Method and device for determining thermal load of component in technical system and for predictive diagnosis of thermal fatigue of component

Abstract

The invention relates to a method for determining the thermal load of a passive component in a technical system having at least one active component which generates waste heat during operation thereof and at least one passive component, the method comprises the following steps: continuously detecting an operating variable and an environmental variable as an operating variable course and an environmental variable course in successive evaluation time windows; a temperature specification is determined in each successive evaluation time window by means of a data-based temperature model, which is trained, as a function of a corresponding operating variable course and environment variable course, in order to associate input data corresponding to or dependent on the course of the operating variable and the course of the environmental variable with a temperature specification of the temperature of the at least one passive component; a warning about the thermal load of the at least one passive component is signaled as a function of the temperature specification.

Description

Technical Field

[0001] The invention relates to a technical system having a plurality of components and a method for monitoring the thermal load of the plurality of components. Background Art

[0002] Technical systems, such as vehicles, etc., can operate with different loads in various ways. Here, active components, such as motors, heating devices, etc., generate waste heat, which also heats passive components according to the thermal connection.

[0003] However, higher temperatures are generally a thermal load for all components of the technical system and can lead to increased wear, increased aging, or other degradation. Here, the degree of aging significantly depends on the level of the thermal load and the duration for which the components involved are subjected to the thermal load. In addition, damage to the components involved occurs when the critical maximum temperature is exceeded.

[0004] Generally, the temperature of the components of a technical system can be determined and monitored by means of temperature sensors. However, temperature sensors are costly and, for cost reasons or due to the assembly effort, cannot be installed for every component in the technical system. In addition, the temperature sensors themselves may be error-prone and may fail or provide incorrect measured values, especially under the influence of environmental factors such as humidity or dirt. Therefore, in particular, the thermal load cannot be determined for passive components for which no dedicated temperature monitoring by means of temperature sensors is provided, such that the effect of the thermal load over a long duration can only be noticed after the function of the components involved is impaired or fails. Summary of the Invention

[0005] The object of the invention is to determine the load state of a component based on the thermal action without having to provide a temperature sensor on the component involved.

[0006] This object is solved by a method for determining the thermal load of a component in a technical system according to claim 1 and a method for training a data-based temperature model according to the dependent claims and the corresponding device.

[0007] Further design options are given in the dependent claims.

[0008] According to a first aspect, there is provided a method for determining the thermal load of a component in a technical system, the technical system having at least one active component and at least one passive component, the active component generating waste heat during operation, the method having the following steps:

[0009] Continuously detect operating parameters and environmental parameters as operating parameter trends and environmental parameter trends in successively consecutive evaluation time windows;

[0010] Evaluate the trends of operating parameters and environmental parameters in an evaluation time window in order to obtain input data;

[0011] Determine a temperature specification in each of a succession of evaluation time windows by means of a data-based temperature model, wherein the data-based temperature model is trained to correlate the input data with a temperature specification for the temperature of the at least one passive component;

[0012] Signal a warning about the thermal load of the at least one passive component based on the temperature specification.

[0013] Technical systems, such as vehicles, usually have active components which generate waste heat during operation depending on the load, and this waste heat is discharged into the surrounding environment and to adjacent other active and passive components of the technical system. The heating of the other components causes a thermal load there, which accelerates their degradation or thermal aging in the sense of thermal fatigue.

[0014] As a rule, the load state of a component cannot be measured due to the thermal effect, but only becomes apparent through impaired function or complete failure of the component concerned after a longer period of time.

[0015] The above method proposes to determine the temperature of a specific component in a technical system by means of a data-based temperature model based on the environmental parameters of the technical system and the operating parameters influencing its load state. In particular, the operating parameters can be the load state of the active component, such as the rotational speed, torque, motor power, etc. in an internal combustion engine, or the current and / or power in an electronic component; thus, components conducting current which are heat sources (e.g. the traction line, power electronics or battery in an electric vehicle) can also be detected. The environmental parameters can, for example, describe the ambient temperature, the degree of sunlight, the humidity level, the weather conditions, such as rain, etc.

[0016] It can be provided that the evaluation of the trends of the operating parameters and the environmental parameters includes selecting the values of the respective operating parameters and environmental parameters at moments within the trends of the operating parameters and the environmental parameters in the evaluation time window and / or determining at least one characteristic parameter which is derived from the trends of the operating parameters and the environmental parameters within the evaluation time window as a collective and / or statistical characteristic, in particular an average value or a median value.

[0017] The data-based temperature model correlates the time series of these input parameters with the corresponding current temperature of the component concerned. For example, the data-based temperature model can be configured as a neural network, LSTM (Long short-term memory), GRU (Gated Recurrent Unit), etc.

[0018] In addition, the data-based temperature model can correspond to a CNN-LSTM model or a CNN-GRU model.

[0019] To map the time series of input parameters to a temperature description, the data-based temperature model can be based on a convolutional neural network (CNN) combined with an LSTM architecture. Here, the operating parameters and environmental parameters in each time segment are provided in the form of a time series and combined in a matrix. This matrix is evaluated by the CNN for feature extraction in successive time segments, and the extracted features are continuously evaluated by the LSTM architecture in each time segment in order to determine the temperature description at each evaluation moment.

[0020] In addition, the temperature description can correspond to a temperature or a description of a temperature range, where, in the case of the temperature as the temperature description, the temperature is associated within the temperature range, and for each temperature range, a frequency is given, since the start of operation of the at least one passive component, for the evaluation time window, the temperature is within the temperature range at the frequency, and a warning about the thermal load of the at least one passive component is signaled according to the frequency in each temperature range within the temperature range.

[0021] The thus determined time series of the temperature description constitutes a temperature curve, and the component involved is subjected to this temperature curve. A temperature load spectrum is derived from this temperature curve, that is, for example, a frequency distribution is determined, which cumulatively determines the frequency at which the temperature is above a specific temperature limit. The temperature ranges defined by the temperature limits overlap correspondingly. The temperature load spectrum is accumulated and evaluated accordingly over the entire operating duration of the component involved.

[0022] The evaluation of the temperature load spectrum can be carried out according to the frequency of the association of the correspondingly generated temperature description with a specific temperature range, which is defined by a corresponding temperature limit. If the corresponding pre-given frequency threshold is exceeded in a specific temperature range, this can be noted in the fault memory of the vehicle for the component involved. Depending on the safety relevance of the component, the corresponding warning can be signaled to the user of the technical system or the component can be continued to operate in an emergency operation mode or the continued operation of the technical system can be blocked.

[0023] To train the data-based temperature model, the time series of operating parameters can be recorded by means of the temperature generated by the component during the actual operation of the technical system. The temperature of the component can be determined, for example, by a temperature sensor installed on the component involved in a test manner. This training is carried out by means of methods known for machine learning models, such as backpropagation, etc., with the provided training data set.

[0024] The input data for a data-based temperature model can correspond to a time series of operating parameters and environmental parameters for successively consecutive evaluation times. The operating parameters and environmental parameters at the evaluation time can correspond to their values at the start or end of an evaluation time window associated with the evaluation time (e.g., a duration of 1 to 5 seconds), or to derived parameters that are derived from the changes in the operating parameters and environmental parameters during the evaluation time window as statistical or aggregated parameters (features), such as the average of the operating parameter change trend and / or the environmental parameter change trend. The training data set then corresponds to a vector of the operating parameters and environmental parameters and / or parameters derived therefrom (within the evaluation time window) and a vector of associated temperature specifications as labels. Such training can be implemented using the training data set and can be monitored and verified using a validation data set (selected as part of the training data set).

[0025] A data-based temperature model can be trained externally to the technical system. After training, the data-based temperature model can be transferred into the technical system for monitoring the thermal load of components.

[0026] According to another aspect, there is a method for training a data-based temperature model for use with the method described above, the method having the following steps:

[0027] Continuously detect operating parameters and environmental parameters as operating parameter change trends and environmental parameter change trends in successively consecutive evaluation time windows, and determine the associated temperature specifications of the passive components;

[0028] Generate a training data set based on the operating parameter change trends, environmental parameter change trends, and the associated temperature specifications;

[0029] Train a data-based temperature model to associate input data corresponding to or depending on the operating parameter change trends and environmental parameter change trends with temperature specifications for the temperature of the at least one passive component.

[0030] In addition, the data-based temperature model can be transferred to the controller of the system. Description of the Drawings

[0031] Embodiments will be described in detail below with reference to the drawings. In the drawings:

[0032] Figure 1 A technical system having a plurality of components whose thermal loads are to be monitored is shown;

[0033] Figure 2 A flowchart for illustrating a method for monitoring components in a technical system is shown;

[0034] Figure 3 Schematic diagram showing a data-based temperature model;

[0035] Figure 4 Chart showing the temperature load spectrum over the entire operating duration of a component; and

[0036] Figure 5 Flowchart showing a method for training a data-based temperature model. Detailed Description

[0037] Figure 1 Schematic diagram of a vehicle as an example of a technical system. The vehicle 1 includes a plurality of components, among which a motor is arranged as an active component 2 and a plurality of passive components 3 are arranged near the motor. The passive components may include, for example, a transmission, a generator, a controller module, a windshield wiper system, a heating system, a steering component, etc. For example, one of the components may contain plastic, the degradation of which depends on the duration and level of the thermal load.

[0038] Thermal energy is introduced into the system via the motor 2. The amount of thermal energy is determined from the efficiency of the motor, which significantly depends on the operating mode and type of the motor. The operating mode of the motor can be determined, for example, by the rotational speed, torque, motor power, vehicle speed, refrigeration power of the motor cooler, motor temperature, etc.

[0039] In addition, the temperature distribution in the vehicle significantly depends on the environmental conditions in terms of the temperature effect on the passive components 3. The environmental conditions can be affected, for example, by the external temperature, sunlight irradiation, wind speed, driving speed, etc.

[0040] A data-based temperature model is implemented in a controller 4 that can be designed to control multiple vehicle functions. The temperature model monitors the temperature of one or more passive components 3 as a virtual sensor to estimate their thermal load. For this purpose, a method in the form of an algorithm is implemented in the controller 4, as further elaborated by the Figure 2 flowchart.

[0041] For this purpose, in step S1, operating parameters and environmental parameters are detected as operating parameter trends and environmental parameter trends at preferably regular time intervals, for example, between 0.1 s and 1 s. The operating parameters of the operating parameter trends may include, for example, the rotational speed, load, motor power, vehicle speed, etc. of the motor. The environmental parameters may include the external temperature, sunlight irradiation, wind speed, and air humidity.

[0042] In step S2, the trends of changes in operating parameters and environmental parameters within an evaluation time window, for example, from 1 s to 10 s, are evaluated as input data for one time step. The duration of the evaluation time window can also be set as a parameterizable hyperparameter, which is optimized during training. For this purpose, the input data can respectively include the values of operating parameters and environmental parameters, such as the corresponding values at the start or end of the corresponding evaluation time window, and / or the values of aggregated characteristic parameters derived therefrom. The characteristic parameters can be derived from the trends of changes in operating parameters and environmental parameters within the evaluation time window as aggregated and / or statistical features (such as mean value, median, etc.).

[0043] The input data can be set in the form of a vector or a matrix and used as the input data for a data-based temperature model. The input data characterizes the operation of the technical system at a specific evaluation moment during the evaluation time window.

[0044] Now, in step S3, at each evaluation moment, the input data is fed into the data-based temperature model, which is trained to associate such input data with the temperature description for a specific component. The temperature model preferably has a CNN-LSTM architecture, as shown, for example, in Figure 3 .

[0045] Figure 3 The block diagram showing the data-based temperature model 10 is presented. The temperature model has a CNN block 11, which includes a convolutional neural network to identify patterns or features from the input data X t The CNN block 11 provides the input data features M t . These input data features M t are fed into the LSTM block 12 (LSTM: Long short term memory), which processes the input data features M t to obtain the temperature description T t at each evaluation moment t. The LSTM block 12 can model dynamic processes such that the past operating states of the technical system (vehicle 1) over time can be considered when modeling the current temperature description.

[0046] The temperature descriptions determined at each evaluation moment are either equally classified into temperature ranges output by the data-based temperature model 10 or associated with temperature ranges after each evaluation moment in the optional step S4. For each temperature range, the frequency is accumulated since the component in question started operating. The temperature ranges are defined as overlapping temperature ranges respectively above a pre-given temperature range. For example, the temperature ranges can be defined as so-called temperature load spectra, such as temperature ranges above 60 °C, above 65 °C, above 70 °C, above 75 °C, above 80 °C and above 85 °C, and the corresponding temperature descriptions are associated with these temperature ranges after each evaluation moment. A graph as shown, for example, in Figure 4 is obtained. Thus, each temperature description to be evaluated can be classified into one or more temperature ranges and thus the frequency in one or more temperature ranges is increased.

[0047] The level of the temperature load spectrum in the graph gives the detection frequency or the entire duration since start of operation, during which the component had a temperature within the corresponding temperature range. The entire duration can be obtained, for example, by multiplying the frequency by the duration of the evaluation time window. The threshold represented in steps gives the limit value of the accumulated operating time above the corresponding temperature limit.

[0048] Figure 5 An alternative illustration is shown, in which the temperature is plotted against the accumulated operating hours. The points shown as crosses show the frequencies at which the respective temperature limits are exceeded. The thick line G gives the corresponding limit value of the accumulated operating time at the respective temperature limit. If the frequency or the entire operating duration at a temperature exceeds the corresponding temperature limit at a location, the component is considered to be thermally fatigued. The step size of the grid points of the temperature (here 5 °C) can be chosen arbitrarily.

[0049] In step S5, it is checked whether the frequency of one of the temperature load spectra exceeds a pre-given frequency threshold for one of the observed temperature limits. If this is the case (selection "yes"), then in step S6 the corresponding warning can be signaled. For example, such a warning can include outputting a prompt, seeking a workshop or writing to a fault memory in the case of a vehicle as a technical system, so that the component can be checked or replaced during the next visit to the workshop. An emergency operation mode can also be adopted. If in step S5 it is determined that the frequency of one of the temperature load spectra does not exceed the pre-given frequency threshold (selection "no"), then the method continues with step S1.

[0050] Figure 6A flowchart is shown for elucidating a method for training a data-based temperature model. To this end, in step S11, a temperature sensor is set up for the component of interest in the vehicle and the vehicle is run in actual operation. The trends of the operating parameter changes and the environmental parameter changes are recorded here.

[0051] At the same time, in step S12, the temperature generated by the component in question is detected.

[0052] In step S13, the input data for successive evaluation time windows is determined from the trends of the operating parameter changes and the environmental parameter changes, and the input data for each evaluation time window is associated with the temperature of the component in question, which has said temperature, for example, at the end of the evaluation time window. The input data can be determined from the trends of the operating parameter changes and the environmental parameter changes in the corresponding evaluation time window as described above. These data constitute the training data set. In addition, a validation data set can be generated in the same way, with which the trained model can be checked later or the training process can be controlled using said validation data set.

[0053] In step S14, the data-based temperature model is trained in a manner known per se using the training data set. This training is preferably carried out outside the vehicle in the cloud using known methods, such as backpropagation, and results in model parameters that characterize the temperature model of the proposed architecture in question.

[0054] In step S15, the model parameters are transmitted from the cloud to the controller 4 in the vehicle 1 so that the corresponding evaluation can be carried out there as described above.

[0055] List of reference numerals

[0056] 1 Vehicle

[0057] 2 Active component

[0058] 3 Passive component

[0059] 4 Controller

Claims

1. A method for determining the thermal load of a passive component (3) in a technical system (1), the technical system having at least one active component (2) and at least one said passive component (3), the active component generating waste heat during its operation, the method having the following steps: continuously detecting (S1) operating parameters and environmental parameters as operating parameter trends and environmental parameter trends in successively consecutive evaluation time windows; Based on a data-based temperature model, a temperature description is determined (S3) in each of the successive evaluation time windows according to the corresponding trends of the operating parameter changes and the environmental parameter changes, where training the data-based temperature model to associate input data corresponding to or depending on the operating parameter trends and the environmental parameter trends with a temperature specification for the at least one passive component (3); signaling (S6) a warning about the thermal load of the at least one passive component (3) according to the temperature specification.

2. The method according to claim 1, wherein, The operating parameters of the operating parameter trends include one of the following parameters: the rotational speed of a motor as an active component, the load and motor power of the motor, the current and / or power of an electronic component, wherein the environmental parameters include at least one of the following parameters: external temperature, sunlight irradiation, wind speed, and air humidity.

3. The method according to claim 1 or 2, wherein Evaluating the operating parameter trends and environmental parameter trends to obtain input data, the input data including the values of the corresponding operating parameters and environmental parameters and / or at least one characteristic parameter at moments within the operating parameter trends and the environmental parameter trends in the evaluation time window, the characteristic parameter being derived from the operating parameter trends and environmental parameter trends within the evaluation time window as aggregated and / or statistical characteristics, in particular mean values or median values.

4. The method according to any one of claims 1 to 3, wherein The data-based temperature model corresponds to a CNN-LSTM model or a CNN-GRU model.

5. The method according to any one of claims 1 to 4, wherein The temperature specification corresponds to a temperature, in the case of the temperature as the temperature specification, associating a temperature range, the temperature range being given by a temperature lower limit, wherein for each temperature range a frequency is given, for an evaluation time window since the at least one passive component starts operating, the temperature being within the temperature range at the frequency, and wherein a warning about the thermal load of the at least one passive component is signaled according to the frequency in each temperature range of the temperature range.

6. The method according to any one of claims 1 to 4, wherein The temperature specification corresponds to a specification of a temperature range, the temperature range being given by a temperature lower limit, wherein for each temperature range the frequency for determining the specification of the temperature range for the evaluation time window is determined, and wherein a warning about the thermal load of the at least one passive component is signaled according to the frequency in each temperature range of the temperature range.

7. The method according to any one of claims 1 to 6, wherein Training the data-based temperature model outside the technical system (1) and transmitting the model parameters of the data-based temperature model to the technical system (1) after completion of the training.

8. A method for training a data-based temperature model for use with one of the methods according to any one of claims 1 to 7, the method having the following steps: Continuously detect (S11, S12) operating parameters and environmental parameters as trends of changes in operating parameters and trends of changes in environmental parameters in successively consecutive evaluation time windows, and determine the temperature description to which the passive component (3) belongs; Generate (S13) a training data set based on the trends of changes in the operating parameters, the trends of changes in the environmental parameters, and the temperature description to which it belongs; Train the data-based temperature model so as to associate input data corresponding to or depending on the trends of changes in the operating parameters and the trends of changes in the environmental parameters with a temperature description for the temperature of the at least one passive component (3).

9. The method according to claim 8, wherein, Transmit the model parameters of the data-based temperature model to the controller (4) of the system (1).

10. Device, in particular a controller, for determining the thermal load of a passive component (3) in a technical system (1), said technical system having at least one active component (2) and at least one of said passive components (3), the active component generating waste heat during its operation, wherein, The device is configured to: Continuously detect operating parameters and environmental parameters as trends of changes in operating parameters and trends of changes in environmental parameters in successively consecutive evaluation time windows; Determine a temperature description for each of the successively consecutive evaluation time windows based on the corresponding trends of changes in the operating parameters and the trends of changes in the environmental parameters by means of a data-based temperature model, wherein the data-based temperature model is trained so as to associate input data corresponding to or depending on the trends of changes in the operating parameters and the trends of changes in the environmental parameters with a temperature description for the temperature of the at least one passive component (3); Signal a warning regarding the thermal load of the at least one passive component (3) based on the temperature description.

Images