Method for determining an unallowable deviation of the system behavior of a technical device from a standard value range

Abstract

In a method for determining an unallowable deviation of a system behavior of a technical device by means of a monitoring algorithm, input data and output data of the technical device are fed to the monitoring algorithm in a learning phase, in a subsequent prediction phase only the input data are fed to the monitoring algorithm and output comparison data are calculated. In a preprocessing step, the input data fed to the monitoring algorithm are standardized to data of a reference signal.

Description

TECHNICAL FIELD

[0001] The invention relates to a method for determining an impermissible deviation of a system behavior of a technical device from a standard value range by means of a monitoring algorithm. BACKGROUND

[0002] A method for predicting a driving maneuver of an object by means of two machine learning systems is described in DE 10 2018 206 805 B3. A first machine learning system determines an output variable characterizing the object from a first input variable, and a second machine learning system determines a second output variable characterizing a state of the object from a second input variable. The future movement of the object is predicted from the output variable. In this document, the first machine learning system comprises a deep neural network, while the second machine learning system comprises a probabilistic graphical model.

[0003] DE 10 2018 209 916 A1 discloses a method for determining a sequence of output signals by means of a sequence of layers of a neural network on the basis of input signals fed to an input layer of the neural network. At defined points in time, new input signals have been fed to the neural network, while previous input signals are still propagating through the neural network. SUMMARY

[0004] By means of the method according to the invention, an impermissible deviation of a system behavior of a technical device from a standard value range can be determined. In this way, a total or partial failure of the technical device can be predicted before the failure actually occurs, so that appropriate countermeasures can be taken in time. In this way, the status of the technical device can be monitored using easily implemented measures. A worsening of the system behavior and system anomalies can be determined in time. By predefining and comparing with the standard value range, the course of the status change of the technical device can be continuously monitored and the point in time can be determined up to which the proper functioning of the technical device is ensured and from which point in time the proper functioning can no longer or at least not completely be ensured.

[0005] The method for determining an impermissible deviation of a technical device uses a monitoring algorithm, to which input data and output data of the technical device are fed in a learning phase. By comparison with the input data and output data of the technical device, corresponding links are created in the monitoring algorithm and the monitoring algorithm is trained with respect to the system behavior of the technical device.

[0006] In the prediction phase following the learning phase, the system behavior of the device can be reliably predicted in the monitoring algorithm. For this purpose, only the input data of the technical device are fed to the monitoring algorithm in the prediction phase, and output comparison data are calculated in the monitoring algorithm which are compared with the output data of the technical device. If the comparison shows that the output data of the technical device, preferably detected as a measurement value, deviate too much from the output comparison data of the monitoring algorithm and exceed a limit value, the system behavior of the technical device deviates unacceptably from the standard value range. Appropriate measures can then be taken, for example a warning signal can be generated or stored or parts of the functionality of the technical device can be deactivated (degradation of the technical device). If necessary, a switch to an alternative technical device can be made in the event of an unacceptable deviation.

[0007] By means of the above-described method, a real technical device can be continuously monitored. In the learning phase, the monitoring algorithm obtains sufficient information from both its input and from its output, so that the technical device can be mapped and simulated in the monitoring algorithm with sufficient accuracy. This allows the technical device to be monitored and the deterioration of the system behavior to be predicted in the subsequent prediction phase. In this way, in particular, the remaining useful life of the technical device can be predicted.

[0008] A neural network is considered in particular as a monitoring algorithm. In the neural network, links are created from the input data and the output data of the technical device in the learning phase, whereby the neural network maps the system behavior of the technical device with high accuracy. In the prediction phase, the neural network can be used correspondingly to reliably predict the deterioration of the system behavior.

[0009] As an alternative to a neural network, a monitoring algorithm implemented in another way can also be considered for monitoring the system behavior of a technical device.

[0010] In the method according to the application, the input data fed to the monitoring algorithm are standardized to data of a reference signal in a preprocessing step carried out before each learning phase step and before each prediction phase step. The advantage of this procedure is that fluctuations in boundary conditions, for example due to natural scattering, can be compensated for or at least largely compensated for by the standardization, whereby depending on the type of scattering, the processing in the learning phase and in the prediction phase is improved, in particular can be carried out more quickly, or only thereby becomes possible. The learning phase and the prediction phase of the monitoring algorithm itself are not influenced by the preprocessing step, since only the input data are standardized in each phase.

[0011] In an advantageous implementation, the standardization is related to the amount of input data fed to the monitoring algorithm. If this amount deviates from the amount of data in the reference signal, standardization is performed to unify the amount of input data to the amount of data in the reference signal. Therefore, the same amount of input data is always fed to the monitoring algorithm after standardization.

[0012] Another advantageous implementation involves a situation where the amount of input data and the amount of reference signal data are the same, but the input data is distorted relative to the reference signal. In this case, normalization can also be performed, where the distorted input data is mapped to the data of the reference signal. This process, for example, makes it possible to map the maximum or minimum value of the shift in the input data to the data of the reference signal.

[0013] According to another advantageous embodiment, the standardization of the input data fed to the monitoring algorithm is performed in three sub-steps. The input data exists in a time-discrete manner, wherein in the first sub-step, time is standardized to a reference signal within the considered time window. In the subsequent second sub-step, the unstandardized input data for different time periods within the considered time window is transformed into the frequency domain. A third sub-step follows, in which frequency segments assigned to different time periods are combined according to the time standardization of the first sub-step. The result is standardized input data in the frequency domain, which is fed as input to the monitoring algorithm. The output comparison data generated in the monitoring algorithm during the prediction phase is also correspondingly located in the frequency domain.

[0014] The comparison between the output comparison data of the monitoring algorithm and the output data of the technical device can be performed in either the time domain or the frequency domain. In the case of a time-domain comparison, the output comparison data present at the output of the monitoring algorithm is inversely transformed from the frequency domain to the time domain, and then compared with the output data of the technical device in the time domain. In the case of a frequency-domain comparison, the output data of the technical device (which is typically located in the time domain, for example, as a measurement sequence) is transformed to the frequency domain. The output comparison data of the monitoring algorithm and the output data of the technical device can then be compared in the frequency domain.

[0015] According to another advantageous implementation, time normalization of the input data to the reference signal, performed in the first sub-step, is achieved through dynamic time warping. In this case, from an optimization perspective, particularly considering the cost function, an optimal path is laid through a matrix that forms the distance from each point of the reference signal to each point of the input data. From an optimization perspective, the most cost-effective path through said matrix is ​​the path whose connection from the starting point to the ending point forms the minimum sum.

[0016] According to a further advantageous embodiment, the transformation of the input data within the considered time window into the frequency domain, which is carried out in the second sub-step, is carried out by means of a short-time Fourier transform (STFT). In this transformation into the frequency domain, a fast Fourier transform (FFT) is carried out for a plurality of time segments, respectively. The advantage of this procedure is that the time information is retained even after the transformation into the frequency domain. Thus, if necessary, an inverse transformation into the time domain can also be carried out, in particular in order to carry out a comparison with the output data of the technical device in the time domain.

[0017] The reference signal for carrying out the standardization is formed, for example, from a plurality of previous input data, for example by forming an average value from a plurality of input signals.

[0018] Alternatively, the reference signal can also follow a defined maneuver which is coordinated with the technical device concerned and which is typical for said technical device. For example, in the automotive sector it is expedient to predefine a defined driving maneuver of a vehicle on the basis of the technical devices used in the vehicle, from which the reference signal is formed.

[0019] The application also relates to an electronic device, for example a control device in a vehicle, which is equipped with means for carrying out the above-described method. These means are, in particular, at least one computing unit and at least one storage unit, respectively, for carrying out the necessary calculations and for storing the input data and the output data.

[0020] The application also relates to a computer program product having a program code which is designed to carry out the above-described method steps. The computer program product can be stored on a machine-readable storage medium and can be run in the above-described electronic device.

[0021] The method can be applied, for example, for monitoring the state of a technical system in a vehicle, for example a steering system or a brake system. In this case, the electronic device is advantageously a control device by means of which components of the technical device can be actuated. Furthermore, it is also possible to monitor only one subsystem as a technical device within a larger system, for example an ESP module (electronic stability program) in a brake system. BRIEF DESCRIPTION OF DRAWINGS

[0022] Further advantages and expedient embodiments result from the other claims, the description of the figures and the figures.

[0023] Figure 1 A block diagram is shown which has a symbolic representation of an ESP module which is supplied with input data and produces output data and which has a neural network connected in parallel,

[0024] Figure 2 A diagram is shown which shows the time course of an input signal and a reference signal,

[0025] Figure 3 a diagram of the input signal transformed into the frequency domain is shown in matrix form,

[0026] Figure 4 is shown in accordance with Figure 2 the input signal transformed into the frequency domain by time normalization. DETAILED DESCRIPTION

[0027] In a block diagram in accordance with Figure 1 a schematic diagram of a technical device 1 in the form of an ESP module for a brake system in a vehicle is shown, which has input data and output data and a neural network 4 connected in parallel. The ESP module 1, for example, serves as a technical device, comprises an ESP pump for generating a desired, modulated brake pressure in the brake system and a control device for operating the ESP pump. The input data 2, for example, an input current for an electrically operable ESP pump of the ESP module 1, are fed to the ESP module 1, wherein the ESP module 1 generates the output data 3, for example, a hydraulic brake pressure, in response to the input data 2.

[0028] The neural network 4, which forms a monitoring algorithm, is connected in parallel to the technical device 1. The neural network 4 is trained for the system behavior of the technical device 1 in a learning phase, for which both the input data 2 and the output data 3 of the technical device 1 are fed to the neural network 4 in the learning phase. In Figure 1 In the learning phase, the dashed arrow from the output data 3 to the neural network 4 corresponds to the learning phase of the neural network, in which the output data 3 are fed to the neural network in addition to the input data 2.

[0029] After the end of the learning phase, the neural network 4 can be used in a prediction phase to determine an early deterioration of the system behavior of the technical device 1. For this purpose, the input data 2 of the technical device 1 are fed as input to the neural network 4 in the prediction phase, wherein the neural network 4 now generates output comparison data on the basis of its learned behavior (output on the neural network 4 is indicated by a solid line). The output comparison data of the neural network 4 can be compared with the output data 3 of the technical device 1. If the deviation between the output comparison data of the neural network 4 and the output data 3 of the technical device 1 exceeds a given standard value range, an unallowable severe deterioration of the system behavior of the technical device 1 occurs, from which it can be concluded that the service life is shortened or that a partial failure of the technical device 5 occurs. Measures can then be taken, for example, a warning signal is generated or the functional scope of the technical device 5 is reduced.

[0030] The neural network 4 can be implemented in the control device of the technical device 1 and run there. However, the neural network 4 can also run in a further control device which is implemented separately from the control device of the technical device 1.

[0031] Figures 2 to 4 A preprocessing step is shown, which is performed before each learning phase step and before each prediction phase step, in which the input data delivered to the monitoring algorithm is standardized to data of a reference signal.

[0032] Figure 2 Two superimposed diagrams are shown, a time-dependent course of a reference signal R (lower diagram) and a time-dependent course of a signal with measured input data M (upper diagram). The input data M corresponds to the input data 2 in Figure 1 The reference signal R has a series of time points a, b, c, d and e. The signal with input data M has a series of time points 1 to 6 at which the values of the input data are measured. The reference signal R can be obtained, for example, from a large number of previous real input data of the technical device or other technical devices of the same design.

[0033] While the signal courses R and M have in principle the same course, they are not identical. In order to standardize the measured signal of the input data M with a total of six measured time points 1 to 6 to the reference signal R with a total of five time points a to e, a dynamic time standardization (dynamic time warping) is performed in a first substep. In this case, the most cost-effective path from the starting point to the end point of the two signal courses R and M is sought from an optimization point of view. As a result, an assignment shown in dashed lines with an assignment pattern 1a, 2b, 3c, 4c, 5d and 6e between the time points in the signal courses R and M is obtained. The measured values of the signal course M at the time points 3 and 4 are both assigned to the time point c in the reference signal R.

[0034] Figure 3 A schematic diagram of the input data M in the frequency domain is shown. In this case, the input data M is transformed into the frequency domain in a second substep by a short-time Fourier transform STFT in such a way that a fast Fourier transform is performed at each time point t = 1 to t = 6, respectively. The advantage of this procedure is that the time information is preserved even during the transformation into the frequency domain. In the matrix according to Figure 3 Each column represents a vector transformed into the frequency domain, which is assigned to one of the time points t = 1 to 6.

[0035] Figure 4 A third substep and the last substep of the input data preprocessing is shown, in which the matrix of the input data M in Figure 2 is combined according to the time standardization of the first substep according to Figure 3 This results in a combination of the input data M in the frequency domain, as Figure 4As shown, the frequency bands assigned to time points 3 and 4 are combined to a common frequency band. The result is a reduction of the frequency bands from six to five. For example, frequency bands 3 and 4 are combined by averaging the information in the respective vectors assigned to time points 3 and 4.

[0036] After the end of the pre-processing, the standardized input data M in the frequency domain can be delivered to a monitoring algorithm implemented as a neural network in the prediction phase, which then determines output comparison data in the frequency domain, which can be compared with the associated output data of the technical device in the frequency domain. In the case of an impermissible deviation that indicates a deterioration of the system behavior of the technical device, for example, an alarm signal can be generated.

[0037] As an alternative to this procedure, the output comparison data calculated in the neural network can also be transformed from the frequency domain to the time domain and compared with the output data of the technical device in the time domain. In this case, if there is an impermissible high deviation that indicates a deterioration of the system behavior, an alarm signal can be generated or other measures can be taken, for example, a functional degradation of the technical device can be carried out or an alternative technical device can be activated.

Claims

1. A method of determining an impermissible deviation of a system behavior of a technical device (1) from a standard value range by means of a monitoring algorithm (4), input data (2, M) and output data (3) of the technical device (1) being fed to the monitoring algorithm in a learning phase, wherein only input data (2, M) of the technical device (1) are fed to the monitoring algorithm (4) and output comparison data are calculated in the monitoring algorithm (4) in a prediction phase following the learning phase, wherein an impermissible deviation of the technical device (1) is determined if the output data (3) of the technical device (1) lie outside the standard value range as a result of a difference from the output comparison data of the monitoring algorithm (4), wherein the input data (2, M) fed to the monitoring algorithm (4) are standardized in a preprocessing step to data of a reference signal (R), wherein in the preprocessing step the input data (2, M) are mapped to the data of the reference signal (R) if the number of the input data (2, M) is the same as the number of the data of the reference signal (R), but the input data (2, M) are distorted with respect to the data of the reference signal (R).

2. The method of claim 1, wherein, In the preprocessing step the number of the input data (2, M) fed to the monitoring algorithm (4) is unified to the number of the data of the reference signal (R).

3. The method of any one of claims 1-2, wherein, The standardization of the input data (2, M) fed to the monitoring algorithm (4) in the preprocessing step is carried out in three sub-steps, the input data (2, M) being present in a time-discrete manner, wherein in a first sub-step the input data (2, M) are time-standardized to the reference signal (R) in a considered time window, in a second sub-step the input data (2, M) of the time period of the time window are transformed into the frequency domain, and in a third sub-step frequency segments of the input data (2, M) assigned to different time periods are combined in accordance with the time standardization of the first sub-step.

4. The method of claim 3, wherein, The time standardization of the input data (2, M) to the reference signal (R) carried out in the first sub-step is carried out by dynamic time warping.

5. The method of claim 3, wherein, The transformation of the input data (2, M) within the considered time window into the frequency domain carried out in the second sub-step is carried out by a short-time Fourier transform (STFT).

6. The method of claim 3, wherein, The output data (3) of the technical device (1) are transformed into the frequency domain and compared in the frequency domain to the output comparison data calculated in the monitoring algorithm (4).

7. The method of claim 3, wherein, The output comparison data calculated in the monitoring algorithm (4) are transformed into the time domain and compared in the time domain to the output data (3) of the technical device (1).

8. The method of any one of claims 1 to 2, wherein, The reference signal (R) is formed from a plurality of previous input data (2, M).

9. The method of any one of claims 1 to 2, wherein, The reference signal (R) corresponds to a defined maneuver, for example a defined driving maneuver of a vehicle.

10. The method of any one of claims 1 to 2, wherein, The monitoring algorithm (4) is configured as a neural network.

11. An electronic device having means configured to carry out the method according to any one of claims 1 to 10.

12. The electronic device of claim 11, wherein, The electronic device is a control device in a vehicle.

11. An electronic device having means configured to carry out the method according to any one of claims 1 to 10.

13. A computer program product having program code designed to perform the steps of the method according to any one of claims 1 to 10.

14. A machine-readable storage medium on which the computer program product according to claim 13 is stored.

Images