Electric drive system state monitoring method, device, equipment and medium

Abstract

The invention discloses an electric driving system state monitoring method, device and equipment and a medium, and relates to the field of electric driving system monitoring, and the method comprises the steps: obtaining a multi-source heterogeneous signal of an electric driving system; the multi-source heterogeneous signals comprise an electrical signal, a mechanical signal, a thermal signal and an environment signal; the multi-source heterogeneous signal is preprocessed, and a processed signal is obtained; performing dynamic weight multi-source signal fusion analysis processing on the processed signal to obtain operation state data of the electric drive system; performing cooperative judgment processing on the operation state data through a pre-constructed digital twinborn model to obtain a target state result; the digital twin model is constructed based on a three-dimensional geometric model, physical characteristics and an operation mechanism of the electric drive system. According to the technical scheme, the accuracy of state monitoring of the electric drive system is remarkably improved, and early faults and potential risks are effectively recognized.

Description

Technical Field

[0001] This application relates to the field of electric drive system monitoring technology, and in particular to a method, device, equipment and medium for monitoring the condition of an electric drive system. Background Technology

[0002] With the rapid development of new energy and intelligent manufacturing, electric drive systems, as core power units in new energy vehicles, industrial intelligent manufacturing equipment, rail transit traction devices, and ship propulsion systems, directly determine the safety, reliability, and operational efficiency of the entire equipment through the stability of their operational status. As these fields evolve towards higher power density, higher integration, and longer lifecycles, electric drive systems face increasingly complex operating conditions. They must withstand frequent start-stop shocks and dynamic load fluctuations, and adapt to harsh environments such as high and low temperatures, high humidity, electromagnetic interference (EMI), and dust corrosion. This significantly increases the risk of failures such as bearing wear, winding insulation aging, rotor eccentricity, and power device failure. Therefore, in order to improve the stability and reliability of electric drive system operation, condition monitoring of electric drive systems has become a core industry requirement.

[0003] Currently, related technologies rely on a single type of information to evaluate the operating status of electric drive systems, which fails to fully cover the multi-physical field coupling characteristics of electric drive systems, resulting in low accuracy in monitoring the status of electric drive systems. Summary of the Invention

[0004] The purpose of this application is to provide a method, apparatus, device, and medium for monitoring the condition of an electric drive system.

[0005] To achieve the above objectives, this application provides the following solution:

[0006] In a first aspect, this application provides a method for monitoring the state of an electric drive system, comprising:

[0007] Acquire multi-source heterogeneous signals from the electric drive system; the multi-source heterogeneous signals include: electrical signals, mechanical signals, thermal signals, and environmental signals; the electrical signals are used to reflect electromagnetic characteristics, the mechanical signals are used to reflect the structural operating state, the thermal signals are used to characterize the temperature field distribution, and the environmental signals are used to correlate with external influencing factors;

[0008] The multi-source heterogeneous signal is preprocessed to obtain the processed signal;

[0009] The processed signal is subjected to dynamic weighted multi-source signal fusion analysis to obtain the operating status data of the electric drive system.

[0010] The operating status data is collaboratively judged and processed by a pre-constructed digital twin model to obtain the target state result; the digital twin model is constructed based on the three-dimensional geometric model, physical characteristics and operating mechanism of the electric drive system.

[0011] In one embodiment, the multi-source heterogeneous signal is preprocessed to obtain a processed signal, including:

[0012] Based on the signal type of the multi-source heterogeneous signals, a corresponding denoising strategy is determined differently; the denoising strategy includes an adaptive Kalman filter algorithm, a wavelet packet threshold denoising algorithm, and a moving average filter algorithm.

[0013] According to the signal type and the corresponding denoising strategy, the multi-source heterogeneous signal is denoised to obtain the denoised signal.

[0014] By using a preset signal quality assessment model, the quality of the denoised signal is assessed based on the signal-to-noise ratio, data integrity, and sensor calibration deviation, resulting in quality coefficients and quality assessment results for each signal type.

[0015] When the quality assessment results of each signal type meet the quality assessment conditions, a normalization algorithm is used to normalize the signals that meet the quality assessment conditions to obtain the processed signals.

[0016] In one embodiment, the processed signal is subjected to dynamic weighted multi-source signal fusion analysis to obtain the operating status data of the electric drive system, including:

[0017] The processed signal is assigned a weight value based on the quality coefficient, and a weighted average algorithm is used to perform data fusion processing on the processed signal according to the weight value to obtain a fused signal.

[0018] Fault features are extracted from the fused signal, and the fault features are then subjected to dimensionality reduction processing to obtain the dimensionality-reduced features.

[0019] The reduced-dimensional features are dynamically assigned their contribution to fault diagnosis using an attention model, and the operating status data of the electric drive system is output. The operating status data includes: operating status and fault confidence. The operating status includes: normal operation or fault type.

[0020] In one embodiment, the attention model includes: an input layer, a bidirectional LSTM layer, an attention weight layer, a fully connected layer, and an output layer, wherein the bidirectional LSTM layer includes a forward LSTM and a backward LSTM;

[0021] The reduced-dimensional features are dynamically assigned their contribution to fault diagnosis using the attention model, and the operating status data of the electric drive system is output, including:

[0022] The dimensionality-reduced features are input into the attention model through the input layer, and the features of each time step are extracted from the forward and backward directions of the time series through the forward LSTM and backward LSTM, respectively.

[0023] The attention weights of the features at each time step are calculated using the softmax function through the attention weight layer.

[0024] The fully connected layer uses the ReLU activation function for processing, and the output layer outputs the operating status data of the electric drive system; the output dimension of the output layer is consistent with the number of fault types of the electric drive system.

[0025] In one embodiment, the operational status data is collaboratively processed using a pre-built digital twin model to obtain a target status result, including:

[0026] The operational status data is mapped to a digital twin model, and the signal response under different fault conditions is simulated in the digital twin model to obtain signal simulation data.

[0027] The simulated signal data is compared with the processed signal to obtain the comparison result;

[0028] The operating status data is corrected based on the comparison results to obtain the target status result.

[0029] In one embodiment, the digital twin model is constructed through the following steps:

[0030] The preset interface is called to create a three-dimensional geometric model of the motor, controller, and reducer;

[0031] The first preset program is called to construct the electromagnetic model of the motor, and the second preset program is called to construct the simulation model. The temperature field distribution of the windings and bearings is analyzed to obtain the digital twin model. The electromagnetic model of the motor is used to simulate the relationship between the stator current and the magnetic field distribution.

[0032] In one embodiment, after obtaining the target state result, the method further includes:

[0033] The target state results are correlated with historical data, and the correlation between "ambient humidity-winding temperature-insulation aging fault" and "vibration kurtosis-bearing wear degree" is mined using the Apriori algorithm; the historical data includes: multi-source heterogeneous data, preprocessed data, and digital twin model simulation parameters;

[0034] When the target state result includes fault type, a clustering algorithm is used to cluster time series data of the same type of fault to divide the fault development stage, and a decision tree algorithm is combined to obtain the feature threshold rules for each stage.

[0035] The correlation and fault stage characteristics are reverse-correlated with the parameter correction of the digital twin model to obtain the optimized digital twin model.

[0036] Secondly, this application provides an electric drive system condition monitoring device, the device comprising:

[0037] The acquisition module is used to acquire multi-source heterogeneous signals from the electric drive system; the multi-source heterogeneous signals include: electrical signals, mechanical signals, thermal signals, and environmental signals; the electrical signals are used to reflect electromagnetic characteristics, the mechanical signals are used to reflect the structural operating state, the thermal signals are used to characterize the temperature field distribution, and the environmental signals are used to correlate with external influencing factors.

[0038] The preprocessing module is used to preprocess the multi-source heterogeneous signal to obtain the processed signal;

[0039] The fusion analysis module is used to perform dynamic weighted multi-source signal fusion analysis on the processed signal to obtain the operating status data of the electric drive system.

[0040] The collaborative judgment module is used to perform collaborative judgment processing on the operating status data through a pre-constructed digital twin model to obtain the target status result; the digital twin model is constructed based on the three-dimensional geometric model, physical characteristics and operating mechanism of the electric drive system.

[0041] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the electric drive system state monitoring method described in any one of the above.

[0042] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the electric drive system state monitoring method described above.

[0043] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0044] This application provides a method, apparatus, device, and medium for monitoring the state of an electric drive system. The method includes: acquiring multi-source heterogeneous signals from the electric drive system; the multi-source heterogeneous signals include electrical signals, mechanical signals, thermal signals, and environmental signals; preprocessing the multi-source heterogeneous signals to obtain processed signals; performing dynamic weighted multi-source signal fusion analysis on the processed signals to obtain operating state data of the electric drive system; and performing collaborative judgment processing on the operating state data through a pre-constructed digital twin model to obtain a target state result; the digital twin model is constructed based on the three-dimensional geometric model, physical characteristics, and operating mechanism of the electric drive system. Compared with existing technologies, this solution comprehensively covers the multi-physics coupling operation characteristics of electric drive systems by collecting electrical signals reflecting electromagnetic properties, mechanical signals reflecting structural operating status, thermal signals characterizing temperature field distribution, and environmental signals. This breaks the limitation that a single signal can only reflect a local state, laying a data foundation for accurate monitoring. Furthermore, preprocessing of multi-source heterogeneous signals effectively eliminates noise interference, data deviations, and dimensional differences, ensuring the quality of the input signals for fusion analysis and avoiding fusion bias caused by poor-quality data. The use of dynamic weighted multi-source signal fusion analysis dynamically allocates weights based on the quality of different signals and their contribution to state assessment, highlighting the role of key signals and improving the accuracy of operating status data. Based on the three-dimensional geometric model, physical characteristics, and operating mechanism of the electric drive system, a digital twin model can be constructed to combine operating status data with virtual and real simulation parameters for collaborative judgment. This not only verifies the rationality of actual operating status data but also supplements implicit state information that is difficult to capture in actual monitoring through simulation, further correcting the state judgment results. Ultimately, this significantly improves the accuracy of electric drive system state monitoring and effectively identifies early faults and potential risks. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a schematic diagram of the application environment of an electric drive system state monitoring method according to an embodiment of this application;

[0047] Figure 2 A flowchart illustrating an embodiment of the electric drive system state monitoring method provided in this application;

[0048] Figure 3This is a flowchart illustrating a method for obtaining operating status data of an electric drive system by performing dynamic weighted multi-source signal fusion analysis on the processed signal according to an embodiment of this application.

[0049] Figure 4 A functional module diagram of an electric drive system status monitoring device provided in an embodiment of this application;

[0050] Figure 5 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0051] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0052] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0053] Related technologies rely on a single type of information to evaluate the operating status of electric drive systems, which fails to fully cover the multi-physical field coupling characteristics of electric drive systems, resulting in low accuracy in monitoring the status of electric drive systems.

[0054] Based on the above-mentioned deficiencies, this application provides a method for monitoring the state of an electric drive system. Compared with existing technologies, this solution comprehensively covers the multi-physics coupling operation characteristics of electric drive systems by collecting electrical signals reflecting electromagnetic properties, mechanical signals reflecting structural operating status, thermal signals characterizing temperature field distribution, and environmental signals. This breaks the limitation that a single signal can only reflect a local state, laying a data foundation for accurate monitoring. Furthermore, preprocessing of multi-source heterogeneous signals effectively eliminates noise interference, data deviations, and dimensional differences, ensuring the quality of the input signals for fusion analysis and avoiding fusion bias caused by poor-quality data. The use of dynamic weighted multi-source signal fusion analysis dynamically allocates weights based on the quality of different signals and their contribution to state assessment, highlighting the role of key signals and improving the accuracy of operating status data. Based on the three-dimensional geometric model, physical characteristics, and operating mechanism of the electric drive system, a digital twin model can be constructed to combine operating status data with virtual and real simulation parameters for collaborative judgment. This not only verifies the rationality of actual operating status data but also supplements implicit state information that is difficult to capture in actual monitoring through simulation, further correcting the state judgment results. Ultimately, this significantly improves the accuracy of electric drive system state monitoring and effectively identifies early faults and potential risks.

[0055] This application provides an embodiment of an electric drive system state monitoring method, which can be applied to, for example... Figure 1 The application environment of the electric drive system state monitoring method shown is as follows. This application environment includes a terminal 102, a server 104, and a data storage system. The terminal 102 communicates with the server 104 via a network. The data storage system can store the multi-source heterogeneous signals of the electric drive system acquired by the server 104. The data storage system can be set up independently, integrated into the server 104, or placed in the cloud or on another server. The terminal 102 can send the acquired multi-source heterogeneous signals of the electric drive system to the server 104. After acquiring the multi-source heterogeneous signals of the electric drive system, the server 104 performs preprocessing, fusion analysis, and collaborative judgment processing to obtain the target state result of the electric drive system. Furthermore, in some embodiments, the electric drive system state monitoring method can also be implemented independently by the server 104 or the terminal 102. For example, the terminal 102 can directly perform preprocessing, fusion analysis, and collaborative judgment processing to obtain the target state result of the electric drive system.

[0056] This electric drive system status monitoring method can be applied to various equipment equipped with electric drive systems, such as new energy vehicles, industrial motor drive equipment, and rail transit traction systems, to achieve real-time and accurate monitoring of the operating status of the electric drive system, timely detection of potential system faults, and ensure the safe and stable operation of the equipment.

[0057] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 104 can be implemented using a standalone server or a server cluster composed of multiple servers, or it can be a cloud server.

[0058] In one exemplary embodiment, such as Figure 2 As shown, a method for monitoring the state of an electric drive system is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps S201 to S204. Wherein:

[0059] Step S201: Obtain multi-source heterogeneous signals from the electric drive system. The multi-source heterogeneous signals include electrical signals, mechanical signals, thermal signals, and environmental signals. Electrical signals are used to reflect electromagnetic characteristics, mechanical signals are used to reflect the structural operating state, thermal signals are used to characterize the temperature field distribution, and environmental signals are used to correlate with external influencing factors.

[0060] It should be noted that the above-mentioned multi-source heterogeneous signals are signals obtained by monitoring the electric drive system through different sensors or other acquisition modules. These signals can include various forms, and this embodiment does not limit the form and source of the multi-source heterogeneous signals.

[0061] Optionally, the aforementioned multi-source heterogeneous signals can be obtained from external devices, imported from blockchain or databases, or obtained through various different sensors. This embodiment does not impose any limitations on the acquisition method of multi-source heterogeneous signals of the electric drive system.

[0062] Understandably, the aforementioned electrical signals include: motor stator three-phase current, rotor current, controller bus voltage, and IGBT module terminal voltage, acquired through Hall sensors and voltage Hall chips, with a sampling frequency of 5-20kHz. Mechanical signals may include motor bearing vibration acceleration, reducer gear meshing vibration, and motor speed, acquired through piezoelectric accelerometers (mounted in the bearing housing and reducer housing) and photoelectric speed sensors, with vibration signal sampling frequencies of 10-30kHz and speed signal sampling frequencies of 200-1000Hz. The aforementioned thermal signals include motor winding temperature, bearing temperature, and IGBT junction temperature, acquired through PT100 platinum resistance sensors and infrared temperature sensors, with a sampling frequency of 2-20Hz. The aforementioned environmental signals include ambient humidity, dust concentration, and electromagnetic interference intensity around the electric drive system, acquired through temperature and humidity sensors, laser dust sensors, and electromagnetic interference detectors, with a sampling frequency of 0.2-2Hz.

[0063] For example, when acquiring multi-source heterogeneous signals from an electric drive system through sensors, a current sensor can be connected in series with the three-phase stator winding circuit of the motor, and a voltage sensor can be connected in parallel across the controller bus. A vibration sensor is mounted on the front bearing housing, rear bearing housing, and input shaft end housing of the motor via a magnetic base, with a mounting surface flatness ≤0.05mm. An infrared temperature sensor can be aligned with the surface of the IGBT module, with the measurement distance controlled within a preset distance to avoid obstruction; this preset distance may, for example, include 5-10cm. An environmental sensor can be installed inside the electric drive system control cabinet, away from the cooling fan exhaust vent, to ensure a stable measurement environment.

[0064] In this embodiment, by acquiring multi-source heterogeneous data from the electric drive system, the acquired data can comprehensively cover the operating characteristics of the electric drive system with multi-physics coupling, breaking the limitation that a single signal can only reflect a local state, and providing good data guidance information for subsequent state monitoring.

[0065] Step S202: Preprocess the multi-source heterogeneous signal to obtain the processed signal.

[0066] It should be noted that, since different types of signals have different forms, it is necessary to use differentiated denoising algorithms for different signal noise characteristics in order to preprocess multi-source heterogeneous signals and obtain the processed signals.

[0067] In one embodiment, the multi-source heterogeneous signal is preprocessed to obtain a processed signal, including:

[0068] Based on the signal types of the multi-source heterogeneous signals, corresponding denoising strategies are determined differently. These strategies include adaptive Kalman filtering, wavelet packet thresholding, and moving average filtering. The multi-source heterogeneous signals are then denoised according to their signal types and corresponding denoising strategies to obtain denoised signals. A preset signal quality assessment model is used to evaluate the quality of the denoised signals based on signal-to-noise ratio, data integrity, and sensor calibration deviation, yielding quality coefficients and evaluation results for each signal type. When the quality evaluation results for each signal type meet the quality evaluation criteria, a normalization algorithm is used to normalize the signals that meet the criteria, resulting in processed signals.

[0069] In this embodiment, the multi-source heterogeneous signals of the electric drive system have distinctly different noise characteristics. Electrical signals are susceptible to high-frequency switching electromagnetic interference, mechanical signals are often mixed with low-frequency background vibrations and transient impact noise, and thermal and environmental signals often contain random pulse interference. Using a uniform denoising algorithm would lead to the loss of useful features or incomplete noise suppression. Therefore, this embodiment requires differentiated denoising strategies based on signal type. For example, for high-frequency random noise in electrical signals, an adaptive Kalman filter algorithm is used, dynamically adjusting the filter gain by estimating the noise covariance matrix in real time, accurately preserving key electrical features such as current harmonics while suppressing switching noise. For complex noise in mechanical signals, a wavelet packet thresholding denoising algorithm can be used, leveraging the wavelet packet's ability to finely divide the frequency domain to separate fault impact features from background noise, avoiding the attenuation of high-frequency impact signals by traditional filtering. For the slow-changing characteristics and pulse interference of thermal and environmental signals, a moving average filtering algorithm is used, smoothing pulse interference while maintaining the signal's trend characteristics by setting the window size. Finally, a specific denoising strategy is used to specifically denoise the signal, resulting in a denoised signal with sufficient noise suppression and complete useful features.

[0070] The size of this window can be customized according to actual needs. For example, 5-10 sampling points can be selected for thermal signals. Electrical signals include, for example, stator current and bus voltage; mechanical signals include, for example, bearing vibration and gear meshing vibration; thermal signals include, for example, winding temperature and IGBT junction temperature; and environmental signals include, for example, humidity and dust concentration.

[0071] After obtaining the denoised signal, its effectiveness needs to be verified using a pre-defined signal quality assessment model. This model constructs an evaluation system based on three core dimensions: signal-to-noise ratio (SNR), data integrity, and sensor calibration deviation. It performs SNR analysis, data missing rate assessment, and deviation calculation respectively to obtain a quality coefficient and a quality assessment result. The SNR reflects the separation effect between signal and noise; for example, for vibration signals, an SNR ≥ 30dB is considered acceptable. Data integrity is assessed through the data missing rate; for example, a data missing rate < 1% is considered acceptable. Sensor calibration deviation is calculated based on periodically calibrated data; for example, a deviation value < ± 2% is considered normal. These three factors are quantified into a quality coefficient and a quality assessment result within a 0-1 range. The closer the quality coefficient is to 1, the better the quality. The quality assessment result includes both acceptable and unacceptable results.

[0072] When the quality coefficient of a certain type of signal is greater than a preset threshold, indicating that the quality assessment result is qualified, it means that it meets the basic requirements for subsequent fusion analysis. At this time, a normalization algorithm is needed to eliminate the dimensional differences between different signals. For example, the unit of current signal is A and the unit of temperature signal is °C. By mapping the signal values ​​to the [0,1] interval, the key features are avoided from being "submerged" during fusion due to differences in numerical range. Finally, a processed signal with qualified quality and unified dimensions is obtained, providing high-quality and standardized data support for subsequent dynamic weight fusion analysis. When the quality coefficient is not greater than the preset threshold, indicating that the quality assessment result is unqualified, the sensor redundancy switching mechanism is triggered, the backup sensor data is used, and preprocessing and quality assessment are performed until the quality assessment is qualified. The preset threshold can be customized according to actual needs.

[0073] In this step, preprocessing multi-source heterogeneous data enables targeted denoising for different signal types, resulting in denoised signals. A quality assessment model is then used to evaluate the quality of the denoised signals from various dimensions, facilitating a more comprehensive determination of the quality assessment results. Subsequently, normalization processing is applied to ensure that the determined processed signals better meet the requirements of data fusion.

[0074] Step S203: Perform dynamic weighted multi-source signal fusion analysis on the processed signal to obtain the operating status data of the electric drive system.

[0075] It should be noted that after acquiring the processed signal, data fusion, feature fusion, and attention allocation through an attention model can be performed on the processed signal to obtain operational status data. This operational status data includes: operational status and corresponding fault confidence; operational status includes: normal operation or fault type. Fault types can include: electrical faults, mechanical faults, thermal faults, or environmentally related faults. Electrical faults can include, for example: inter-turn / phase-to-phase short circuits in motor windings, broken rotor bars / eccentricity, controller power device (IGBT / MOSFET) failure, and bus capacitor bulging / capacitance decay. Mechanical faults can include, for example: motor bearing wear / jamming / ball bearing peeling, and reducer gear meshing wear / broken teeth / tooth surface adhesion. Thermal faults can include, for example: motor winding overheating, bearing lubrication failure, excessively high IGBT module junction temperature, or cooling system blockage. Environmentally related faults can include: high humidity causing motor winding insulation breakdown, and strong electromagnetic interference causing controller signal turbulence. Fault confidence can be the probability of detecting a fault type.

[0076] In one embodiment, the processed signal undergoes dynamic weighted multi-source signal fusion analysis to obtain the operating status data of the electric drive system. Please refer to [link to relevant documentation]. Figure 3 As shown, the method includes:

[0077] Step S301: Assign weight values ​​to the processed signal based on the quality coefficient, and perform data fusion processing on the processed signal using a weighted average algorithm according to the weight values ​​to obtain the fused signal.

[0078] Step S302: Extract fault features from the fused signal and perform dimensionality reduction processing on the fault features to obtain the dimensionality-reduced features.

[0079] Step S303: The dimensionality-reduced features are dynamically assigned their contribution to fault judgment through an attention model, and the operating status data of the electric drive system is output.

[0080] Specifically, after acquiring the processed signal, which is of acceptable quality and normalized, weight values ​​are assigned based on the quality coefficient of each signal type. Signals with higher quality coefficients (closer to 1) are considered more resistant to noise and have higher data reliability, thus receiving higher weights. Conversely, signals with lower quality coefficients are considered less resistant to noise and have lower data reliability, thus receiving lower weights. For example, a vibration signal with a quality coefficient of 0.9 might be assigned a weight of 0.4, and a temperature signal with a quality coefficient of 0.7 might be assigned a weight of 0.2. After determining the weight values, a weighted average algorithm is used for data fusion. The fused signal is calculated by summing the products of "weight value × corresponding signal value," ensuring that high-quality signals dominate the fusion result while the influence of low-quality signals is appropriately weakened. For example, in bearing fault monitoring, if the quality coefficient of the vibration signal is 0.85, the corresponding weight value is 0.35; the quality coefficient of the current signal is 0.75, the corresponding weight value is 0.25; the quality coefficient of the temperature signal is 0.8, the corresponding weight value is 0.3; and the quality coefficient of the environmental signal is 0.6, the corresponding weight value is 0.1. The fused signal will primarily reflect the abnormal characteristics of vibration and temperature, while also taking into account changes in the current signal. This effectively avoids fusion bias caused by distortion of a single signal, providing more robust foundational data for subsequent feature extraction.

[0081] After obtaining the fused signal, the fault features extracted from it encompass multi-dimensional information, such as electrical, mechanical, and thermal features. These features may contain redundancy or noise interference. Dimensionality reduction algorithms, such as Principal Component Analysis (PCA), are used to retain core features with a high cumulative contribution rate, reducing computational complexity while focusing on key information. Electrical features could be, for example, harmonic amplitude; mechanical features could be vibration kurtosis; and thermal features could be temperature change rate. The dimensionality-reduced features are then input into an attention model, which dynamically assigns contribution weights by learning the correlation between different features and fault types. For example, when a bearing fault is identified, the attention weight of the vibration kurtosis feature is automatically increased to a value much higher than other features, while when a winding short circuit is identified, the weight of the current harmonic feature is strengthened to above 0.5. Finally, the model output includes the operating status and fault confidence. The operating status can include: early bearing wear or normal operation.

[0082] The attention model described above includes an input layer, a bidirectional LSTM layer, an attention weight layer, a fully connected layer, and an output layer. The bidirectional LSTM layer includes forward LSTM and backward LSTM. The attention model dynamically assigns the contribution of each feature to fault diagnosis based on the dimensionality reduction features, outputting the operating status data of the electric drive system, including:

[0083] The dimensionality-reduced features are input into the attention model through the input layer, and features at each time step are extracted from the forward and backward directions of the time series using forward LSTM and backward LSTM, respectively. The attention weights of the features at each time step are calculated using the softmax function through the attention weight layer. The ReLU activation function is used for processing through the fully connected layer, and the operating status data of the electric drive system is output through the output layer. The output dimension of the output layer is consistent with the number of fault types of the electric drive system.

[0084] Specifically, after obtaining the dimensionality-reduced features, which include, for example, principal components of core features such as the fused vibration kurtosis, current harmonics, and temperature change rate, the dimensionality-reduced features are fed into the model through the input layer. Although these features have been redundant, they still retain time-series attributes, such as the dynamic evolution of the system's operating state corresponding to feature changes at different times. At this point, the forward LSTM and backward LSTM work together. The forward LSTM is used to extract features frame by frame from the start to the end of the time series, capturing the increasing trend of features over time, such as the gradual increase in vibration kurtosis during bearing wear. The backward LSTM is used to extract features in reverse from the end to the start, mining the correlation information of features decreasing over time, and obtaining the features at each time point. For example, the pattern of the temperature change rate falling from its peak when the fault is alleviated. The combination of the two can fully cover the bidirectional correlation of time-series features, avoiding the omission of fault precursor information in the reverse time series by traditional unidirectional LSTM, and avoiding the possibility of misjudging random interference by only forward extraction. Combining the features extracted before and after the reverse extraction can confirm whether it is a fault initiation signal. The precursory information of this fault is, for example, an abnormal fluctuation in the current harmonics at a certain moment.

[0085] After the bidirectional LSTM outputs the hidden state features at each time step, an attention weight layer calculates the attention weight of each feature based on the softmax function. This assigns higher attention weights to features more critical to fault identification. For example, when identifying a short-circuit fault in a motor winding, the current harmonic features 10 seconds before the fault and the vibration peak features at the fault location contribute significantly more to the identification result than features during normal operation. In this case, the softmax function uses exponential normalization to calculate the weights of these two features, making them significantly higher than those at other times. This strengthens the influence of key time-series information and weakens the interference of irrelevant data in subsequent processing. Subsequently, the weighted hidden state features are input into a fully connected layer, where the ReLU activation function performs a non-linear mapping on the features. This effectively solves the gradient vanishing problem and highlights positive and effective fault features by suppressing negative eigenvalues. For instance, a positive rate of temperature change indicates a temperature increase, which may be associated with fault risk, and the ReLU function retains this feature; a negative rate of temperature change indicates a temperature decrease, which is usually not associated with faults, and the ReLU function sets it to 0, further optimizing the feature representation. Ultimately, the output of the fully connected layer is fed into the output layer, since the output dimension of the output layer is consistent with the number of fault types in the electric drive system.

[0086] For example, if the system needs to monitor four types of faults: bearing wear, winding short circuit, IGBT failure, and gear tooth breakage, and the output dimension is set to five dimensions, then four dimensions correspond to each type of fault, and one dimension corresponds to normal operation. The output is transformed into a probability distribution through the softmax function. The probability distribution of the output can be [0.02, 0.89, 0.03, 0.05, 0.01], which correspond to the probabilities of normal operation, bearing wear, winding short circuit, IGBT failure, and gear tooth breakage, respectively. The system can then directly output the operating status data of "operating status is bearing wear, fault confidence 0.89", achieving accurate matching and risk quantification of fault types.

[0087] In this embodiment, attention weights are assigned to the dimensionality-reduced features using an attention model, and the fully connected layer and output layer are used for processing. This allows for the assignment of corresponding weight values ​​to different features, and based on these attention weight values, the operating status data of the electric drive system can be accurately determined.

[0088] Step S204: The operating status data is collaboratively judged and processed using a pre-constructed digital twin model to obtain the target status result; the digital twin model is constructed based on the three-dimensional geometric model, physical characteristics and operating mechanism of the electric drive system.

[0089] Specifically, a digital twin model is first constructed. This can be achieved by calling a preset interface to create three-dimensional geometric models of the motor, controller, and reducer. A first preset program is then called to construct the motor's electromagnetic model, and a second preset program is called to construct a simulation model. The temperature field distribution of the windings and bearings is analyzed to obtain the digital twin model. The motor's electromagnetic model is used to simulate the relationship between stator current and magnetic field distribution. The preset interface can be a SolidWorks interface, the first preset program can be Maxwell software, and the second preset program can be ANSYS software. The three-dimensional geometric structure includes, for example, the dimensional parameters of the bearing's inner and outer rings and balls; the physical properties include, for example, the elastic modulus and friction coefficient of the bearing material; and the operating mechanism model includes, for example, the bearing contact mechanics model and vibration transmission model.

[0090] Furthermore, when mapping operational status data to the digital twin model, the core is to achieve a precise correlation between the "physical state and the virtual model." The digital twin model pre-integrates the three-dimensional geometric structure, physical properties, and operational mechanism model of the electric drive system. Through real-time data transmission protocols such as OPCUA, the characteristic parameters in the operational status data are synchronized to the corresponding components in the virtual model. For example, "vibration kurtosis 3.2, temperature 75℃" is synchronized to the vibration sensor and temperature monitoring point of the virtual bearing in the digital twin model. Subsequently, based on a pre-set fault condition library, which covers typical parameters of various faults such as bearing wear and winding short circuits, the model simulates the signal response at different stages of fault development. For example, for the operational state of "early bearing wear," the model simulates the changes in the time-domain impact signal and frequency-domain characteristic frequency collected by the virtual vibration sensor as the wear increases from 0.05mm to 0.1mm, generating signal simulation data containing time-series dimensions. The sampling frequency and characteristic dimensions of the simulation data are completely consistent with the actual processed signal.

[0091] After generating the simulated signal data, it can be compared with the processed actual signal in terms of "characteristic consistency" and "trend matching degree": Regarding characteristic consistency, the numerical deviations of key characteristic parameters between the simulated signal data and the processed data are compared, such as the actual processed vibration peak value being 4.8 m / s². 2 The peak vibration rate under the same working conditions in the simulation data is 5.1 m / s². 2 The calculated deviation rate was 6.25%. Simultaneously, the amplitude proportions of fault characteristic frequencies in the frequency domain were compared; for example, the amplitude proportion of the 230Hz frequency in the actual signal was 25%, while in the simulated data it was 23%. At the trend matching level, the dynamic time warping (DTW) algorithm was used to compare the temporal change trends of the two signals, yielding a comparison result. For example, if the actual temperature signal rose from 70℃ to 75℃ within 10 minutes, while the simulated data rose from 69℃ to 74℃ in the same timeframe, the trend similarity was judged to be 92%. This comparison result, for example, could be considered a high match if the feature deviation rate was <8% and the trend similarity was >90%.

[0092] During the process of correcting the operating status data based on the comparison results, if the comparison results show a high degree of matching, it indicates that the virtual simulation is consistent with the actual operating pattern. The original fault confidence level of 0.85 is then corrected to 0.92 to enhance the reliability of the judgment. If there is a feature deviation in the comparison results, the simulation data of different fault conditions in the digital twin model are retrieved and compared again. For example, it is found that the original simulation data of "early wear of bearing" deviates significantly from the actual signal, while the simulation data of "slight peeling of bearing outer ring" has a deviation rate of only 5%. In this case, the operating status is corrected to "slight peeling of bearing outer ring", and the confidence level is adjusted to 0.88, finally obtaining the target status result that is more consistent with the actual operating conditions.

[0093] In this embodiment, by mapping operational status data to a digital twin model and simulating fault condition signal responses, and then comparing and correcting the results with the actual processed signals, the accuracy and reliability of electric drive system status monitoring can be significantly improved. On the one hand, the signal simulation data generated by the digital twin model based on real physical properties and operating mechanisms can provide a "virtual reference system" for the actual operating status, compensating for the feature distortion problems that may occur due to local sensor failures or environmental interference when relying solely on actual signals. For example, when the actual vibration signal exhibits false peaks due to electromagnetic interference, comparison with simulated data can quickly identify deviations and eliminate interference. On the other hand, through dual comparison of feature consistency and trend matching, the rationality of the original operational status data can be verified, and the status judgment can be accurately corrected when deviations occur, avoiding misjudgments of fault types due to feature similarity. The final output target status result is more consistent with the actual operating conditions of the electric drive system, providing a more accurate decision-making basis for subsequent fault warnings and preventive maintenance, and effectively reducing the risk of equipment downtime and maintenance costs caused by monitoring deviations.

[0094] Furthermore, after obtaining the target state result, the above method also includes:

[0095] The target state results are correlated with historical data, and the Apriori algorithm is used to mine the correlations between "ambient humidity-winding temperature-insulation aging fault" and "vibration kurtosis-bearing wear degree". Historical data includes: multi-source heterogeneous data, preprocessed data, and simulation parameters of the digital twin model. When the target state results include fault types, a clustering algorithm is used to cluster time-series data of the same type of fault to divide the fault development stages. The characteristic threshold rules for each stage are obtained by combining the decision tree algorithm. The correlations and fault stage characteristics are then back-linked to the parameter correction of the digital twin model to obtain the optimized digital twin model.

[0096] In this embodiment, after obtaining the target state result, data mining processing can be performed. The target state result is used as the result label, and a mapping relationship is established with the corresponding multi-source heterogeneous raw data, preprocessed data, and digital twin model simulation parameters in the historical database to form a complete dataset of "state-data-parameter". On this basis, the Apriori algorithm is used to scan the dataset to count frequent itemsets and mine implicit associations. For example, for "ambient humidity-winding temperature-insulation aging fault", the algorithm will select the frequent itemset "humidity continuously > 85% (support 12%) and winding temperature > 120℃ (support 15%)", calculate its association confidence with "insulation aging fault", and clarify that "high humidity superimposed with high temperature" is the key condition for inducing insulation aging. For "vibration kurtosis-bearing wear degree", the association rule of "vibration kurtosis > 4.5 (support 18%)" and "bearing wear > 0.1mm (confidence 79%)" is mined, providing a basis for fault cause tracing and risk prediction.

[0097] When the target state result includes a fault type, the fault evolution pattern is further analyzed using clustering and decision tree algorithms. First, K-means clustering can be used to perform unsupervised clustering of time-series data for similar faults. Based on the similarity of time-series features, the faults are divided into three development stages: early, middle, and late. Each stage corresponds to a cluster. The early stage corresponds to a vibration kurtosis of 2.5-3.5 and a temperature <70℃; the middle stage corresponds to a vibration kurtosis of 3.5-5 and a temperature of 70-85℃; and the late stage corresponds to a vibration kurtosis >5 and a temperature >85℃. Then, the stage labels obtained from clustering and the feature data at each time point are input into the decision tree algorithm. By recursively partitioning the feature space, feature threshold rules for each stage are extracted. These threshold rules could be, for example, that vibration kurtosis ≤3.5 and temperature <70℃ indicate early wear; vibration kurtosis >5 or temperature >85℃ indicate late wear. Finally, the mined correlations and fault stage features are back-linked to the digital twin model to obtain the optimized digital twin model. The optimized model outputs new target state results, which in turn supplement the historical database as new data, providing richer associated samples for the next round of data mining, and realizing deep coupling and cyclical empowerment of data at each step.

[0098] In this embodiment, the environmental impact parameters of the model are corrected according to the correlation relationship. For example, the attenuation coefficient of winding insulation resistance under high humidity is adjusted from 0.02 to 0.03, which is closer to the actual fault causes. On the other hand, the fault simulation logic of the model is optimized according to the fault stage characteristics. For example, when simulating bearing wear, the vibration kurtosis of 3.5 is used as the trigger condition for the early stage. Finally, an optimized digital twin model that can more accurately simulate the actual fault conditions and is closer to the system operation law is obtained, which provides more reliable virtual simulation support for subsequent condition monitoring.

[0099] This application provides a method for monitoring the state of an electric drive system. The method includes: acquiring multi-source heterogeneous signals from the electric drive system; the multi-source heterogeneous signals include electrical signals, mechanical signals, thermal signals, and environmental signals; preprocessing the multi-source heterogeneous signals to obtain processed signals; performing dynamic weighted multi-source signal fusion analysis on the processed signals to obtain operating state data of the electric drive system; and performing collaborative judgment processing on the operating state data through a pre-constructed digital twin model to obtain a target state result; the digital twin model is constructed based on the three-dimensional geometric model, physical characteristics, and operating mechanism of the electric drive system. Compared with existing technologies, this solution comprehensively covers the multi-physics coupling operation characteristics of electric drive systems by collecting electrical signals reflecting electromagnetic properties, mechanical signals reflecting structural operating status, thermal signals characterizing temperature field distribution, and environmental signals. This breaks the limitation that a single signal can only reflect a local state, laying a data foundation for accurate monitoring. Furthermore, preprocessing of multi-source heterogeneous signals effectively eliminates noise interference, data deviations, and dimensional differences, ensuring the quality of the input signals for fusion analysis and avoiding fusion bias caused by poor-quality data. The use of dynamic weighted multi-source signal fusion analysis dynamically allocates weights based on the quality of different signals and their contribution to state assessment, highlighting the role of key signals and improving the accuracy of operating status data. Based on the three-dimensional geometric model, physical characteristics, and operating mechanism of the electric drive system, a digital twin model can be constructed to combine operating status data with virtual and real simulation parameters for collaborative judgment. This not only verifies the rationality of actual operating status data but also supplements implicit state information that is difficult to capture in actual monitoring through simulation, further correcting the state judgment results. Ultimately, this significantly improves the accuracy of electric drive system state monitoring and effectively identifies early faults and potential risks.

[0100] Based on the same inventive concept, this application also provides an electric drive system condition monitoring device for implementing the above-mentioned electric drive system condition monitoring device. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more electric drive system condition monitoring device embodiments provided below can be found in the limitations of the electric drive system condition monitoring method above, and will not be repeated here.

[0101] In one exemplary embodiment, such as Figure 4 As shown, an electric drive system condition monitoring device is provided, the device comprising:

[0102] The acquisition module 510 is used to acquire multi-source heterogeneous signals of the electric drive system. The multi-source heterogeneous signals include electrical signals, mechanical signals, thermal signals and environmental signals. The electrical signals are used to reflect electromagnetic characteristics, the mechanical signals are used to reflect the structural operating state, the thermal signals are used to characterize the temperature field distribution, and the environmental signals are used to correlate with external influencing factors.

[0103] Preprocessing module 520 is used to preprocess multi-source heterogeneous signals to obtain processed signals;

[0104] The fusion analysis module 530 is used to perform dynamic weighted multi-source signal fusion analysis on the processed signal to obtain the operating status data of the electric drive system.

[0105] The collaborative judgment module 540 is used to perform collaborative judgment processing on the operating status data through a pre-built digital twin model to obtain the target status result; the digital twin model is constructed based on the three-dimensional geometric model, physical characteristics and operating mechanism of the electric drive system.

[0106] As an optional implementation, the preprocessing module 520 is specifically used for:

[0107] Based on the signal type of the multi-source heterogeneous signals, the corresponding denoising strategies are determined differently; the denoising strategies include adaptive Kalman filtering algorithm, wavelet packet threshold denoising algorithm, and moving average filtering algorithm.

[0108] According to the signal type and the corresponding denoising strategy, the multi-source heterogeneous signal is denoised to obtain the denoised signal.

[0109] By using a preset signal quality assessment model, the quality of the denoised signal is assessed based on the signal-to-noise ratio, data integrity, and sensor calibration deviation, resulting in the quality coefficients and quality assessment results for each signal type.

[0110] When the quality assessment results of each signal type meet the quality assessment conditions, a normalization algorithm is used to normalize the signals that meet the quality assessment conditions, resulting in the processed signals.

[0111] As an optional implementation, the fusion analysis module 530 is specifically used for:

[0112] Weight values ​​are assigned to the processed signal based on the quality coefficient, and a weighted average algorithm is used to perform data fusion processing on the processed signal according to the weight values ​​to obtain the fused signal;

[0113] Fault features are extracted from the fused signal, and the fault features are then subjected to dimensionality reduction processing to obtain the dimensionality-reduced features.

[0114] The reduced-dimensional features are dynamically assigned their contribution to fault diagnosis using an attention model, and the operating status data of the electric drive system is output. The operating status data includes: operating status and fault confidence. The operating status includes: normal operation or fault type.

[0115] As an optional implementation, the fusion analysis module 530 is also used for:

[0116] The dimensionality-reduced features are input into the attention model through the input layer, and the features of each time step are extracted from the forward and backward directions of the time series respectively through forward LSTM and backward LSTM.

[0117] Attention weights for each time step are calculated using the softmax function in the attention weight layer.

[0118] The data is processed using the ReLU activation function through a fully connected layer, and the operating status data of the electric drive system is output through the output layer; the output dimension of the output layer is consistent with the number of fault types of the electric drive system.

[0119] As an optional implementation, the collaborative judgment module 540 is specifically used for:

[0120] The operational status data is mapped to a digital twin model, and the signal response under different fault conditions is simulated in the digital twin model to obtain signal simulation data.

[0121] The analog signal data is compared with the processed signal to obtain the comparison result;

[0122] The operational status data is corrected based on the comparison results to obtain the target status result.

[0123] As an alternative implementation method, a digital twin model is constructed through the following steps:

[0124] The preset interface is called to create a three-dimensional geometric model of the motor, controller, and reducer;

[0125] The first preset program is called to build the electromagnetic model of the motor, and the second preset program is called to build the simulation model. The temperature field distribution of the windings and bearings is analyzed to obtain the digital twin model. The electromagnetic model of the motor is used to simulate the relationship between the stator current and the magnetic field distribution.

[0126] As an optional implementation, the above-described apparatus is further used for:

[0127] The target state results are correlated with historical data, and the correlation between "ambient humidity-winding temperature-insulation aging fault" and "vibration kurtosis-bearing wear degree" is mined using the Apriori algorithm; historical data includes: multi-source heterogeneous data, preprocessed data, and simulation parameters of digital twin model;

[0128] When the target state result includes fault type, a clustering algorithm is used to cluster time series data of the same type of fault to divide the fault development stage, and a decision tree algorithm is combined to obtain the feature threshold rules for each stage.

[0129] By inversely linking the correlation and fault stage characteristics to the parameter correction of the digital twin model, an optimized digital twin model is obtained.

[0130] The electric drive system status monitoring device provided in this application comprehensively covers the multi-physical field coupling operation characteristics of the electric drive system by collecting electrical signals reflecting electromagnetic characteristics, mechanical signals reflecting structural operating status, thermal signals characterizing temperature field distribution, and environmental signals. This breaks the limitation that a single signal can only reflect a local state, laying a data foundation for accurate monitoring. Furthermore, the device preprocesses multi-source heterogeneous signals, effectively eliminating noise interference, data deviations, and dimensional differences, ensuring the quality of the input signals for fusion analysis and avoiding fusion deviations caused by poor-quality data. The use of dynamic weighted multi-source signal fusion analysis dynamically allocates weights based on the quality of different signals and their contribution to status assessment, highlighting the role of key signals and improving the accuracy of operating status data. Based on the three-dimensional geometric model, physical characteristics, and operating mechanism of the electric drive system, a digital twin model can be constructed to combine operating status data with virtual and real simulation parameters for collaborative judgment. This not only verifies the rationality of actual operating status data but also supplements implicit state information that is difficult to capture in actual monitoring through simulation, further correcting the state judgment results. Ultimately, this significantly improves the accuracy of electric drive system status monitoring and effectively identifies early faults and potential risks.

[0131] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 5 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores video tag processing data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When executed by the processor, the computer program implements a method for monitoring the status of an electrically driven system.

[0132] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0133] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0134] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0135] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0136] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0137] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0138] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0139] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0140] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for monitoring the state of an electric drive system, characterized in that, The electric drive system status monitoring method includes: Acquire multi-source heterogeneous signals from the electric drive system; the multi-source heterogeneous signals include: electrical signals, mechanical signals, thermal signals, and environmental signals; the electrical signals are used to reflect electromagnetic characteristics, the mechanical signals are used to reflect the structural operating state, the thermal signals are used to characterize the temperature field distribution, and the environmental signals are used to correlate with external influencing factors; The multi-source heterogeneous signal is preprocessed to obtain the processed signal; The processed signal is subjected to dynamic weighted multi-source signal fusion analysis to obtain the operating status data of the electric drive system. The target state result is obtained by collaboratively judging and processing the operating status data through a pre-constructed digital twin model; the digital twin model is constructed based on the three-dimensional geometric model, physical characteristics and operating mechanism of the electric drive system.

2. The method for monitoring the state of an electric drive system according to claim 1, characterized in that, The multi-source heterogeneous signal is preprocessed to obtain a processed signal, including: Based on the signal type of the multi-source heterogeneous signals, a corresponding denoising strategy is determined differently; the denoising strategy includes adaptive Kalman filtering algorithm, wavelet packet threshold denoising algorithm, and moving average filtering algorithm. According to the signal type and the corresponding denoising strategy, the multi-source heterogeneous signal is denoised to obtain the denoised signal. By using a preset signal quality assessment model, the quality of the denoised signal is assessed based on the signal-to-noise ratio, data integrity, and sensor calibration deviation, resulting in quality coefficients and quality assessment results for each signal type. When the quality assessment results of each signal type meet the quality assessment conditions, a normalization algorithm is used to normalize the signals that meet the quality assessment conditions to obtain the processed signals.

3. The method for monitoring the state of an electric drive system according to claim 2, characterized in that, The processed signal is subjected to dynamic weighted multi-source signal fusion analysis to obtain the operating status data of the electric drive system, including: The processed signal is assigned a weight value based on the quality coefficient, and a weighted average algorithm is used to perform data fusion processing on the processed signal according to the weight value to obtain a fused signal. Fault features are extracted from the fused signal, and the fault features are then subjected to dimensionality reduction processing to obtain the dimensionality-reduced features. The reduced-dimensional features are dynamically assigned their contribution to fault diagnosis using an attention model, and the operating status data of the electric drive system is output. The operating status data includes: operating status and fault confidence. The operating status includes: normal operation or fault type.

4. The method for monitoring the state of an electric drive system according to claim 3, characterized in that, The attention model includes: an input layer, a bidirectional LSTM layer, an attention weight layer, a fully connected layer, and an output layer. The bidirectional LSTM layer includes a forward LSTM and a backward LSTM. The reduced-dimensional features are dynamically assigned their contribution to fault diagnosis using the attention model, and the operating status data of the electric drive system is output, including: The dimensionality-reduced features are input into the attention model through the input layer, and the features of each time step are extracted from the forward and backward directions of the time series through the forward LSTM and backward LSTM, respectively. The attention weights of the features at each time step are calculated using the softmax function through the attention weight layer. The fully connected layer uses the ReLU activation function for processing, and the output layer outputs the operating status data of the electric drive system; the output dimension of the output layer is consistent with the number of fault types of the electric drive system.

5. The method for monitoring the state of an electric drive system according to claim 1, characterized in that, The operational status data is collaboratively processed using a pre-built digital twin model to obtain the target status result, including: The operational status data is mapped to a digital twin model, and the signal response under different fault conditions is simulated in the digital twin model to obtain signal simulation data. The simulated signal data is compared with the processed signal to obtain the comparison result; The operating status data is corrected based on the comparison results to obtain the target status result.

6. The method for monitoring the state of an electric drive system according to claim 5, characterized in that, The digital twin model is constructed through the following steps: The preset interface is called to create a three-dimensional geometric model of the motor, controller, and reducer; The first preset program is called to construct the electromagnetic model of the motor, and the second preset program is called to construct the simulation model. The temperature field distribution of the windings and bearings is analyzed to obtain the digital twin model. The electromagnetic model of the motor is used to simulate the relationship between the stator current and the magnetic field distribution.

7. The method for monitoring the state of an electric drive system according to claim 1, characterized in that, After obtaining the target state result, the method further includes: The target state results are correlated with historical data, and the correlation between "ambient humidity-winding temperature-insulation aging fault" and "vibration kurtosis-bearing wear degree" is mined using the Apriori algorithm; the historical data includes: multi-source heterogeneous data, preprocessed data, and digital twin model simulation parameters; When the target state result includes fault type, a clustering algorithm is used to cluster time series data of the same type of fault to divide the fault development stage, and a decision tree algorithm is combined to obtain the feature threshold rules for each stage. The correlation and fault stage characteristics are reverse-correlated with the parameter correction of the digital twin model to obtain the optimized digital twin model.

8. A condition monitoring device for an electric drive system, characterized in that, The electric drive system status monitoring device includes: The acquisition module is used to acquire multi-source heterogeneous signals of the electric drive system; the multi-source heterogeneous signals include: electrical signals, mechanical signals, thermal signals and environmental signals; the electrical signals are used to reflect electromagnetic characteristics, the mechanical signals are used to reflect the structural operating state, the thermal signals are used to characterize the temperature field distribution, and the environmental signals are used to correlate external influencing factors. The preprocessing module is used to preprocess the multi-source heterogeneous signal to obtain the processed signal; The fusion analysis module is used to perform dynamic weighted multi-source signal fusion analysis on the processed signal to obtain the operating status data of the electric drive system. The collaborative judgment module is used to perform collaborative judgment processing on the operating status data through a pre-constructed digital twin model to obtain the target status result; the digital twin model is constructed based on the three-dimensional geometric model, physical characteristics and operating mechanism of the electric drive system.

9. A computer device, comprising: The memory and processor contain a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the electric drive system state monitoring method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the electric drive system state monitoring method according to any one of claims 1-7.

Images