TBM electric drive system fault diagnosis method of adaptive sparse fusion network

Abstract

The invention discloses a TBM electric drive system fault diagnosis method of an adaptive sparse fusion network, and belongs to the technical field of TBM electric drive control. De-noising and time-frequency domain feature extraction are carried out on the multi-source operation data of the electric drive system, and a multi-modal fusion data set is constructed; establishing a multi-task diagnosis framework, and dynamically allocating an optimal modal combination through a learnable modal weight vector; a double-support vector quantile regression algorithm is put forward to fit upper and lower boundary thresholds of temperature change, and real-time early warning of temperature abnormity is achieved; constructing a task conflict graph based on the task gradient similarity, and controlling a neuron activation range of a conflict task by using a dynamic sparse path; and the fusion modulator performs confidence weighting and decision compensation on each task result, and outputs a fault type and a diagnosis basis. The problems that weak fault signals are covered under strong coupling of multiple motors, and the misdiagnosis and missed diagnosis rate is high are solved, the fault diagnosis robustness and accuracy are improved, and the operation stability and safety of a TBM electric drive system are guaranteed.

Description

Technical Field

[0001] This invention belongs to the technical field of TBM electric drive control, and in particular relates to a fault diagnosis method for TBM electric drive systems using an adaptive sparse fusion network. Background Technology

[0002] TBMs (Tunnel Boring Machines) are crucial equipment used in hard rock tunnel construction, boasting advantages such as high tunneling efficiency, good safety, and environmental friendliness, and are widely used in tunnel boring operations. The TBM's electric drive system, as the core power source, undertakes several key tasks, including driving the cutterhead rotation, propulsion system movement, and guide mechanism adjustment. Its stable operation is fundamental to ensuring normal tunneling. In the long-distance, high-intensity, and continuous operation environment of tunnel construction, a failure in the electric drive system can easily lead to problems such as cutterhead jamming, propulsion failure, and cooling system abnormalities, causing interruptions, delays, or even halts in tunneling operations, thus affecting project timelines and costs. Furthermore, electric drive system failures may be accompanied by phenomena such as electric arcing, overcurrent, and overheating, which can severely damage electrical equipment and even cause safety accidents such as fires, explosions, and electric shocks, posing a direct threat to operators and equipment. Therefore, the operating status of the TBM's electric drive system not only determines the operating efficiency and economy of the tunneling equipment but also directly relates to the operational safety at the construction site.

[0003] In recent years, research on fault modeling and health status monitoring of TBM electric drive systems has gradually gained attention. Some theoretical methods have been successfully applied to key component anomaly detection, abnormal data detection, motor temperature rise monitoring and control strategy optimization, effectively improving the fault response capability and system stability of the tunneling process and promoting the development of intelligent operation and maintenance of TBMs.

[0004] However, TBM electric drive systems typically involve 9-14 motors operating in concert, resulting in diverse fault types, large data volumes, and strong signal coupling. Traditional fault diagnosis methods usually focus on a single fault type, only monitoring short-circuit or open-circuit faults. When multiple fault types occur concurrently, misdiagnosis or missed diagnosis may occur. TBMs operate under harsh conditions, often accompanied by strong vibrations, high temperatures, high humidity, and heavy dust, leading to signals containing strong disturbances and noise. The coordination of multiple motors also introduces signal coupling problems, making it difficult for traditional methods to extract effective fault features, especially weak features in the early stages of faults, which are easily "submerged." Although some multi-task fault diagnosis models exist, these models are prone to problems such as task conflicts, unreasonable resource allocation, diagnostic delays, and low diagnostic accuracy when dealing with complex faults in TBM electric drive systems. Summary of the Invention

[0005] To address the above technical problems, this invention provides a fault diagnosis method for TBM electric drive systems using adaptive sparse fusion networks.

[0006] This invention adopts the following technical solution: a fault diagnosis method for a TBM electric drive system based on an adaptive sparse fusion network, comprising the following steps: The original data is denoised and effective signals are obtained. Time-domain features and frequency-domain features are extracted based on the signal features and fused into a multimodal fusion dataset. A multi-task diagnostic model is built, which includes a temperature anomaly detection auxiliary module and multiple task feature recognition modules. Each module shares the core diagnostic module while retaining an independent feature recognition branch. The multi-task diagnostic model is trained, and a learnable modality selection weight vector is introduced for each module. The optimal modality data combination is dynamically assigned to the corresponding module using the modality selection weight vector. The temperature anomaly detection auxiliary module uses a dual support vector quantile regression algorithm to fit the temperature change threshold and establishes a dynamic threshold update mechanism. The task feature recognition module calculates the conflict degree between tasks based on task gradient similarity and constructs a task conflict graph; the task conflict graph guides the generation of tasks. sparse mask parameters Preliminary decoding results were obtained by regulating the activation range of neurons; A temperature monitoring and fault correlation mechanism is established, and the initial decoding results are weighted by confidence and compensated by a fusion modulator to generate the final fusion decision.

[0007] In a further embodiment, the raw data includes: current data, torque data, and temperature data; The time-domain features include: peak value, skewness, and kurtosis; The frequency domain characteristics include: spectral energy, spectral bandwidth, and peak frequency.

[0008] In a further embodiment, for the task The allocation process for the optimal modal data combination is as follows: Define the selection weight vector by the task For modes Selection weight The composition is achieved by weighted fusion of multimodal features using the following formula: ; In the formula, For the task A weighted fusion mechanism for multimodal features. For the number of modes, , For modality The output of the feature extractor For modality The input to the feature extractor; The selected weight vector is optimized using gradient descent. Combined with the loss function for each task, the total loss is minimized using backpropagation and the Adam optimizer, and the selected weights are then updated. ; Updated selection weights The modality corresponding to the weight threshold is determined as the task. The optimal combination of modal data.

[0009] In a further embodiment, the fitting process for the temperature change threshold is as follows: A TSVQR algorithm supporting parallel computation of biquantiles is constructed, and a biobjective boundary function for the TSVQR algorithm is established, wherein the biobjective boundary function is a nonlinear fitting of kernel function mapping; the lower and upper bound quantiles are constrained according to the temporal characteristics of temperature data. Define the boundary constraints of the TSVQR algorithm, use the boundary constraints to simultaneously optimize the parameters of the dual-objective boundary function, and record the change data of the dual-objective boundary function; save the upper and lower bounds of the change data as temperature change thresholds; Correspondingly, the update process of the threshold dynamic update mechanism is as follows: If the number of temperature data exceeds the specified number, the input temperature data is updated and the TSVQR algorithm is run again to update the temperature change threshold.

[0010] In a further embodiment, the process for constructing the task conflict graph is as follows: For any task and tasks Calculate the gradient vector of the backbone network and the gradient vector of the backbone network Cosine similarity between And calculate the task and tasks The degree of conflict between , Main network gradient vector and the gradient vector of the backbone network The angle between them; Based on the conflict level Construct a task conflict matrix , R is the real number field, and N is the total number of tasks; Set conflict threshold Traverse the task conflict matrix Each element in, if Then, in representing the task and tasks Add a conflict edge between nodes and the degree of conflict Assigning weights to the edges creates a task conflict graph. : , A collection of task nodes. It is a set of weighted conflicting edges.

[0011] In a further embodiment, the sparse mask parameters Express it in the following form: ,in, The elements representing the mask parameters are either 0 or 1, where 0 represents masking irrelevant features and 1 represents preserving relevant features. The dimension of the mask.

[0012] In a further embodiment, the process of obtaining the preliminary decoding result is as follows: ; In the formula, Represents element-wise product. The basic feature extraction function, For the task The optimal combination of input modal data, These are the model parameters for the multi-task diagnostic model. For the task Preliminary decoding results.

[0013] In a further embodiment, the temperature monitoring and fault association mechanism is as follows: If the current temperature If the temperature change threshold is exceeded, the temperature-fault correlation judgment process is triggered. Based on the preliminary decoding results of each task feature recognition module, the following correlation verification is performed: Check the diagnostic confidence of each task. If the diagnostic confidence of a task is higher than the corresponding confidence threshold, it is determined that the temperature anomaly was caused by that task. Output the correlation results between the fault type and the temperature anomaly. If the diagnostic confidence level for all tasks is lower than the corresponding confidence threshold, an alert will be issued and manual intervention will be requested.

[0014] In a further embodiment, the process of confidence weighting and decision compensation for the preliminary decoding result is as follows: The corresponding task is obtained from the preliminary decoding result. Task results Task Results Diagnostic confidence The final fusion decision is generated using the following formula. : ,in, For the task Normalized weights, N is the total number of tasks.

[0015] The beneficial effects of this invention are as follows: This invention integrates multi-source operating data such as current, torque, and temperature of the TBM electric drive system with wavelet denoising and time-frequency domain analysis to ingeniously construct a multi-source information fusion dataset for the TBM electric drive system. Simultaneously, it constructs a high-dimensional, multi-modal fusion feature vector, fundamentally improving the distinguishability of fault features and effectively enhancing the detectability of fault signals.

[0016] This invention proposes a learnable mode selection weighting mechanism for three different tasks: short circuit, open circuit, and temperature anomaly. This mechanism enables the autonomous selection of the optimal mode for each task and the accurate construction of task-specific modes, effectively improving the accuracy of fault feature extraction and mitigating the problem of fault feature "overwhelming" caused by signal coupling in TBM multi-motor systems.

[0017] For the first time, temperature anomaly detection is combined with a fault diagnosis model as an auxiliary algorithm for fault diagnosis tasks. The TSVQR algorithm used for temperature anomaly detection has good human boundary fitting ability and data sensitivity, and can accurately fit the variation pattern of complex data and identify outliers. In addition, temperature anomalies, as a manifestation of faults in TBM electric drive systems, can effectively assist other fault diagnosis tasks and avoid misjudgments and missed judgments caused by complex data and heavy coupling.

[0018] This invention integrates gradient similarity calculations from different tasks to construct a task conflict matrix, and then uses this matrix to build a task conflict graph. The task conflict graph displays the degree of conflict between different tasks, enhancing model interpretability and effectively mitigating the shortcomings of traditional deep learning models in terms of interpretability, thus facilitating later maintenance by engineers.

[0019] This invention constructs a sparse mask to control the activation regions of neurons. Based on the inter-task conflict degree indicated by the task conflict graph, the sparse mask is controlled, allowing tasks with high similarity to share more neurons, and tasks with low similarity to share fewer neurons. Through sparse mask control, gradient interference between conflicting tasks is reduced, improving the stability of the feature extractor and its task generalization ability. The introduction of the sparse mask effectively alleviates the problem of complex fault information caused by multi-motor coordination in TBM electric drive systems, helps improve the efficiency of computing resource utilization, and makes computing resource allocation more rational. Attached Figure Description

[0020] Figure 1 This is a flowchart of the fault diagnosis method for TBM electric drive systems.

[0021] Figure 2 (a) in the diagram is the original waveform of the current signal.

[0022] Figure 2 (b) in the figure is the waveform of the current signal after filtering.

[0023] Figure 3 This is a graph showing the fitting results of the upper and lower bounds of the temperature change using the TSVQR algorithm.

[0024] Figure 4 This is a flowchart of the training process for the adaptive sparse fusion network and the temperature anomaly diagnosis auxiliary module. Detailed Implementation

[0025] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0026] Example 1 like Figure 1 As shown, a fault diagnosis method for a TBM electric drive system based on an adaptive sparse fusion network includes the following steps: The original data is denoised and effective signals are obtained. Time-domain features and frequency-domain features are extracted based on the signal features and fused into a multimodal fusion dataset. A multi-task diagnostic model is built, which includes a temperature anomaly detection auxiliary module and multiple task feature recognition modules. Each module shares the core diagnostic module while retaining an independent feature recognition branch. The multi-task diagnostic model is trained, and a learnable modality selection weight vector is introduced for each module. The optimal modality data combination is dynamically assigned to the corresponding module using the modality selection weight vector. The temperature anomaly detection auxiliary module uses a dual support vector quantile regression algorithm to fit the temperature change threshold and establishes a dynamic threshold update mechanism. The task feature recognition module calculates the conflict degree between tasks based on task gradient similarity and constructs a task conflict graph; the task conflict graph guides the generation of tasks. sparse mask parameters Preliminary decoding results were obtained by regulating the activation range of neurons; A temperature monitoring and fault correlation mechanism is established, and the initial decoding results are weighted by confidence and compensated by a fusion modulator to generate the final fusion decision.

[0027] It should be noted that the method of the present invention consists of the following tasks: (1) Short circuit fault diagnosis: mainly manifested as a sudden change or asymmetrical abnormality in the current of a certain phase, with current data as the main mode and transient detection used.

[0028] (2) Open circuit fault diagnosis: It is generally manifested as the disappearance of single-phase current and the increase of torque. The torque is the main mode, and its steady-state characteristic changes are explored.

[0029] (3) Temperature anomaly diagnosis: This task is a separate branch.

[0030] In a further embodiment, the raw data is collected from the TBM electric drive system switch cabinet, including current data, torque data, and temperature data; furthermore, each type of data may have 3600 samples, and each fault task type may have 600 samples.

[0031] To address the severe noise interference during TBM operation, this embodiment employs wavelet filtering technology to denoise the original data. Denoising experiments were conducted on the original data of the electric drive system by experimenting with different combinations of wavelet bases and decomposition levels. Simultaneously, the signal-to-noise ratio (SNR) of the denoised signal under each combination was calculated. Finally, the optimal db8 wavelet base and 3-level decomposition parameters were selected to complete the denoising operation on the original data. A comparison between the effective signal and the original signal is shown in Figure 2(a) and... Figure 2 As shown in (b) of the table, the commonly used wavelet functions and wavelet bases are listed in Table 1.

[0032] Table 1 Common wavelet functions and wavelet bases Based on the signal characteristics of the TBM electric drive system, the time-domain and frequency-domain characteristics of the signal are calculated respectively. The time-domain characteristics include: peak value, skewness, and kurtosis; the frequency-domain characteristics include: spectral energy, spectral bandwidth, and peak frequency.

[0033] Based on the fusion of the aforementioned effective signals, time-domain features, and frequency-domain features, a multimodal fusion dataset is obtained. To facilitate subsequent processing, each sample in the multimodal fusion dataset is labeled with a corresponding fault, and the dataset is divided into three parts: 70% training set, 20% validation set, and 10% test set. This ensures a balanced distribution of task categories across the datasets, meeting the model training requirements.

[0034] In a further embodiment, the constructed multi-modal fusion dataset described above is used to train the multi-task diagnostic model. During training, a learnable modality selection weight vector is introduced for each task in the model; this vector is optimized using a gradient descent algorithm, which can automatically learn and determine the optimal modality combination for a specific task, thereby improving the accuracy of task feature extraction. Modalities with larger weights are considered more important for that task.

[0035] To further understand, the allocation process for the optimal modal data combination is as follows: Define the selection weight vector by the task For modes Selection weight The composition is achieved by weighted fusion of multimodal features using the following formula: In the formula, For the task A weighted fusion mechanism for multimodal features. For the number of modes, , For modality The output of the feature extractor For modality The input to the feature extractor; The selected weight vector is optimized using gradient descent. Combined with the loss function for each task, the total loss is minimized using backpropagation and the Adam optimizer, and the selected weights are then updated. The loss function for each task can be a predefined loss function, such as cross-entropy loss for the fault detection task and binary cross-entropy loss for the temperature anomaly detection auxiliary module. Therefore, during the update, the parameters of the core diagnostic module and the parameters of each feature recognition branch are updated simultaneously. Finally, the updated selection weights are... The modality corresponding to the weight threshold is determined as the task. The optimal combination of modal data.

[0036] In a further embodiment, the fitting process for the temperature change threshold is as follows: A TSVQR algorithm (Dual Support Vector Quantile Regression) supporting parallel computation of dual quantiles is constructed, and a dual-objective boundary function for the TSVQR algorithm is established. This dual-objective boundary function is a nonlinear fit of a kernel function mapping. In this embodiment, the dual-objective boundary function is: ; ;in, This represents the input data for the temperature anomaly detection auxiliary module, such as the optimal modal data combination assigned above. and These are the upper boundary function for temperature and the lower boundary function for temperature, respectively. and All are weighting coefficients. and All are offset tops.

[0037] The lower and upper bound quantiles are constrained based on the temporal characteristics of the temperature data. In the application of this embodiment, the temperature changes in the TBM electric drive system are complex and nonlinear, making it difficult for existing conventional algorithms to fit the variation patterns at specified locations. Based on the temperature characteristics of the TBM electric drive system, quantiles of 0.1 and 0.99 are set, corresponding to the upper and lower parts of the data respectively, prompting the algorithm to focus on the boundaries of data changes.

[0038] Define the boundary constraints of the TSVQR algorithm, use the boundary constraints to simultaneously optimize the parameters of the dual-objective boundary function, and record the change data of the dual-objective boundary function; save the upper and lower bounds of the change data as temperature change thresholds.

[0039] In this embodiment, for the upper bound boundary function of temperature The boundary constraints are expressed as follows: ; For the temperature upper boundary function The boundary constraints are expressed as follows: .

[0040] In the boundary constraint formulas for the upper and lower bound boundary functions, and These represent the weight vectors of the upper and lower temperature boundary functions, respectively. and These represent the biases of the upper and lower temperature boundary functions, respectively. τ This is the quantile parameter, ranging from 0 to... τ <1; and These are all regularization parameters, used to control model complexity; and This represents an insensitive parameter that controls the regression bandwidth. Let represent the relaxation variables of the upper and lower bound boundary functions of temperature, respectively, and represent the upper and lower bound errors; A is the input matrix. Each row represents one sample; Y is the response variable. Each element is the output value of the corresponding sample; A vector consisting entirely of 1s , The number of training samples. n For feature dimension, This is the transpose of the vector. To be constrained by.

[0041] Therefore, the obtained upper and lower bounds of data change are saved as temperature change thresholds. Correspondingly, the update process of the threshold dynamic update mechanism is as follows: if the number of temperature data exceeds a specified number, the input temperature data is updated and the TSVQR algorithm is run again to update the temperature change threshold.

[0042] The TSVQR algorithm is used, and by setting quantiles, the algorithm focuses on different parts of the data, thus obtaining the upper and lower bounds of temperature changes. If the temperature exceeds the upper or lower bound, an alarm is triggered. Simultaneously, temperature also serves as an auxiliary feature for judging motor faults; if an abnormal temperature is detected, a diagnostic program is run. This effectively reduces the obscuring of fault signals by normal signals.

[0043] Based on the training of the temperature anomaly detection auxiliary module, the model parameters are fixed after the training reaches a preset number of iterations. For the task feature recognition module, the construction process of the task conflict graph is as follows: For any task and tasks Calculate the gradient vector of the backbone network and the gradient vector of the backbone network Cosine similarity between And calculate the task and tasks The degree of conflict between , Main network gradient vector and the gradient vector of the backbone network The angle between them; Furthermore, cosine similarity The calculation formula is as follows: .

[0044] Then, the degree of conflict The following formula is used to obtain: .

[0045] Based on the conflict level Construct a task conflict matrix , R is the real number field, and N is the total number of tasks; Set conflict threshold p Traverse the task conflict matrix Each element in, if Then, in representing the task and tasks Add a conflict edge between nodes and the degree of conflict Assigning weights to the edges creates a task conflict graph. : ; A collection of task nodes. It is a set of weighted conflicting edges.

[0046] Using the constructed task conflict graph as a guide, a task-related sparse mask parameter is introduced into the shared backbone network. This parameter controls whether neurons are active in gradient updates and predictions for specific tasks, so that neurons for tasks with high conflict overlap less, and tasks with low conflict share more neurons.

[0047] Therefore, the sparse mask parameters in this embodiment Express it in the following form: ,in, The elements representing the mask parameters are either 0 or 1, where 0 represents masking irrelevant features and 1 represents preserving relevant features. The dimension of the mask.

[0048] Correspondingly, the process of obtaining the preliminary decoding results is as follows: ; In the formula, Represents element-wise product. The basic feature extraction function, For the task The optimal combination of input modal data, These are the model parameters for the multi-task diagnostic model.

[0049] Based on this, the temperature monitoring and fault association mechanism is as follows: if the current temperature If the temperature change threshold is exceeded, the temperature-fault correlation judgment process is triggered. Based on the preliminary decoding results of each task feature recognition module, the following correlation verification is performed: Check the diagnostic confidence of each task. If the diagnostic confidence of a task is higher than the corresponding confidence threshold, it is determined that the temperature anomaly was caused by that task. Output the correlation results between the fault type and the temperature anomaly. If the diagnostic confidence level for all tasks is lower than the corresponding confidence threshold, an alert will be issued and manual intervention will be requested.

[0050] Finally, for cases where the results of a single task are unreliable, reliability is increased by weighting the confidence level of the preliminary decoding results and applying decision compensation. The process is as follows: Obtain the corresponding task from the preliminary decoding results. Task results Task Results Diagnostic confidence The final fusion decision is generated using the following formula. : ,in, For the task Normalized weights, .

[0051] In summary, this embodiment proposes a multi-task fault diagnosis method for TBM electric drive systems based on an adaptive sparse fusion network. It dynamically assigns optimal mode combinations for short-circuit faults, open-circuit faults, and temperature anomaly tasks using learnable modal weight vectors. A dual support vector quantile regression algorithm is proposed to fit upper and lower bound thresholds for temperature changes, enabling real-time early warning of temperature anomalies. A task conflict graph is constructed based on task gradient similarity, and dynamic sparse paths are used to control the activation range of neurons in conflicting tasks. Finally, a fusion modulator performs confidence weighting and decision compensation on the results of each task, outputting the fault type and diagnostic basis. This invention solves the problem of weak fault signals being masked and high misdiagnosis / missed diagnosis rates under strong coupling of multiple motors, improving the robustness and accuracy of fault diagnosis and ensuring the continuity and safety of TBM tunneling operations.

[0052] Example 2 This embodiment discloses a TBM electric drive system fault diagnosis system, used to implement the TBM electric drive system fault diagnosis method based on the adaptive sparse fusion network described in Embodiment 1, including: The first module is configured to denoise the original data and obtain effective signals, extract time-domain features and frequency-domain features based on the signal features, and fuse them into a multimodal fusion dataset. The second module is set up to build a multi-task diagnostic model that includes a temperature anomaly detection auxiliary module and multiple task feature recognition modules. Each module shares the core diagnostic module while retaining an independent feature recognition branch. The third module is designed to train the multi-task diagnostic model, introducing a learnable modality selection weight vector for each module and dynamically assigning the optimal modality data combination to the corresponding module using the modality selection weight vector. The temperature anomaly detection auxiliary module uses a dual support vector quantile regression algorithm to fit the temperature change threshold and establishes a dynamic threshold update mechanism. The task feature recognition module calculates the conflict degree between tasks based on task gradient similarity and constructs a task conflict graph. The task conflict graph guides the generation of tasks. sparse mask parameters Preliminary decoding results were obtained by regulating the activation range of neurons; The fourth module is set up to establish a temperature monitoring and fault correlation mechanism. It uses a fusion modulator to perform confidence weighting and decision compensation on the preliminary decoding results to generate the final fusion decision.

[0053] The above embodiments are only used to illustrate the design concept and features of the present invention, and their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. The protection scope of the present invention is not limited to the above embodiments. Therefore, all equivalent changes or modifications made based on the principles and design ideas disclosed in the present invention are within the protection scope of the present invention.

Claims

1. A fault diagnosis method for a TBM electric drive system based on an adaptive sparse fusion network, characterized in that, Includes the following steps: The original data is denoised and effective signals are obtained. Time-domain features and frequency-domain features are extracted based on the signal features and fused into a multimodal fusion dataset. A multi-task diagnostic model is built, which includes a temperature anomaly detection auxiliary module and multiple task feature recognition modules. Each module shares the core diagnostic module while retaining an independent feature recognition branch. The multi-task diagnostic model is trained, and a learnable modality selection weight vector is introduced for each module. The optimal modality data combination is dynamically assigned to the corresponding module using the modality selection weight vector. The temperature anomaly detection auxiliary module uses a dual support vector quantile regression algorithm to fit the temperature change threshold and establishes a dynamic threshold update mechanism. The task feature recognition module calculates the conflict degree between tasks based on task gradient similarity and constructs a task conflict map. The task conflict diagram serves as a guide for generating tasks. sparse mask parameters Preliminary decoding results were obtained by regulating the activation range of neurons; A temperature monitoring and fault correlation mechanism is established, and the initial decoding results are weighted by confidence and compensated by a fusion modulator to generate the final fusion decision.

2. The method for fault diagnosis of a TBM electric drive system based on an adaptive sparse fusion network according to claim 1, characterized in that, The raw data includes: current data, torque data, and temperature data; The time-domain features include: peak value, skewness, and kurtosis; The frequency domain characteristics include: spectral energy, spectral bandwidth, and peak frequency.

3. The fault diagnosis method for a TBM electric drive system based on an adaptive sparse fusion network according to claim 1, characterized in that, For the task The allocation process for the optimal modal data combination is as follows: Define the selection weight vector by the task For modes Selection weight The composition is achieved by weighted fusion of multimodal features using the following formula: ； In the formula, For the task A weighted fusion mechanism for multimodal features. For the number of modes, , For modality The output of the feature extractor For modality The input to the feature extractor; The selected weight vector is optimized using gradient descent. Combined with the loss function for each task, the total loss is minimized using backpropagation and the Adam optimizer, and the selected weights are then updated. ; Updated selection weights The modality corresponding to the weight threshold is determined as the task. The optimal combination of modal data.

4. The method for fault diagnosis of a TBM electric drive system based on an adaptive sparse fusion network according to claim 1, characterized in that, The fitting process for the temperature change threshold is as follows: A TSVQR algorithm supporting parallel computation of biquantiles is constructed, and a biobjective boundary function for the TSVQR algorithm is established, wherein the biobjective boundary function is a nonlinear fitting of kernel function mapping; the lower and upper bound quantiles are constrained according to the temporal characteristics of temperature data. Define the boundary constraints of the TSVQR algorithm, use the boundary constraints to simultaneously optimize the parameters of the dual-objective boundary function, and record the change data of the dual-objective boundary function; save the upper and lower bounds of the change data as temperature change thresholds; Correspondingly, the update process of the threshold dynamic update mechanism is as follows: If the number of temperature data exceeds the specified number, the input temperature data is updated and the TSVQR algorithm is run again to update the temperature change threshold.

5. The method for fault diagnosis of a TBM electric drive system based on an adaptive sparse fusion network according to claim 1, characterized in that, The process for constructing the task conflict graph is as follows: For any task and tasks Calculate the gradient vector of the backbone network and the gradient vector of the backbone network Cosine similarity between And calculate the task and tasks The degree of conflict between , Main network gradient vector and the gradient vector of the backbone network The angle between them; Based on the conflict level Construct a task conflict matrix , R is the real number field, and N is the total number of tasks; Set conflict threshold Traverse the task conflict matrix Each element in, if Then, in representing the task and tasks Add a conflict edge between nodes and the degree of conflict Assigning weights to the edges creates a task conflict graph. : , A collection of task nodes. It is a set of weighted conflicting edges.

6. The fault diagnosis method for a TBM electric drive system based on an adaptive sparse fusion network according to claim 1, characterized in that, The sparse mask parameters Express it in the following form: ,in, The elements representing the mask parameters are either 0 or 1, where 0 represents masking irrelevant features and 1 represents preserving relevant features. The dimension of the mask.

7. The method for fault diagnosis of a TBM electric drive system based on an adaptive sparse fusion network according to claim 1, characterized in that, The process of obtaining the preliminary decoding results is as follows: ； In the formula, Represents element-wise product. The basic feature extraction function, For the task The optimal combination of input modal data, These are the model parameters for the multi-task diagnostic model. For the task Preliminary decoding results.

8. The fault diagnosis method for a TBM electric drive system based on an adaptive sparse fusion network according to claim 1, characterized in that, The temperature monitoring and fault correlation mechanism is as follows: If the current temperature If the temperature change threshold is exceeded, the temperature-fault correlation judgment process is triggered. Based on the preliminary decoding results of each task feature recognition module, the following correlation verification is performed: Check the diagnostic confidence of each task. If the diagnostic confidence of a task is higher than the corresponding confidence threshold, it is determined that the temperature anomaly was caused by that task. Output the correlation results between the fault type and the temperature anomaly. If the diagnostic confidence level for all tasks is lower than the corresponding confidence threshold, an alert will be issued and manual intervention will be requested.

9. The fault diagnosis method for a TBM electric drive system based on an adaptive sparse fusion network according to claim 1, characterized in that, The confidence-weighted and decision-compensation process for the preliminary decoding results is as follows: The corresponding task is obtained from the preliminary decoding results. Task results Task Results Diagnostic confidence The final fusion decision is generated using the following formula. : ,in, For the task Normalized weights, N is the total number of tasks.

Images