difference. If the first measured difference corresponds to the second measured difference, it can be concluded that the condition of the test bench is OK and the reason for the deviation or drift between the current measured value of the control unit under test and the target value lies in the control units under test.

[0023] Preferably, a plurality of first measurement differences are determined, and wherein determining the state of the test bench preferably comprises creating a density curve of the first measurement difference. In other words, a plurality of control units can be tested. Accordingly, a plurality of first measurement differences can be determined. The plurality of control units can, for example, be a production batch of control units. The batches can be selected randomly or in a sliding window. According to the above, it can happen that the measurement differences between the individual measured values ​​of the control units and the target values ​​of the control units lie within the threshold range, which means that the respective control unit has passed the test, but that the measured values ​​deviate from the predicted target values.If the measured values ​​deviate significantly from the target values, this may indicate a slow deterioration of the test bench and, consequently, a need for recalibration. This deviation can also be referred to as drift.

[0024] The density curve can represent a distribution of the values ​​of the first measurement difference. In other words, the density curve can represent the frequency with which a certain measurement difference occurs. The area under the density curve can be one. Consequently, all events that can occur (i.e. all measurement differences) are described by the density curve. In this case, the density curve can be symmetrical. Preferably, the mode (i.e. the highest point of the density curve) can be exactly zero. In other words, the probability that the first measurement difference is greater than zero can be just as large as the probability that the first measurement difference is less than zero. The density curve can be used to indicate which events occur how frequently. In other words, the density curve can be used to describe how frequently which first measurement differences occur.Consequently, the density curve of the first measurement differences can be used to determine if the frequency of occurring first measurement differences is shifting. Based on this, it can be determined whether the test bench is deteriorating over time. If the above procedure, in which the first measurement difference is compared with the second measurement difference, is carried out before this procedure, the influence of the control unit can be excluded and only the condition of the test bench can be determined. In other words, it can be confirmed that the cause of a shift in probabilities lies in the test bench itself and not in the tested control units. For this purpose, the standard control units can be used as described above. These can be tested on the same test bench to determine whether the target values ​​are reproduced by the standard control units (to test whether the calibration value corresponds to the target measurement value).If the calibration values ​​can be reproduced without any shift, then it can be determined that the problem lies with the tested batch of control units and not with the test bench. If the target measurement values ​​cannot be reproduced, it can be concluded that the problem lies with the test bench itself. It can then be determined that the test bench needs to be inspected and recalibrated, or other appropriate countermeasures need to be taken.

[0025] Preferably, the condition of the test bench is determined based on a shift in the median of the density curve. The median or mode can be the highest point of the density curve. In other words, the first measurement difference that occurs most frequently can be located at this point. If this point or location shifts, it can be concluded that the condition of the test bench is deteriorating. A shift can be viewed with reference to the x-axis, which indicates the measurement differences that occur. In contrast, the y-axis can indicate the relative frequency of the respective measurement differences. Consequently, only one point needs to be monitored, which simplifies the determination of the condition of the test bench. Therefore, a large number of conditions can be monitored in parallel without requiring excessive computing power.

[0026] The method preferably comprises determining a variance of the density curve. This makes it possible to determine whether a certain first measurement difference is statistically significant or not. Consequently, a high quality of the density curve can be ensured. The variance in the density curve can be obtained via an integral over the product of the squared deviation and the density function of the distribution. It can therefore be integrated over the space of all possible characteristics (possible value of a static characteristic). Furthermore, the variance can indicate how the first measurement differences are distributed around the mean of all measurement differences. Since the variance describes the dispersion of the first measurement differences around the mean, the variance is one of the measures of dispersion. The variance can therefore indicate how far the first measurement differences differ from the mean. The larger the variance, the further the first measurement differences are from the mean.Therefore, using the variance, a probability can be determined at which no shift (for example, of the median) is present. For example, it may be useful to specify that the probability that there is no shift should be 95%. This allows individual outliers in the measured values ​​to be ignored, which can further increase the efficiency of the test procedure. By using a variance when determining the condition of the test bench, a certain quantile of the density curve can be considered as a limit. For example, 2.5% of the first measurement differences furthest from the mean can be ignored. The remaining measured values ​​that are still below the density curve can then correspond to 95% of all measured values. With a normal distribution around the mean, this position (or quantile) corresponds to a value between +1.96 and -1.96. 1.96 can be the 0.975 - quantile of the normal distribution.

[0027] Preferably, the method comprises testing at least one control unit to obtain the measured value. As already described above, at least one control unit can be tested in parallel with the evaluation of the measurement results. The measured value can correspond to the current measured value of the control unit under test. Thus, the method can be executed in real time in parallel with testing of a plurality of control units.

[0028] Preferably, testing involves applying a voltage to the control unit under test. The test bench can thus input signals and / or information into the control unit and record and test the control unit's outputs. For example, individual circuits of the control unit can be tested by applying a voltage to a first location and measuring the resulting voltage at a second location. This can be used to detect, for example, faulty contacts and / or short circuits in the control units under test.

[0029] Preferably, the measured value, the target measured value, and the calibration value are a numerical value and / or a binary value. The use of a binary value may also be referred to, for example, as the use of a binary system or a binary system.

[0030] Therefore, the binary value can only have two states. For example, the binary value can represent a "pass" or a "fail." Furthermore, the binary value can also take the numbers one or zero. In contrast, the numerical value can represent the actual measurement result.

[0031] According to a further aspect of the present invention, a computer program with program code is provided which, when executed by a computing unit, carries out the method according to one of the above embodiments. The computer program can be in any code, in particular in a code suitable for test bench controls. According to a further aspect, the invention is directed to a computer-readable medium comprising a computer program as defined above. The computer-readable medium can be any digital data storage device, such as a USB stick, a hard disk, a CD-ROM, an SD card, or an SSD card. Of course, the computer program does not have to be stored on such a computer-readable medium in order to be made available to a customer, but can also be obtained via the Internet.

[0032] According to a further aspect of the present invention, a control device for determining a state of a test bench is provided, wherein the control device is configured to carry out the method according to one of the above embodiments. The control device can be a computer-like element that can receive data, process data, and output data. The control device can be, for example, a computer.

[0033] According to a further aspect of the present invention, a test bench for testing control units with a control device according to the above embodiment is provided. The test bench can be a device or apparatus with which a technical object can be reproducibly tested for its properties. In addition to mechanical equipment for holding the test object (for example, the control unit), the test bench can also comprise corresponding sensors and controls to test the properties of the test object and to record the measured values ​​obtained. The test bench is preferably an "end-of-line test bench." Such a test bench can be provided at the end of a manufacturing process for the objects to be tested (in this case, for example, the control unit) in order to test the functionality of the manufactured test objects. The test bench can thus serve for quality assurance.

[0034] According to one embodiment of the present invention, a specific set of control units known to be fault-free is used. These control units are referred to as standard control units. The measured values ​​provided by these standard control units can be referred to as calibration values. These calibration values ​​should match the target values ​​or target measured values ​​of the control units under test. The standard control units can be sent over the test bench for testing at regular intervals to ensure that the standard control units consistently reproduce the calibration values ​​and match the target measured values. Detecting serious errors

[0035] It may happen that discrepancies occur between the measured values ​​and the target values ​​or target measured values ​​for control units under test. If the test bench consistently returns a "failure" result (i.e., indicates that the control unit is defective), which contradicts the target measured value, an "anomaly" can be assumed. In the event of such an anomaly, the standard control units can be tested on the test bench. This allows the calibration values ​​to be determined. If the standard control units also fail the test, it can be assumed that there is a problem with the test bench, and the test bench must be further investigated. Detecting slow deterioration

[0036] Target readings can be used not only for obvious faults such as the anomaly described above, but also to detect a slow deterioration in the test bench's condition over time. When control units are continuously tested using the test bench, the measured readings from a batch of control units are compared with the target readings. The batches can be selected randomly or within a sliding window. It may happen that the readings from the tested control units are within an acceptable threshold range, meaning the control units have passed the test, but that the readings may deviate from the target readings. If these readings deviate significantly from the target readings, this may indicate a slow deterioration of the test bench and, consequently, the need for recalibration.This drift can be detected using the procedure described below. However, this is only an example procedure, and other procedures are also suitable. To confirm that the cause of the drift lies in the test bench itself and not in the control unit under test, the standard control units are used. They are tested with the same test bench to obtain the calibration values ​​and to see if the calibration values ​​are reproduced and whether they correspond to the target measurements. If the calibration measurements can be reproduced without drift, then the problem lies with the tested batch of control units and not with the test bench. If the calibration measurements cannot be reproduced, it can be concluded that the problem lies with the test bench itself. In this case, the test bench must be examined and recalibrated, or appropriate countermeasures must be taken. Statistical test for drift in the test bench

[0037] The measured values ​​of a batch of control units can be compared and plotted together with the target measured values. Then, an initial measurement difference between the measured values ​​and the target measured values ​​of the control units can be determined.

[0038] The first measurement differences can be plotted as a probability density function of the residuals. X̃ represents the median of the residuals. If the measured values ​​do not deviate from the target values, the residuals are symmetrically distributed around the mean zero. For symmetric distributions: X˜=0 The error probability>X˜ is 50% The probability of error <X˜ ist 50%

[0039] Thus, the density function can be considered as a binomial distribution with p = 0.5. If the number of measured values ​​considered in the batch is sufficiently large, the binomial distribution can be replaced by a normal distribution with mean µ = np and a variance σ 2 = np(1 - p) can be approximated.

[0040] A 95% probability that no drift occurs is equal to µ ± 1.96 σ. =N2±1.96(N12(1−12)) =N2±1.96(N4) =N2(1±1.96N)

[0041] 1.96 is the 0.975 quantile of the normal distribution.

[0042] The above equation gives the control limits that can be used to determine with 95% confidence that drift is absent. For example, if the sample size of a batch is N = 100, the number of samples for which the prediction error is positive should be between 40 and 60. This means there is a 95% confidence that drift is absent. If this limit is exceeded, it can be concluded that drift is likely present and further analysis is required. In general, it is good to have a large sample size of a batch.

[0043] Individual features of embodiments can be combined with other features or other embodiments to form new embodiments. The advantages and configurations mentioned in connection with the individual features also apply analogously to the new embodiments. Advantages and configurations described in connection with the device also apply analogously to the method, and vice versa.

[0044] In the following, embodiments of the present invention are described by way of example with reference to the accompanying figures. Fig. 1 is a schematic diagram of a test bench according to an embodiment of the present invention. Fig. 2 is a schematic flow diagram showing a method according to an embodiment of the present invention. Fig. 3 is a schematic flow diagram showing a method according to an embodiment of the present invention. Fig. 4 is a schematic flow diagram schematically showing a flow of a method according to another embodiment of the present invention.

[0045] Fig. 1 is a schematic representation of a test bench 1 according to an embodiment of the present invention. The test bench 1 is designed to test a control unit 2. For this purpose, the control unit 2 can be inserted into a test fixture 4. The test fixture 4 has an interface that establishes a connection between the test bench 1 and the control unit 2. The test fixture 1 can communicate bidirectionally with the control unit 2 via the test fixture. The test bench 1 also has a control device 3 that is designed to record data, process data, and output data. The control device 3 communicates with an interface 5, which is also provided in the test bench 1. The interface 5 is further designed to obtain data and information from outside the test bench 1 and to forward it to the control device 3. In addition, the interface 5 can output data.The test bench 1 of the present embodiment is arranged at the end of a production line for control units 2. The control units 2 are fed into and removed from the test fixture 4 of the test bench 1 in a fully automatic manner.

[0046] Fig. 2 is a schematic flowchart showing a flow of a method according to an embodiment of the present invention. In step S1, a current measured value of a control unit 2 is obtained or received. In other words, the measured value is a current measured value of the control unit 2 under test. In step S2, the target value is obtained or received. The target value or target measured value corresponds to a measured value at which the control unit 2 under test functions optimally. The target value is generated by a linear machine learning model. The linear machine learning model outputs the target value based on input information about the control unit 2 under test (such as structure and components). The target value is fed to the control device 3 via the interface 5 of the test bench 1. In step S3, the target value is compared with the measured value. In other words, in step S3, it is determined whether the target value is equal to the measured value.According to the result, it is either determined in step S7 that the condition of the test bench 1 is good or it is determined in step S4 that an abnormality exists.

[0047] In the event of an abnormality, the calibration value is obtained or received in step S5. In step S6, it is checked whether the calibration value corresponds to the target value. In other words, in step S6, the target value is compared with the calibration value and it is determined whether the target value corresponds to the calibration value. If the answer to this question is "yes", the system proceeds to step S7 and determines that the condition of the test bench is good. If the answer to the above question from step S6 is "no", the system proceeds to step S8 and determines that the condition of test bench 1 is not good and further steps are necessary. The further steps can, for example, be a recalibration of test bench 1.

[0048] Fig. 3 is a flowchart of a method according to a further embodiment of the present invention. In step S1, a current measured value is again obtained for a control unit 2 under test. In step S2, the corresponding target value of the tested control unit 2 is obtained. Here, the same steps S1 and S2 can be carried out as in the previous embodiment. In step S9, the measured value is subtracted from the target value to obtain a first measured difference C1. In other words, the measured difference C1 is the result of a substruction between the target value and the measured value, or vice versa. In step S10, the first measured difference C1 is compared with a threshold value range. In other words, in step S10, it is determined whether the first measured difference C1 lies within the threshold value range or not.The threshold range is a predefined range that is provided, for example, via the interface 5 of the control device 3 of the test bench 1. Thus, in step S10, the question is answered as to whether the first measurement difference C1 lies within the threshold range or not. If the question is answered with "no," the method proceeds to step S4, in which it is determined that an anomaly exists. Thus, the method of the present embodiment can also be incorporated into the method illustrated in . Fig. 2. More specifically, steps S1, S2 and S4 can be mapped to the corresponding steps from the Fig. 2 illustrated embodiment.

[0049] If, in the present embodiment, the question in step S10 is answered with "yes," the process proceeds to step S5. A calibration value is obtained here. Step S5 can also be assigned to step S5 from the Fig. 2. Subsequently, in step S12, a second measurement difference C2 between the target value and the calibration value is determined. The second measurement difference C2 is determined analogously to the first measurement difference C1. In other words, the target value is subtracted from the calibration value, or vice versa. In a next step S13, the first measurement difference C1 is compared with the second measurement difference C2. In other words, in step S13 it is determined whether the first measurement difference C1 corresponds to the second measurement difference C2. If the answer to the question in step S13 is “yes,” it can be concluded that it is the control unit that is responsible for the first measurement difference not matching the second measurement difference. If a large number of control units 2 are tested, it can then be determined that there is a shift in the measurement values ​​in the batch of control units 2 tested.If, however, the answer to the question from step S13 is "no," it can be concluded that the cause of the inequality between the first measurement difference C1 and the second measurement difference C2 lies in test bench 1. Thus, it is possible to proceed to step S15, where it is determined that there is a gradual deterioration in the condition of test bench 1. Appropriate countermeasures, such as recalibration of test bench 1, can then be initiated.

[0050] Fig.4 is a schematic flow diagram schematically showing a sequence of a method according to a further embodiment of the present invention. Step S15, which defines that a gradual deterioration of the test bench 1 is present, can serve as a starting point. In a subsequent step S12, a density curve can be created from the first measurement differences C1. In one embodiment of the present invention, the density curve is realized as a normal distribution. In step S14, it is then monitored whether the median of the density curve is shifting. In other words, in step S14 it is observed whether a value (i.e. a measured value) of the most frequently occurring measurement results is shifting. If this question is answered with "yes", the process can continue to step S15, in which it is defined that the condition of the test bench 1 is deteriorating gradually.Accordingly, an immediate countermeasure is necessary to ensure the smooth running of the test procedure for control units 2 with test bench 1. If, on the other hand, the answer to the question from step S14 is "no," the process can proceed to step S16, which defines that the current state of test bench 1 is maintained constant. Thus, a deterioration in the condition of test bench 1 is not expected in the foreseeable future. List of reference symbols 1 test bench 2 control unit 3 Control device 4 Test recording 5 Interface

Claims

[1] Computer-implemented method for determining a state of a test bench (1) for testing control units (2), the method comprising: Obtaining at least one target measured value for a control unit to be tested (2), Obtaining at least one measured value for the control unit to be tested (2), Obtaining at least one calibration value of a standard control unit, and Determining the condition of the test bench (1) based on the target measurement value, the measured value and the calibration measurement value, wherein determining the state of the test bench (1) comprises comparing the target value with the measured value. [2] Method according to claim 1, wherein determining the state of the test bench (1) comprises comparing the target value with the calibration value. [3] A method according to any one of the preceding claims, wherein the target measurement value is obtained by a linear machine learning model. [4] Method according to one of the preceding claims, wherein determining the state of the test bench (1) comprises: Determining at least a first measurement difference (C1) between the at least one target measurement value and the at least one measurement value, and Comparing the at least one first measurement difference (C1) with a threshold value range. [5] Method according to claim 4, wherein determining the state of the test bench (1) comprises: Determining at least one second measurement difference (C2) between the at least one target measurement value and the at least one calibration value, and Comparing the at least one first measurement difference (C1) with the at least one second measurement difference (C2). [6] Method according to claim 4 or 5, wherein a plurality of first measurement differences (C1) are determined, and wherein determining the condition of the test bench comprises: Create a density curve of the first measurement differences (C1). [7] Method according to claim 6, wherein the state of the test bench (1) is determined based on a shift of the median of the density curve. [8] A method according to any one of the preceding claims, the method further comprising: Testing at least one control unit to obtain the measured value. [9] Computer program with program code which, when executed on a computing unit, carries out the method according to one of the preceding claims. [10] Control device (3) for determining a state of a test bench, wherein the control device (3) is designed to carry out the method according to one of claims 1 to 8. [11] Test bench (1) for testing control units (2) with a control device (3) according to claim 10.

Images