Coal mine equipment fault diagnosis method and related device

Abstract

The invention discloses a coal mine equipment fault diagnosis method and a related device, and relates to the technical field of coal mine equipment fault diagnosis, and the method comprises the steps: obtaining and preprocessing historical coal mine equipment data; the preprocessed historical coal mine equipment data are input into a coal mine equipment fault diagnosis model, loss is calculated according to a coal mine equipment fault diagnosis result output by the coal mine equipment fault diagnosis model and a corresponding real label, model parameters are optimized based on the loss, and a trained coal mine equipment fault diagnosis model is obtained after iterative training is completed; wherein the coal mine equipment fault diagnosis model is a neural network model based on a dynamic kernel principal component constraint mechanism and a quantum de-coherence suppression mechanism; and acquiring target real-time coal mine equipment data, and performing fault diagnosis on the target real-time coal mine equipment data by using the trained coal mine equipment fault diagnosis model to obtain a final coal mine equipment fault diagnosis result. According to the invention, the accuracy of coal mine equipment fault diagnosis can be improved.

Description

Technical Field

[0001] This application relates to the technical field of coal mine equipment fault diagnosis, and particularly to a coal mine equipment fault diagnosis method and related device. Background Art

[0002] Coal mine equipment undertakes important tasks during the production process. The operating status of coal mine equipment is directly related to production safety and the lives of miners. Therefore, the fault diagnosis of coal mine equipment has always been an important research topic in the coal mining industry. Most traditional coal mine equipment fault diagnosis methods rely on manual experience, rule-based judgment, or simple statistical analysis. These methods often cannot cope with the complex, variable operating environment of coal mine equipment and the diversity of fault modes. Therefore, the accuracy of coal mine equipment fault diagnosis is relatively low. Summary of the Invention

[0003] The purpose of this application is to provide a coal mine equipment fault diagnosis method and related device, which can effectively improve the accuracy of coal mine equipment fault diagnosis.

[0004] To achieve the above purpose, the following solutions are provided in this application.

[0005] In the first aspect, this application provides a coal mine equipment fault diagnosis method, which specifically includes the following steps.

[0006] Obtain historical coal mine equipment data; the historical coal mine equipment data refers to the historical operation data of coal mine equipment, and each piece of historical coal mine equipment data is labeled with a true label reflecting the actual working status of the coal mine equipment.

[0007] Preprocess the historical coal mine equipment data to obtain preprocessed historical coal mine equipment data.

[0008] Input the preprocessed historical coal mine equipment data into a coal mine equipment fault diagnosis model, calculate the loss based on the coal mine equipment fault diagnosis result output by the coal mine equipment fault diagnosis model and the corresponding true label, optimize the model parameters based on the loss, and obtain a trained coal mine equipment fault diagnosis model after iterative training; wherein, the coal mine equipment fault diagnosis model is a neural network model based on a dynamic kernel principal component constraint mechanism and a quantum decoherence suppression mechanism. The dynamic kernel principal component constraint mechanism refers to dynamically adjusting the kernel function width according to the characteristics of different equipment fault types during the training process, and the quantum decoherence suppression mechanism refers to suppressing quantum noise in the weight update and gradient calculation during the training process.

[0009] Obtain the target real-time coal mine equipment data, and use the trained coal mine equipment fault diagnosis model to perform fault diagnosis on the target real-time coal mine equipment data to obtain the final coal mine equipment fault diagnosis result; the target real-time coal mine equipment data refers to the target operation data that is collected in real time and used for fault diagnosis of coal mine equipment.

[0010] Optionally, preprocess the historical coal mine equipment data to obtain the preprocessed historical coal mine equipment data, which specifically includes the following steps.

[0011] Perform data cleaning on the historical coal mine equipment data to obtain the historical coal mine equipment data after data cleaning.

[0012] Perform data normalization or standardization on the historical coal mine equipment data after data cleaning to obtain the normalized or standardized historical coal mine equipment data.

[0013] Perform feature extraction in the time domain and frequency domain on the normalized or standardized historical coal mine equipment data to obtain the historical coal mine equipment data features, which are used as the preprocessed historical coal mine equipment data.

[0014] Optionally, the dynamic kernel principal component constraint mechanism specifically includes the following content.

[0015] Use the Gaussian kernel function to perform mapping transformation on the coal mine equipment data, which is expressed as the following formula.

[0016] ; Among them, is the kernel function, which characterizes the input coal mine equipment data point and projections in the high-dimensional space; is the th coal mine equipment data sample; is the th coal mine equipment data sample, which characterizes the neighboring coal mine equipment data sample of.

[0017] Dynamically adjust the kernel width according to the characteristics of the coal mine equipment data, and the dynamic adjustment method of the kernel width is expressed as the following formula.

[0018] ; Among them, is the adjustment value of the kernel function width after the th iteration, which characterizes the adaptability of the kernel function; is the adjustment value of the kernel function width after the th iteration, which characterizes the adaptability of the kernel function; is the square of the Euclidean distance between the input coal mine equipment data and the current mapping result, representing the difference between coal mine equipment data; is the adjustment factor, controlling the adjustment speed of the kernel width; is the preset tolerance error, controlling the flexibility of dynamic adjustment; is the number of samples of coal mine equipment data input to the neural network in the current batch.

[0019] Optionally, the quantum decoherence suppression mechanism specifically includes the following content.

[0020] When dynamically adjusting the width of the kernel function, use the quantum decoherence suppression mechanism to update the weight of the kernel function. Let the decoherence effect of quantum noise be expressed as and its update process is expressed as the following formula.

[0021] ; where is the value of the kernel weight of the th coal mine equipment data sample at the th iteration, representing the weight after decoherence suppression; is the learning rate, controlling the amplitude of parameter update; is the gradient of the objective function, representing the direction of kernel function adjustment; is the quantum decoherence suppression term; is the hyperparameter controlling the noise suppression intensity, representing the intensity of quantum noise suppression.

[0022] Calculate the quantum decoherence suppression term using the following formula.

[0023] ; where is the sign function; is the parameter controlling the suppression intensity; is the weight value of the connection between the th neuron and the th neuron in the neural network at the th iteration; is the decoherence attenuation factor.

[0024] The gradient calculation in backpropagation is expressed as the following formula.

[0025] ; where is the gradient in backpropagation; is the loss function of the neural network, representing the error between the predicted value and the true value; is the L2 regularization term; is the weight of the neural network, representing the connection strength between the input features of the coal mine equipment data and the neurons.

[0026] Under the quantum decoherence suppression mechanism, the gradient calculation is expressed as the following formula.

[0027] ; where, is the gradient after quantum noise suppression, representing the weight update gradient adjusted by the quantum decoherence suppression mechanism; is the quantum noise model, representing the suppression effect of the noise.

[0028] The update process of the coal mine equipment data features is expressed as the following formula.

[0029] ; where, is the updated coal mine equipment data feature vector, representing the coal mine equipment data features after data feature mapping and quantum noise suppression; is the kernel function, representing the input coal mine equipment data points and projections in the high-dimensional space; is the coal mine equipment data feature vector mapped by the kernel function, representing the coal mine equipment data feature space after non-linear transformation; is the corresponding quantum decoherence suppression term.

[0030] Optionally, obtain the target real-time coal mine equipment data, and use the trained coal mine equipment fault diagnosis model to perform fault diagnosis on the target real-time coal mine equipment data to obtain the final coal mine equipment fault diagnosis result, which specifically includes the following steps.

[0031] Obtain the target real-time coal mine equipment data.

[0032] Preprocess the target real-time coal mine equipment data to obtain the preprocessed target real-time coal mine equipment data.

[0033] Input the preprocessed target real-time coal mine equipment data into the trained coal mine equipment fault diagnosis model to obtain the final coal mine equipment fault diagnosis result.

[0034] Optionally, preprocess the target real-time coal mine equipment data to obtain the preprocessed target real-time coal mine equipment data, which specifically includes the following steps.

[0035] Perform data cleaning on the target real-time coal mine equipment data to obtain the target real-time coal mine equipment data after data cleaning.

[0036] Perform data normalization or standardization on the target real-time coal mine equipment data after data cleaning to obtain the normalized or standardized target real-time coal mine equipment data.

[0037] Extract the time-domain and frequency-domain features from the normalized or standardized target real-time coal mine equipment data to obtain the target real-time coal mine equipment data features, which are used as the preprocessed target real-time coal mine equipment data.

[0038] Optionally, after the step of obtaining the trained coal mine equipment fault diagnosis model after iterative training, the coal mine equipment fault diagnosis method further includes the following steps.

[0039] Use the model incremental training mechanism driven by perturbation perception to continuously perform model incremental training on the trained coal mine equipment fault diagnosis model to obtain the coal mine equipment fault diagnosis model after model incremental training; the coal mine equipment fault diagnosis model after model incremental training is used as the trained coal mine equipment fault diagnosis model to perform fault diagnosis on the target real-time coal mine equipment data to obtain the final coal mine equipment fault diagnosis result.

[0040] Optionally, the model incremental training mechanism driven by perturbation perception specifically includes the following content.

[0041] Real-time monitor the distribution change of the input coal mine equipment data features, and judge whether the distribution change of the coal mine equipment data features satisfies the feature distribution drift trigger condition to obtain the first judgment result; the feature distribution drift trigger condition is the following formula.

[0042] ; Among them, represents the feature distribution drift trigger condition, is the Kullback-Leibler divergence; is the feature distribution of the current coal mine equipment data; is the historical distribution during model training; is the preset trigger threshold.

[0043] When the first judgment result is yes, it is determined that the distribution change of the coal mine equipment data features has a feature distribution drift. At this time, trigger the local perturbation reconstruction mechanism to perform perturbation incremental update on the affected hidden layer neurons. The weight of each neuron to be updated is expressed as the following formula.

[0044] ; Among them, is the original neural network connection weight; is the perturbation step size; is the perturbation gradient of the input feature, expressed as the following formula.

[0045] ; Among them, is the sliding mean of historical features, used to measure the feature offset direction; is the regularization adjustment factor, used to control the perturbation scale; is the sign function; is the loss function of the neural network, representing the error between the predicted value and the true value; is the th coal mine equipment data sample.

[0046] Introduce a memory retention regularization term to update the objective function, and the updated objective function is expressed as the following formula.

[0047] ; Among them, represents the updated objective function; is the key parameter snapshot of the original model; is the key parameter of the original neural network model; is the set of parameters that have the greatest impact on performance; is the retention intensity factor.

[0048] In a second aspect, the present application proposes a computer device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the computer program to implement the coal mine equipment fault diagnosis method described above.

[0049] In a third aspect, the present application proposes a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the coal mine equipment fault diagnosis method described above.

[0050] According to the specific embodiments provided by the present application, the present application has the following technical effects.

[0051] This application provides a coal mine equipment fault diagnosis method and related devices. A neural network model based on a dynamic kernel principal component constraint mechanism and a quantum decoherence suppression mechanism is used as the coal mine equipment fault diagnosis model. On the one hand, through the dynamic kernel principal component constraint mechanism, the width of the kernel function is dynamically adjusted according to the characteristics of different equipment fault types during the training process, enabling the coal mine equipment fault diagnosis model to adapt to different fault modes, effectively improving the generalization ability of the coal mine equipment fault diagnosis model, reducing interference between features, and enhancing the accuracy of coal mine equipment fault diagnosis. On the other hand, through the quantum decoherence suppression mechanism, quantum noise suppression is performed on the weight update and gradient calculation during the training process, which can effectively reduce the impact of quantum noise on model training, avoid noise amplification, and thus ensure the stability and accuracy of the coal mine equipment fault diagnosis model, improving the fault diagnosis accuracy of the coal mine equipment fault diagnosis model. Therefore, by introducing a dynamic kernel principal component constraint mechanism and a quantum decoherence suppression mechanism during the model training process, this application realizes continuous updating of coal mine equipment data features during the training process, enabling the coal mine equipment fault diagnosis model to better capture the diversity and complexity of coal mine equipment data, ensuring that the coal mine equipment fault diagnosis model can learn more accurate and effective features during the iterative training process, and obtaining more accurate and reliable coal mine equipment fault diagnosis results. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0053] Figure 1 It is an application environment diagram of a coal mine equipment fault diagnosis method provided by an embodiment of the present application.

[0054] Figure 2 It is a schematic flowchart of a coal mine equipment fault diagnosis method provided by an embodiment of the present application.

[0055] Figure 3 It is a feature space distribution diagram of the traditional DNN method provided by an embodiment of the present application.

[0056] Figure 4 It is a feature space distribution diagram of the method of the present invention provided by an embodiment of the present application.

[0057] Figure 5 It is a schematic diagram of the anti-noise performance of the traditional DNN method provided by an embodiment of the present application.

[0058] Figure 6A schematic diagram of the anti-noise performance of the method of the present invention provided in one embodiment of the present application.

[0059] Figure 7 A schematic diagram of how the kernel width and gradient modulus vary with training rounds provided in an embodiment of the present application.

[0060] Figure 8 A comparison diagram of gradient noise distribution between the method of the present invention and the traditional DNN method provided in one embodiment of the present application.

[0061] Figure 9 A schematic diagram of the structure of a computer provided in one embodiment of the present application. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0063] At present, with the development of sensor technology, coal mine equipment can collect a large amount of operating data in real time through a variety of sensors (such as vibration, temperature, pressure, current, speed, etc.), which can provide an important basis for fault diagnosis. At the same time, the rapid development of machine learning technology has also provided new ideas and methods for coal mine equipment fault diagnosis. The fault diagnosis method based on machine learning can be trained through a large amount of historical data to automatically identify the fault mode of the equipment and perform rapid diagnosis when the equipment is abnormal, which greatly improves the accuracy and efficiency of diagnosis. Traditional fault diagnosis methods lack effective feature fusion and multi-dimensional data processing capabilities when dealing with multi-sensor data of coal mine equipment. Existing methods often rely on single sensor data, or the fusion processing of multi-sensor data is not sufficient, which makes it difficult to fully capture the operating status and potential fault modes of the equipment. Therefore, how to effectively fuse the time domain and frequency domain features of multiple sensors to improve the accuracy of coal mine equipment fault diagnosis has become a technical problem that needs to be solved in this field.

[0064] In addition, the types of failures in coal mine equipment are complex and diverse. Traditional machine learning methods often lack adaptability to different types of failures when processing equipment data, resulting in weak generalization capabilities of fault diagnosis models. Existing diagnostic systems mostly rely on static feature selection and fixed algorithm structures, and fail to fully consider the diversity and complexity of coal mine equipment failure modes, resulting in low diagnostic accuracy of the model under different equipment and failure types. Therefore, how to improve the adaptability of the model to different types of failures through a dynamic adjustment mechanism has become one of the key technical issues to be solved in this application.

[0065] Traditional machine learning methods often do not consider the impact of quantum noise on the model during the training process. Especially when dealing with high-dimensional data and complex non-linear features, the noise has a significant impact on the stability and accuracy of model training. This noise interference may lead to unstable weight updates during the model training process, affecting the accuracy and reliability of fault diagnosis results. Therefore, this application needs to solve the problem of how to effectively suppress the interference of quantum noise on the neural network training process and ensure the stability and accuracy of training in the case of high-dimensional data.

[0066] Traditional coal mine equipment fault diagnosis methods usually have the problem of overfitting in the training process. Especially when facing complex data and diverse fault patterns, it often leads to the model being unable to accurately identify the actual fault state of the equipment, thus affecting the effect of fault diagnosis. Therefore, how to design a feature update and learning mechanism with strong adaptability to prevent overfitting and improve the robustness and generalization ability of the model is another key technical problem to be solved in this application.

[0067] To make the above objects, features, and advantages of this application more obvious and understandable, the following further details this application in conjunction with the accompanying drawings and specific embodiments.

[0068] The coal mine equipment fault diagnosis method provided by the embodiments of this application can be applied to, for example Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set separately, integrated on the server 104, placed on the cloud or other servers. The terminal 102 can send historical coal mine equipment data and target real-time coal mine equipment data to the server 104. After receiving the historical coal mine equipment data and target real-time coal mine equipment data, for the historical coal mine equipment data and target real-time coal mine equipment data, the server 104 first preprocesses the historical coal mine equipment data; inputs the preprocessed historical coal mine equipment data into the coal mine equipment fault diagnosis model, calculates the loss according to the coal mine equipment fault diagnosis result output by the coal mine equipment fault diagnosis model and the corresponding true label, optimizes the model parameters based on the loss, and obtains a trained coal mine equipment fault diagnosis model after iterative training; wherein, the coal mine equipment fault diagnosis model is a neural network model based on a dynamic kernel principal component constraint mechanism and a quantum decoherence suppression mechanism; then for the target real-time coal mine equipment data, uses the trained coal mine equipment fault diagnosis model to perform fault diagnosis on the target real-time coal mine equipment data, and obtains the final coal mine equipment fault diagnosis result. The server 104 can feedback the final coal mine equipment fault diagnosis result to the terminal 102. In addition, in some embodiments, the coal mine equipment fault diagnosis method can also be implemented by the server 104 or the terminal 102 alone. For example, the terminal 102 can directly perform coal mine equipment fault diagnosis processing on the historical coal mine equipment data and target real-time coal mine equipment data, or the server 104 can obtain the historical coal mine equipment data and target real-time coal mine equipment data from the data storage system and perform coal mine equipment fault diagnosis processing on the historical coal mine equipment data and target real-time coal mine equipment data.

[0069] Among them, the terminal 102 can be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, and Internet of Things devices. The server 104 can be implemented by an independent server or a server cluster composed of multiple servers, and can also be a cloud server.

[0070] In an exemplary embodiment, as Figure 2 shown, a coal mine equipment fault diagnosis method is provided. This method is executed by a computer device, and can specifically be executed alone by a computer device such as a terminal or a server, or jointly executed by a terminal and a server. In the embodiments of the present application, taking this method applied to Figure 1 the server 104 in

[0071] S1: Obtain historical coal mine equipment data.

[0072] Among them, the historical coal mine equipment data refers to the historical operation data of coal mine equipment, which is used in the model training stage. It is mainly to complete the model training of the coal mine equipment fault diagnosis model and obtain a trained coal mine equipment fault diagnosis model. Each piece of the historical coal mine equipment data is labeled with a true label reflecting the actual working state of the coal mine equipment.

[0073] S2: Preprocess the historical coal mine equipment data to obtain preprocessed historical coal mine equipment data.

[0074] In this embodiment, step S2 preprocesses the historical coal mine equipment data to obtain preprocessed historical coal mine equipment data, which specifically includes the following steps.

[0075] S21: Clean the historical coal mine equipment data to obtain the historical coal mine equipment data after data cleaning.

[0076] S22: Perform data normalization or standardization on the historical coal mine equipment data after data cleaning to obtain the normalized or standardized historical coal mine equipment data.

[0077] S23: Extract time-domain and frequency-domain features from the normalized or standardized historical coal mine equipment data to obtain historical coal mine equipment data features, which serve as the preprocessed historical coal mine equipment data.

[0078] S3: Input the preprocessed historical coal mine equipment data into the coal mine equipment fault diagnosis model, calculate the loss based on the coal mine equipment fault diagnosis result output by the coal mine equipment fault diagnosis model and the corresponding true label, optimize the model parameters based on the loss, and obtain a trained coal mine equipment fault diagnosis model after iterative training.

[0079] In this embodiment, the coal mine equipment fault diagnosis model is a neural network model based on a dynamic kernel principal component constraint mechanism and a quantum decoherence suppression mechanism. The dynamic kernel principal component constraint mechanism refers to dynamically adjusting the kernel function width according to the characteristics of different equipment fault types during the training process. The quantum decoherence suppression mechanism refers to suppressing quantum noise in the weight update and gradient calculation during the training process.

[0080] S4: Obtain target real-time coal mine equipment data, and use the trained coal mine equipment fault diagnosis model to diagnose the faults of the target real-time coal mine equipment data to obtain the final coal mine equipment fault diagnosis result.

[0081] Among them, the target real-time coal mine equipment data refers to the target operation data collected in real time for fault diagnosis of coal mine equipment, which is used in the actual application stage of the model. The main purpose is to utilize the trained coal mine equipment fault diagnosis model to complete the real-time fault diagnosis of coal mine equipment. The final coal mine equipment fault diagnosis result is the coal mine equipment fault diagnosis result corresponding to the target real-time coal mine equipment data.

[0082] In this embodiment, step S4 obtains the target real-time coal mine equipment data, and uses the trained coal mine equipment fault diagnosis model to perform fault diagnosis on the target real-time coal mine equipment data, obtaining the final coal mine equipment fault diagnosis result, which specifically includes the following steps.

[0083] S41: Obtain the target real-time coal mine equipment data.

[0084] S42: Preprocess the target real-time coal mine equipment data to obtain the preprocessed target real-time coal mine equipment data.

[0085] S43: Input the preprocessed target real-time coal mine equipment data into the trained coal mine equipment fault diagnosis model to obtain the final coal mine equipment fault diagnosis result.

[0086] In this embodiment, step S42 preprocesses the target real-time coal mine equipment data to obtain the preprocessed target real-time coal mine equipment data, which specifically includes the following steps.

[0087] S421: Clean the target real-time coal mine equipment data to obtain the target real-time coal mine equipment data after data cleaning.

[0088] S422: Perform data normalization or standardization processing on the target real-time coal mine equipment data after data cleaning to obtain the normalized or standardized target real-time coal mine equipment data.

[0089] S423: Extract the time-domain and frequency-domain features of the normalized or standardized target real-time coal mine equipment data to obtain the target real-time coal mine equipment data features, which are used as the preprocessed target real-time coal mine equipment data.

[0090] In an exemplary embodiment, after the step of obtaining the trained coal mine equipment fault diagnosis model after the iteration training in step S3 is completed, the coal mine equipment fault diagnosis method further includes the following steps.

[0091] Using a perturbation-aware driven model incremental training mechanism, continuously perform model incremental training on the trained coal mine equipment fault diagnosis model to obtain a coal mine equipment fault diagnosis model after model incremental training. Among them, the coal mine equipment fault diagnosis model after model incremental training is used as the trained coal mine equipment fault diagnosis model to perform fault diagnosis on the target real-time coal mine equipment data to obtain the final coal mine equipment fault diagnosis result.

[0092] In this embodiment, the perturbation-aware driven model incremental training mechanism specifically includes the following content.

[0093] Real-time monitor the distribution change of the input coal mine equipment data features, judge whether the distribution change of the coal mine equipment data features meets the feature distribution drift trigger condition, and obtain a first judgment result, including the following two situations.

[0094] (1) When the first judgment result is yes, it is determined that the distribution change of the coal mine equipment data features triggers feature distribution drift. At this time, trigger the local perturbation reconstruction mechanism, perform perturbation incremental update on the affected hidden layer neurons, introduce a memory retention regularization term to update the objective function, and use the updated objective function to train to obtain a coal mine equipment fault diagnosis model after model incremental training, and use the coal mine equipment fault diagnosis model after model incremental training as the trained coal mine equipment fault diagnosis model to perform fault diagnosis on the coal mine equipment.

[0095] (2) When the first judgment result is no, it is determined that the distribution change of the coal mine equipment data features does not have feature distribution drift. At this time, do not trigger the perturbation-aware driven model incremental training mechanism, and still continue to use the trained coal mine equipment fault diagnosis model to perform fault diagnosis on the coal mine equipment.

[0096] To make the technical solution of this embodiment clearer, the following takes the form of an example to detail the specific implementation process of the technical solution of this embodiment, including the following implementation steps.

[0097] S1: Collection and annotation of historical coal mine equipment data.

[0098] Perform real-time monitoring on coal mine equipment, and collect coal mine equipment data within a period of time as historical coal mine equipment data for model training, including various sensor data such as vibration, temperature, pressure, current, and rotation speed. Each sensor will record the operation data of the coal mine equipment in real time, and these historical coal mine equipment data need to be labeled with real labels according to the actual working state of the coal mine equipment. The labeling content includes normal operation state, different types of fault states, and corresponding fault levels, etc.

[0099] S2: Preprocessing of historical coal mine equipment data.

[0100] First, clean the historical coal mine equipment data to remove the noise and outliers generated during the acquisition process and ensure the accuracy of the coal mine equipment data. Then, data normalization or standardization is required so that sensor data in different dimensions can be compared and analyzed on the same scale. For multi-dimensional time series data, feature extraction techniques are used to extract meaningful features from the time domain and frequency domain, such as mean, variance, peak value, kurtosis, spectral energy, etc., thereby converting the original data into feature vectors that can be used for model training.

[0101] S3: Model training of the coal mine equipment fault diagnosis model.

[0102] In this embodiment, a neural network model is used as the coal mine equipment fault diagnosis model, and its training process is as follows.

[0103] S301: Initialize the parameters of the neural network.

[0104] At the beginning of training, determine the structure of the neural network, including the number of layers of the input layer, hidden layer, and output layer and the number of neurons in each layer.

[0105] In an exemplary embodiment, the structure of the neural network is as follows: the input layer contains the coal mine equipment data features of multiple sensors (such as vibration, temperature, pressure, etc.), the hidden layer contains multiple neurons for extracting the relationship of coal mine equipment data features, and the output layer corresponds to the fault categories of the coal mine equipment. The specific number of layers and neurons will be adjusted according to the complexity of the coal mine equipment and the specific requirements of the task. For example, the input layer has 10 neurons corresponding to the coal mine equipment data features collected by 10 sensors; the hidden layer has 2 layers, the first hidden layer has 20 neurons, and the second hidden layer has 15 neurons; the output layer has 3 neurons corresponding to 3 types of coal mine equipment fault types (such as mechanical faults, electrical faults, and hydraulic faults).

[0106] Among them, the number of hidden layers can be multiple, such as 5 layers.

[0107] During the training process of this embodiment, the weights and bias terms of the network start with random initialization to avoid symmetry breaking in the initial stage, and then the gradient descent algorithm is used for optimization.

[0108] In view of the characteristics of coal mine equipment data, the input coal mine equipment data is multi-dimensional coal mine equipment data collected by multiple sensors. Each coal mine equipment data sample contains different sensor data features (such as vibration, temperature, pressure, etc.). In one embodiment, the coal mine equipment data is time-series data collected by 6 types of sensors (vibration triaxial accelerometers, temperature sensors, pressure sensors, current sensors, rotational speed sensors, oil particle sensors). Each sensor extracts 5 time-domain data features (mean value, variance, peak value, kurtosis, waveform factor) and 3 frequency-domain data features (main frequency amplitude, spectral energy, frequency band entropy) within a 1-second time window, forming a 6×8 = 48-dimensional coal mine equipment data feature vector.

[0109] In this embodiment, the connection weights and bias terms in the neural network are randomly initialized, and an appropriate dynamic kernel principal component constraint mechanism is set according to the distribution of the training coal mine equipment data. The initialization of the neural network includes randomly initializing the weights and bias terms, which is expressed as the following formula.

[0110] (1).

[0111] (2).

[0112] Among them, is the weight of the neural network, representing the connection strength between the input features of the coal mine equipment data and the neurons; is the bias of the neural network, representing the offset of the neuron activation function; indicates that the weight initialization follows a normal distribution with a mean of 0 and a variance of to ensure the symmetry of the model.

[0113] In this embodiment, during the initialization process, it is also necessary to initialize the parameters of the kernel function and design a dynamic adjustment mechanism corresponding to the dynamic kernel principal component constraint mechanism. In one embodiment, a Gaussian kernel function is used for the mapping transformation of the coal mine equipment data, which is expressed as the following formula.

[0114] (3).

[0115] Among them, is the kernel function, representing the projection of the input coal mine equipment data points and in the high-dimensional space; is the th coal mine equipment data sample; is the th coal mine equipment data sample, representing the neighboring coal mine equipment data sample of

[0116] At this time, the kernel width is dynamically adjusted according to the characteristics of the coal mine equipment data of the coal mine equipment to adapt to different types of equipment failures. During the training process, the dynamic adjustment method of the kernel width is expressed by the following formula.

[0117] (4).

[0118] Where, is the adjustment value of the kernel function width after the th iteration, characterizing the adaptability of the kernel function; is the th coal mine equipment data sample; is the th coal mine equipment data sample, characterizing the nearest neighbor coal mine equipment data sample; is the adjustment value of the kernel function width after the th iteration, characterizing the adaptability of the kernel function; is the square of the Euclidean distance between the input coal mine equipment data and the current mapping result, characterizing the difference between coal mine equipment data; is the adjustment factor, controlling the adjustment speed of the kernel width; is the preset tolerance error, controlling the flexibility of dynamic adjustment; is the number of samples of the coal mine equipment data input to the neural network in the current batch. In this embodiment is set to 0.2, is set to 0.00001.

[0119] S302: Coal mine equipment data feature constraint and adaptive learning.

[0120] During the training process, each input feature of the coal mine equipment data will be non-linearly mapped through the dynamic kernel principal component constraint mechanism to ensure that the coal mine equipment data feature space can better adapt to the diversity and complexity of coal mine equipment failures. Each input vector is mapped to a higher-dimensional space through the kernel function in the coal mine equipment data feature space so that the neural network can learn more complex relationships of coal mine equipment data features.

[0121] At the same time, the coal mine equipment data feature mapping is represented by the kernel function, reducing the interference caused by quantum noise and ensuring the stability of the coal mine equipment data feature transformation process, which is expressed by the following formula.

[0122] (5).

[0123] Where, is the coal mine equipment data feature vector after being mapped through the kernel function (corresponding to the th coal mine equipment data sample), characterizing the coal mine equipment data feature space after non-linear transformation; is the input of the function; is the weight of the kernel function of the th coal mine equipment data sample, representing the contribution of each training coal mine equipment data sample to the feature space of coal mine equipment data; is the kernel function, representing the input coal mine equipment data point and

[0124] In this embodiment, during dynamic adjustment, the weights of the kernel function are updated through the quantum decoherence suppression mechanism. Let the decoherence effect of quantum noise be expressed as , and its update process is expressed as the following formula.

[0125] (6).

[0126] where is the value of the kernel weight of the th coal mine equipment data sample at the th iteration, representing the weight after decoherence suppression; is the learning rate, controlling the amplitude of parameter update; is the gradient of the objective function, representing the direction of kernel function adjustment; is the quantum decoherence suppression term, reducing the influence of noise; is the hyperparameter controlling the noise suppression intensity, representing the intensity of quantum noise suppression.

[0127] In this embodiment, the design of the quantum decoherence suppression term enables stronger suppression of small weights during weight update, preventing noise amplification while retaining the learning ability of key coal mine equipment data features. The calculation method is expressed as the following formula.

[0128] (7).

[0129] where is the decoherence attenuation factor (preset to 0.1), suppressing the quantum noise interference of large weight values through exponential decay; is the sign function, retaining the gradient direction, is the parameter controlling the suppression intensity; is the weight value of the connection between the th neuron and the th neuron in the neural network at the th iteration.

[0130] S303: Backpropagation and gradient adjustment.

[0131] During the backpropagation process, the traditional gradient calculation method will be affected by the quantum decoherence suppression mechanism to ensure the accuracy and stability of gradient calculation, thereby optimizing the weight update of the neural network, improving the convergence speed and accuracy of the training process. The gradient calculation method in backpropagation is expressed as the following formula.

[0132] (8).

[0133] Where, is the gradient in backpropagation, is the loss function of the neural network, representing the error between the predicted value and the true value; is the L2 regularization term to prevent overfitting; is the weight of the neural network, characterizing the connection strength between the input features of coal mine equipment data and neurons.

[0134] In this embodiment, under the quantum decoherence suppression mechanism, the gradient calculation method is adjusted to the following formula.

[0135] (9).

[0136] Where, is the gradient after quantum noise suppression, characterizing the weight update gradient adjusted by the quantum decoherence suppression mechanism; is the quantum noise model, representing the suppression effect of noise; is the hyperparameter controlling the noise suppression intensity, characterizing the intensity of quantum noise suppression. In this embodiment is set to 0.01.

[0137] The suppression of quantum noise is not only reflected in the adjustment of the kernel function, but also in the update process of each weight and gradient.

[0138] By introducing the quantum decoherence suppression mechanism in the backpropagation of the neural network in this embodiment, the influence of noise on the network training process can be reduced, ensuring more accurate model updates, and ultimately improving the accuracy of fault diagnosis of the coal mine equipment fault diagnosis model.

[0139] S304: Update and reorganization of coal mine equipment data features.

[0140] During the training process of the neural network, as the coal mine equipment data features of each layer are continuously updated and reorganized, the network gradually learns more effective representations of coal mine equipment data features.

[0141] At this time, through the dynamic kernel principal component constraint mechanism, the coal mine equipment data features will be continuously mapped to a higher-dimensional coal mine equipment data feature space to better capture the non-linear relationships in the coal mine equipment data.

[0142] Meanwhile, the quantum decoherence suppression mechanism plays a role in this process, reducing the impact of noise on the update of the data characteristics of coal mine equipment, and ensuring that each neuron can effectively learn the core characteristics of the coal mine equipment data. Therefore, there is a synchronous and cooperative correlation between the dynamic kernel principal component constraint mechanism and the quantum decoherence suppression mechanism. Through the coordinated cooperation between the dynamic kernel principal component constraint mechanism and the quantum decoherence suppression mechanism, the accuracy of the coal mine equipment fault diagnosis model can be effectively improved, and the accuracy of coal mine equipment fault diagnosis can be enhanced.

[0143] Based on this, the update process of the coal mine equipment data characteristics is expressed as the following formula.

[0144] (10).

[0145] Among them, is the updated coal mine equipment data characteristic vector, representing the coal mine equipment data characteristics after data characteristic mapping and quantum noise suppression; is the coal mine equipment data characteristic vector after being mapped by the kernel function, representing the coal mine equipment data characteristic space after nonlinear transformation; is the corresponding quantum decoherence suppression term, representing the reduction of the influence of coal mine equipment data characteristic noise through quantum noise suppression.

[0146] S305: Multi-level integration and final output stage.

[0147] After multiple rounds of optimization and coal mine equipment data characteristic processing, the neuron information of all layers is finally integrated into the output layer of the neural network.

[0148] The output layer of the neural network performs a nonlinear transformation using a specific activation function and generates the final classification result. In this process, the dynamic kernel principal component constraint mechanism and the quantum decoherence suppression mechanism act together to ensure the accurate modeling of complex relationships by the neural network algorithm. The calculation formula for the output result is as follows.

[0149] (11).

[0150] Among them, is the final predicted output, representing the prediction result of the coal mine equipment fault diagnosis model; is the Sigmoid activation function, representing the nonlinear transformation process; is the weight of the neural network, representing the connection strength between the input and the output; is the updated coal mine equipment data characteristic vector, representing the coal mine equipment data characteristics after data characteristic mapping and quantum noise suppression; is the bias term of the neural network, representing the activation offset of the neuron.

[0151] S306: Repeat the above steps in step S3 until the preset iteration stop condition is met, indicating that the model training is completed. In one embodiment, the preset iteration stop condition is reaching the preset maximum number of iterations, and the preset maximum number of iterations is set to 1000 times.

[0152] S4: Coal mine equipment fault diagnosis.

[0153] After the coal mine equipment fault diagnosis model is trained, obtain the target real-time coal mine equipment data according to the method in step S1, preprocess the target real-time coal mine equipment data according to the method in step S2, and use the trained coal mine equipment fault diagnosis model to perform coal mine equipment fault diagnosis on the preprocessed target real-time coal mine equipment data to obtain the final coal mine equipment fault diagnosis result.

[0154] S5: Model incremental training.

[0155] For new fault types or working condition changes that occur in the coal mine equipment operation environment, this embodiment proposes a perturbation-aware driven model incremental training mechanism to achieve continuous learning and online adaptation of the coal mine equipment fault diagnosis model in practical applications.

[0156] In some scenarios, it is not necessary to perform full-scale training on the original coal mine equipment fault diagnosis model. After the coal mine equipment fault diagnosis model is deployed, the system monitors the distribution change of the input coal mine equipment data features in real time. If it is detected that the distribution change of the input coal mine equipment data features has a feature distribution drift, that is, the trigger condition for the feature distribution drift is as follows.

[0157] (12).

[0158] Where, represents the trigger condition for feature distribution drift; is the Kullback-Leibler divergence, which measures the difference between the distributions of new and old input data; is the feature distribution of the current coal mine equipment data; is the historical distribution during model training; is the preset trigger threshold.

[0159] When it is determined that the distribution change of the coal mine equipment data features has a feature distribution drift, the system will automatically trigger the perturbation-aware driven model incremental training mechanism. The perturbation-aware driven model incremental training mechanism is a local perturbation reconstruction mechanism, which is used to perform perturbation incremental update on the affected hidden layer neurons after perceiving the feature distribution drift. The weight of each neuron that needs to be updated is expressed as follows.

[0160] (13).

[0161] Among them, is the original neural network connection weight, is the perturbation step size; is the perturbation gradient of the input feature, which is defined as the following formula.

[0162] (14).

[0163] Among them, is the moving average of the historical feature, which is used to measure the feature offset direction; is the regularization adjustment factor, which is used to control the perturbation scale.

[0164] In this embodiment, to prevent the model from catastrophic forgetting under new tasks, a memory retention regularization term is introduced in the incremental training stage, and its objective function is updated as the following formula.

[0165] (15).

[0166] Among them, represents the updated objective function; is the key parameter snapshot of the original neural network model; is the key parameter of the original neural network model; is the set of parameters that have the greatest impact on performance, which is screened by the Fisher information matrix; is the retention intensity factor.

[0167] In this embodiment, the real-time monitored target real-time coal mine equipment data is input into the trained coal mine equipment fault diagnosis model. The trained coal mine equipment fault diagnosis model will predict and classify the operating state of the coal mine equipment according to the coal mine equipment data features and rules it has learned. Specifically, the trained coal mine equipment fault diagnosis model will perform a non-linear transformation on the input coal mine equipment data features. After being processed by a multi-layer neural network, the fault category of the equipment is finally output. Through the activation function of the output layer, the trained coal mine equipment fault diagnosis model can accurately judge whether the coal mine equipment is currently in a fault state.

[0168] Traditional neural network models usually rely on fixed feature extraction and linear mapping when dealing with coal mine equipment fault diagnosis. In this embodiment, by adopting a dynamic kernel principal component constraint mechanism, non-linear mapping is achieved, and the width of the kernel function can be dynamically adjusted according to the characteristics of different equipment fault types to adapt to different fault modes. This mechanism can effectively improve the generalization ability of the model and reduce the interference between features. Moreover, during the model training process, the quantum decoherence suppression mechanism can effectively reduce the impact of quantum noise on model training. By suppressing quantum noise in the weight update and gradient calculation during the training process, noise amplification can be avoided, ensuring the stability and accuracy of the coal mine equipment fault diagnosis model and improving the accuracy of coal mine equipment fault diagnosis. The feature mapping in this embodiment not only depends on static feature extraction methods, but also continuously updates the coal mine equipment data features during the training process through the dynamic kernel principal component constraint mechanism and the quantum decoherence suppression mechanism. The adaptive learning mechanism can better capture the diversity and complexity of coal mine equipment data, ensuring that the coal mine equipment fault diagnosis model can learn more accurate and effective features in multiple rounds of iteration. This embodiment combines various sensor data (such as vibration, temperature, pressure, current, etc.) and time-domain and frequency-domain features. By fusing multi-dimensional feature data, it ensures that the coal mine equipment fault diagnosis model can comprehensively capture the operating state and potential faults of coal mine equipment and can perform fault diagnosis more accurately. In addition, the gradient calculation in the backpropagation process of this embodiment is optimized by adopting the quantum noise suppression mechanism, thus ensuring the accuracy and convergence speed during the training process of the coal mine equipment fault diagnosis model, effectively reducing noise interference during training, ensuring that each weight update can be carried out in the correct direction, and further ensuring the accuracy of coal mine equipment fault diagnosis.

[0169] To verify the effectiveness of the method of the present invention, the method of the present invention is compared with the traditional DNN (Deep Neural Network) method in the following comparative experiments.

[0170] 1) Through the visualization of the three-dimensional feature space distribution, compare the feature extraction capabilities of the method of the present invention and the traditional DNN method. To verify the optimization effect of the dynamic kernel principal component constraint mechanism on the feature space structure, the separability of different fault types in the feature space is mainly investigated. Figure 3 It is the feature space distribution diagram of the traditional DNN method. Figure 4This is the feature space distribution diagram of the method of the present invention. Among them, data points of different colors represent not in the same plane in three-dimensional space. The darker the color of the data point, the closer the position of the feature space distribution of the data point is to the bottom layer. It can be seen that the features extracted by the method of the present invention present a clear time-series spiral structure, and data points of different fault categories form obviously separated cluster distributions in three-dimensional space, while the feature distribution of the traditional method presents a state of disordered random scatter points. This difference intuitively reflects that the dynamic kernel principal component constraint mechanism can transform the original sensor data into a feature representation with high separability through non-linear mapping, laying a solid foundation for subsequent classification tasks.

[0171] 2) The three-dimensional response surface was used to compare the performance stability of different methods in the complex noise environment. The experiment focused on verifying the dual resistance ability of the quantum decoherence suppression mechanism to sensor noise and quantum noise, simulating complex noise interference in industrial scenarios. Figure 5 This is the schematic diagram of the noise resistance performance of the traditional DNN method. Figure 6 This is the schematic diagram of the noise resistance performance of the method of the present invention. The results show that with the synchronous increase of the intensities of the two types of noise, the classification accuracy of the method of the present invention only shows a gentle decline, and its response surface presents a stable plateau feature; while the performance of the traditional method deteriorates sharply with the increase of noise, forming a steep descent gradient, indicating that the quantum decoherence suppression mechanism can effectively separate the noise signal and the real feature, and still maintain a reliable feature learning ability in a strong interference environment.

[0172] 3) The co-evolution law of the kernel function adjustment mechanism and the training process was revealed through the dynamic parameter change curve. The experiment tracked the real-time changes of the kernel width parameter and the gradient norm length during the training process, aiming to clarify the improvement effect of the dynamic adjustment mechanism on the training stability. As Figure 7 shown, the kernel width parameter shows regular oscillatory decay in the middle and early training stages, forming an accurate match with the gradient convergence process, and finally stabilizing in the theoretically expected interval. This adaptive adjustment characteristic indicates that the kernel function parameter can automatically optimize the mapping scale according to the feature learning progress, avoiding both the unstable fluctuations in the initial training and the excessive contraction in the later training, and realizing the dynamic matching of the feature space scale and the model convergence state.

[0173] 4) The anti-interference abilities of different methods were quantitatively compared through the gradient noise distribution histogram. The experiment focused on the quantum noise interference in the neural network training process, and mainly analyzed the statistical distribution characteristics of the gradient update amount. Figure 8This is a comparison chart of the gradient noise distributions between the method of the present invention and traditional DNN methods. It can be seen that the gradient noise generated by the method of the present invention is concentrated in a narrow interval near zero, presenting a symmetric sharp peak shape; while the noise distribution of the traditional method shows a bimodal diffusion characteristic. This distribution difference directly reflects the effect of the quantum decoherence suppression mechanism. By means of the dynamic weight decay strategy, abnormal gradient pulses are suppressed and high-frequency noise components are filtered, so as to ensure that the parameter update steadily advances in the correct direction, significantly enhancing the robustness of the training process.

[0174] In an exemplary embodiment, a computer device is provided. The computer device can be a server or a terminal, and its internal structure diagram can be as Figure 9 shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store historical coal mine equipment data and target real-time coal mine equipment data. The input / output interface of the computer device is used for exchanging information between the processor and external devices. The communication interface of the computer device is used for communicating with an external terminal through a network connection. The computer program, when executed by the processor, implements a coal mine equipment fault diagnosis method.

[0175] Those skilled in the art can understand that Figure 9 the structure shown in

[0176] is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0177] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, which, when executed by a processor, implements the steps in the above-mentioned method embodiments.

[0178] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. 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), magnetoresistive 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 be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0179] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, 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, it should be considered as the scope described in this specification.

[0180] Specific examples are used in this article to elaborate on the principles and implementation manners of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A method for diagnosing faults in coal mine equipment, characterized in that, The coal mine equipment fault diagnosis method includes: Obtaining historical coal mine equipment data; the historical coal mine equipment data refers to the historical operation data of coal mine equipment, and each piece of the historical coal mine equipment data is labeled with a true label reflecting the actual working state of the coal mine equipment; Preprocessing the historical coal mine equipment data to obtain preprocessed historical coal mine equipment data; Inputting the preprocessed historical coal mine equipment data into a coal mine equipment fault diagnosis model, calculating the loss based on the coal mine equipment fault diagnosis result output by the coal mine equipment fault diagnosis model and the corresponding true label, optimizing the model parameters based on the loss, and obtaining a trained coal mine equipment fault diagnosis model after iterative training; wherein, the coal mine equipment fault diagnosis model is a neural network model based on a dynamic kernel principal component constraint mechanism and a quantum decoherence suppression mechanism, the dynamic kernel principal component constraint mechanism refers to dynamically adjusting the kernel function width according to the characteristics of different equipment fault types during the training process, and the quantum decoherence suppression mechanism refers to suppressing quantum noise for weight update and gradient calculation during the training process; Obtaining target real-time coal mine equipment data, and using the trained coal mine equipment fault diagnosis model to perform fault diagnosis on the target real-time coal mine equipment data to obtain the final coal mine equipment fault diagnosis result; the target real-time coal mine equipment data refers to the target operation data collected in real time for fault diagnosis of coal mine equipment.

2. The coal mine equipment fault diagnosis method according to claim 1, wherein Preprocessing the historical coal mine equipment data to obtain preprocessed historical coal mine equipment data, specifically including: Performing data cleaning on the historical coal mine equipment data to obtain the historical coal mine equipment data after data cleaning; Performing data normalization or standardization processing on the historical coal mine equipment data after data cleaning to obtain the normalized or standardized historical coal mine equipment data; Performing time-domain and frequency-domain feature extraction on the normalized or standardized historical coal mine equipment data to obtain historical coal mine equipment data features as the preprocessed historical coal mine equipment data.

3. The coal mine equipment fault diagnosis method according to claim 1, characterized in that The dynamic kernel principal component constraint mechanism specifically includes: Using a Gaussian kernel function for coal mine equipment data mapping transformation, expressed as: ； Among them, is the kernel function, representing the input coal mine equipment data point and 's projection in the high-dimensional space; is the th coal mine equipment data sample; is the th coal mine equipment data sample, representing the neighboring coal mine equipment data sample of ; Dynamically adjusting the kernel width according to the coal mine equipment data characteristics of the coal mine equipment, and the dynamic adjustment method of the kernel width is expressed as: ； Among them, is the adjustment value of the kernel function width after the -th iteration, which characterizes the adaptability of the kernel function; is the adjustment value of the kernel function width after the -th iteration, which characterizes the adaptability of the kernel function; is the square of the Euclidean distance between the input coal mine equipment data and the current mapping result, which characterizes the difference between coal mine equipment data; is the adjustment factor, which controls the adjustment speed of the kernel width; is the preset tolerance error, which controls the flexibility of dynamic adjustment; is the number of samples of the coal mine equipment data input to the neural network in the current batch.

4. The coal mine equipment fault diagnosis method according to claim 1, characterized in that The quantum decoherence suppression mechanism specifically includes: When dynamically adjusting the kernel function width, the weight of the kernel function is updated using the quantum decoherence suppression mechanism. Let the decoherence effect of quantum noise be expressed as , and its update process is expressed as: ； Among them, is the value of the nuclear weight of the -th coal mine equipment data sample at the -th iteration, representing the weight after decoherence suppression; is the learning rate, controlling the amplitude of parameter update; is the gradient of the objective function, representing the direction of kernel function adjustment; is the quantum decoherence suppression term; is the hyperparameter controlling the noise suppression intensity, representing the intensity of quantum noise suppression; Calculating the quantum decoherence suppression term using the following formula: ； Among them, is the sign function; is the parameter for controlling the suppression intensity; At the -th iteration, is the weight value of the connection between the -th neuron and the -th neuron in the neural network; is the decoherence attenuation factor; The gradient calculation in backpropagation is expressed as: ； Among them, is the gradient in backpropagation; is the loss function of the neural network, representing the error between the predicted value and the true value; is the L2 regularization term; is the weight of the neural network, characterizing the connection strength between the input features of coal mine equipment data and neurons; Under the quantum decoherence suppression mechanism, the gradient calculation is expressed as: ； Among them, is the gradient after quantum noise suppression, representing the weight update gradient adjusted by the quantum decoherence suppression mechanism; is the quantum noise model, representing the suppression effect of the noise. The update process of the coal mine equipment data features is expressed as: ； Among them, is the updated characteristic vector of coal mine equipment data, representing the characteristics of coal mine equipment data after data characteristic mapping and quantum noise suppression; is the kernel function, representing the input coal mine equipment data point and projected in the high-dimensional space; is the characteristic vector of coal mine equipment data mapped by the kernel function, representing the characteristic space of coal mine equipment data after non-linear transformation; is the corresponding quantum decoherence suppression term.

5. The coal mine equipment fault diagnosis method according to claim 1, characterized in that, Obtaining target real-time coal mine equipment data, and using the trained coal mine equipment fault diagnosis model to perform fault diagnosis on the target real-time coal mine equipment data to obtain the final coal mine equipment fault diagnosis result, specifically including: Obtaining target real-time coal mine equipment data; Preprocessing the target real-time coal mine equipment data to obtain preprocessed target real-time coal mine equipment data; Inputting the preprocessed target real-time coal mine equipment data into the trained coal mine equipment fault diagnosis model to obtain the final coal mine equipment fault diagnosis result.

6. The coal mine equipment fault diagnosis method according to claim 5, characterized in that Preprocess the target real-time coal mine equipment data to obtain the preprocessed target real-time coal mine equipment data, specifically including: Perform data cleaning on the target real-time coal mine equipment data to obtain the target real-time coal mine equipment data after data cleaning; Perform data normalization or standardization on the target real-time coal mine equipment data after data cleaning to obtain the target real-time coal mine equipment data after normalization or standardization; Extract time-domain and frequency-domain features from the target real-time coal mine equipment data after normalization or standardization to obtain the target real-time coal mine equipment data features, which are used as the preprocessed target real-time coal mine equipment data.

7. The coal mine equipment fault diagnosis method according to claim 1, wherein After the step of obtaining the trained coal mine equipment fault diagnosis model after iterative training, the coal mine equipment fault diagnosis method further includes: Utilize the model incremental training mechanism driven by perturbation perception to continuously perform model incremental training on the trained coal mine equipment fault diagnosis model to obtain the coal mine equipment fault diagnosis model after model incremental training; the coal mine equipment fault diagnosis model after model incremental training is used as the trained coal mine equipment fault diagnosis model to perform fault diagnosis on the target real-time coal mine equipment data to obtain the final coal mine equipment fault diagnosis result.

8. The coal mine equipment fault diagnosis method according to claim 7, characterized in that, Utilize the model incremental training mechanism driven by perturbation perception to continuously perform model incremental training on the trained coal mine equipment fault diagnosis model to obtain the coal mine equipment fault diagnosis model after model incremental training, specifically including: Real-time monitor the distribution change of the input coal mine equipment data features, and judge whether the distribution change of the coal mine equipment data features meets the feature distribution drift trigger condition to obtain the first judgment result; the feature distribution drift trigger condition is: ； Among them, represents the feature distribution drift trigger condition, is the Kullback-Leibler divergence; is the feature distribution of the current coal mine equipment data; is the historical distribution during model training; is the preset trigger threshold; When the first judgment result is yes, it is determined that the distribution change of the coal mine equipment data features has a feature distribution drift. At this time, trigger the local perturbation reconstruction mechanism to perform perturbation incremental update on the affected hidden layer neurons. The weight of each neuron to be updated is expressed as: ； Among them, is the original neural network connection weight; is the perturbation step size; is the perturbation gradient of the input feature, expressed as: ； Among them, is the moving average of historical features, which is used to measure the feature offset direction; is the regularization adjustment factor, which is used to control the perturbation scale; is the sign function; is the loss function of the neural network, which represents the error between the predicted value and the true value; is the th coal mine equipment data sample; Introduce a memory retention regularization term to update the objective function, and the updated objective function is expressed as: ； Among them, represents the updated objective function; is the key parameter snapshot of the original model; are the key parameters of the original neural network model; is the parameter set with the greatest impact on performance; is the retention strength factor.

9. A computer device, comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the coal mine equipment fault diagnosis method according to any one of claims 1-8.

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 coal mine equipment fault diagnosis method according to any one of claims 1-8.

Images