Mining equipment fault diagnosis method based on multi-parameter change monitoring

Abstract

The invention provides a mining equipment fault diagnosis method based on multi-parameter change monitoring. The mining equipment fault diagnosis method comprises the following steps: firstly, constructing a fault diagnosis model through a neural network; digital signals of the running state of the mining equipment are obtained in real time through a plurality of sensors; the digital signals are preprocessed through a data preprocessing module; then, features capable of representing the running state of the equipment are extracted from the preprocessed digital signals through a feature extraction module based on a neural network, and then the extracted features are spliced and fused through a feature fusion module; and finally, analyzing the fused features through a fault diagnosis training model, carrying out fault diagnosis, generating fault information according to a fault diagnosis result, and sending out a fault alarm. According to the method, various parameter data of the operation state of the mining equipment are fully considered, feature extraction and feature splicing fusion are performed based on the twin neural network, the feature discrimination is effectively enhanced, and the intelligent fault diagnosis accuracy and efficiency of the mining equipment are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent diagnosis of mining equipment, and in particular to a mining equipment fault diagnosis method based on multi-parameter change monitoring. Background Art

[0002] With the rapid development of the coal industry towards intelligence and automation, the complexity of mining equipment and the harsh operating environment have put forward higher requirements for fault diagnosis technology. Therefore, fault diagnosis technology based on parameter change monitoring has come into being. This technology integrates the latest achievements in sensor technology, computer technology, information processing technology and artificial intelligence technology, and provides strong guarantees for the healthy management and safe operation of mining equipment.

[0003] Existing sensor technologies, such as fiber optic sensors and wireless sensor networks, have the characteristics of high precision, high sensitivity, and strong anti-interference ability, and can monitor the changes of various parameters of mining equipment in real time; sensors are used as the source of data collection, and their accuracy and reliability directly affect the accuracy of fault diagnosis. Computer technology provides strong support for data processing and analysis; the application of high-speed computing, big data processing, cloud computing and other technologies makes it possible to process and analyze massive equipment operation data in real time.

[0004] Existing information processing technologies include signal processing, feature extraction, pattern recognition and other aspects. By preprocessing, denoising, and extracting features from the collected data, useful information reflecting the characteristics of equipment failures can be extracted. Pattern recognition technology can automatically classify and identify equipment failures based on the extracted feature information.

[0005] The application of existing artificial intelligence technology in fault diagnosis is becoming more and more extensive; intelligent diagnosis algorithms based on machine learning can automatically learn the characteristics and patterns of equipment failures, improving the accuracy and efficiency of fault diagnosis; at the same time, the introduction of technologies such as deep learning and neural networks enables fault diagnosis systems to have stronger adaptability and generalization capabilities.

[0006] In summary, the intelligent fault diagnosis technology for mining equipment based on multi-parameter change monitoring has played an important role in improving the maintenance efficiency of mining equipment and ensuring production safety. However, the existing technology still has some shortcomings in practical applications, mainly including the following aspects:

[0007] 1) Limited data processing capabilities: Mining equipment generates a large amount of multi-parameter data during operation, including temperature, pressure, vibration, current, voltage, etc. When processing these massive amounts of data, existing technologies may face problems such as insufficient computing power and slow processing speed, which affects the real-time and accuracy of data processing.

[0008] 2) Poor adaptability of intelligent algorithms: Mining equipment is of various types and complex structures, and the fault characteristics of different equipment vary greatly. Existing intelligent algorithms may have poor adaptability when processing equipment data of different types and fault characteristics, resulting in reduced accuracy and efficiency of fault diagnosis.

[0009] 3) Low system integration: At present, some intelligent fault diagnosis systems for mining equipment are still operating independently, and the integration with other systems (such as production management systems, equipment management systems, etc.) is not high. This leads to obstacles in data sharing and collaborative operations, affecting the efficiency and accuracy of fault diagnosis.

[0010] 4) Soft fault detection is difficult: Soft faults usually involve small changes in multiple parameters, and these changes may be interrelated and affect each other. Summary of the invention

[0011] The purpose of the present invention is to provide a mining equipment fault diagnosis method based on multi-parameter change monitoring to solve the deficiencies of the above-mentioned prior art.

[0012] To achieve the above object, the present invention adopts the following technical solution:

[0013] A method for fault diagnosis of mining equipment based on multi-parameter change monitoring comprises the following steps:

[0014] S1. A general data set containing mapping relationships between the status of mining equipment and fault categories with multiple parameters is learned through a neural network to build a fault diagnosis model;

[0015] The fault diagnosis model includes a data preprocessing module, a feature extraction module based on a twin neural network, a feature fusion module and a fault diagnosis module;

[0016] S2, collect analog signals of the operating status of mining equipment with different parameters in real time through a number of sensors, and convert them into digital signal output;

[0017] S3, receiving the digital signal output by the sensor through a data preprocessing module and performing preprocessing;

[0018] S4, dividing the preprocessed digital signals into two groups, normalizing the two groups of digital signals through the data preprocessing module, and unifying them into the same dimensional space as input data of the twin neural network;

[0019] S5, extracting features that can characterize the operating status of the equipment through the feature extraction module, and then splicing and fusing the extracted features through the feature fusion module;

[0020] S6. Analyze the spliced ​​and fused features through the fault diagnosis module and make a fault diagnosis. At the same time, generate fault information and issue a fault alarm according to the fault diagnosis result.

[0021] S7. Optimize the parameters of the fault diagnosis model according to the fault diagnosis results and the training loss of the fault diagnosis model, and repeat steps S2-S6 to monitor the mining equipment in real time.

[0022] Furthermore, the sensor is arranged on the mining equipment to collect vibration, temperature, pressure or other analog signals of the mining equipment that can characterize the operating status of the mining equipment, and the sensor has a built-in analog-to-digital converter that can convert analog signals into digital signals.

[0023] Furthermore, the preprocessing of the digital signal by the data preprocessing module includes: filtering, denoising and signal amplification processing of the digital signal based on Fourier transform and wavelet transform, and dimensionality reduction processing of the digital signal based on principal component analysis method.

[0024] Furthermore, the data preprocessing module performs normalization processing on the two groups of digital signals, including: performing data cleaning on the two groups of digital signals by a binarization method, filtering noise in the data, and unifying the data into the same dimensional space by segmentation and zero-filling processing methods.

[0025] Furthermore, the feature extraction module based on the twin neural network adopts the Transformer network structure. Each network in the twin network builds two neural networks containing 6 layers of encoders and decoders, and uses SwiGLU as the activation function to extract features from the input data.

[0026] Furthermore, the fusion method of the feature fusion module is: the two sets of features extracted by the feature extraction module are initially fused by feature addition, and further fused through two convolutional layers.

[0027] Furthermore, the fault diagnosis module includes a classifier, which can input the features fused by the feature fusion module into the classifier for classification through a SoftMax function. The classification process of the SoftMax function is represented by the following formula (1):

[0028]

[0029] In formula (1), K represents the number of elements of feature z, where the size of K is equal to the number of fault categories; z i represents the value of the i-th element of the feature; j represents the output of the j-th neuron in a certain layer of the neural network, which is a natural constant; e is a natural constant; p irepresents the probability of belonging to the i-th category.

[0030] Furthermore, in the parameter optimization process of the fault diagnosis model, the model network structure is cut according to the training loss to reduce the network parameters that have a smaller impact on the training loss. The training loss is calculated by the cross entropy loss function, and the cross entropy loss function is expressed by the following formula (2):

[0031]

[0032] In formula (2), K represents the number of fault categories; y ab is the characteristic information of the operating status of mining equipment in a certain period of time, indicating the probability that sample a belongs to class b, y ab =1 means it belongs to, y ab =0 means not belonging; Indicates the predicted category.

[0033] It can be seen from the above technical solutions that the present invention fully considers the various parameter data of the operating status of mining equipment, and at the same time filters, denoises, amplifies the signal, reduces the dimension and normalizes the parameter data, and performs feature extraction and feature splicing and fusion based on the twin neural network, which effectively enhances the discriminability of the features and improves the accuracy and efficiency of intelligent fault diagnosis of mining equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 Schematic diagram of the steps of the diagnostic method of the present invention;

[0035] Figure 2 It is a schematic diagram of the process of feature extraction and fusion for characterizing the operating status of the equipment according to the present invention;

[0036] Figure 3 The figure is a flow chart of feature fusion for characterizing the operating status of a device according to the present invention. DETAILED DESCRIPTION

[0037] A preferred embodiment of the present invention is described in detail below with reference to the accompanying drawings.

[0038] like Figure 1 The method for intelligent fault diagnosis of mining equipment based on multi-parameter change monitoring shown includes the following steps:

[0039] S1. A general data set containing the mapping relationship between the status of mining equipment and fault categories with multiple parameters is learned through a neural network to build a fault diagnosis model.

[0040] Specifically, the general data set described in this preferred embodiment can be obtained from a general data set library, and the fault diagnosis model includes a data preprocessing module, a feature extraction module based on a twin neural network, a feature fusion module and a fault diagnosis module.

[0041] S2. Use a number of sensors to collect analog signals of the operating status of mining equipment with different parameters in real time and convert them into digital signal output.

[0042] In this preferred embodiment, a variety of sensors are installed on the mining equipment, such as vibration sensors, temperature sensors, pressure sensors, etc., to collect vibration, temperature, pressure or other analog signals that can characterize the operating status of the mining equipment. Specifically, the sensor is built with an analog-to-digital converter, and the analog-to-digital converter (ADC) is a device that converts a continuous signal in analog form into a discrete signal in digital form. Because most modern digital systems and algorithms cannot directly process the analog signals generated by the sensor, these analog signals can be converted into digital signals through the analog-to-digital converter, so that the sensor data can be read, processed and analyzed by a computer, a microprocessor or other digital device.

[0043] S3. Receive the digital signal output by the sensor through a data preprocessing module and perform preprocessing.

[0044] Specifically, the preprocessing of the digital signal by the data preprocessing module includes: filtering, denoising and signal amplification processing of the digital signal based on Fourier transform and wavelet transform, and dimensionality reduction processing of the digital signal based on the principal component analysis method.

[0045] S4. Divide the preprocessed digital signals into two groups, normalize the two groups of digital signals through the data preprocessing module, and unify them into the same dimensional space as input data of the twin neural network.

[0046] Specifically, the data preprocessing module performs normalization processing on the two groups of digital signals, including: performing data cleaning on the two groups of digital signals by binarization, filtering the noise in the data, and unifying the data into the same dimensional space by segmentation and zero filling processing. By first filtering, denoising, signal amplification, dimensionality reduction and normalization of the digital signal, and then extracting features through a neural network, the feature dimension is unified, and the feature extraction capability of the model is effectively enhanced, and the data processing speed is improved.

[0047] The Siamese neural network is a coupled architecture based on two artificial neural networks. The Siamese neural network takes two samples as input and outputs their representation embedded in a high-dimensional space to compare the similarity of the two samples; the Siamese neural network usually has a deep structure and can be composed of a convolutional neural network, a recurrent neural network, etc.

[0048] S5. The feature extraction module based on the twin neural network extracts the features that can characterize the operating status of the equipment respectively, and then the extracted features are spliced ​​and fused through the feature fusion module.

[0049] The feature extraction module based on the twin neural network described in this preferred embodiment adopts the Transformer network structure. Each network in the twin network builds two neural networks containing 6 layers of encoders and decoders, and uses SwiGLU as the activation function for feature extraction of input data.

[0050] The specific fusion method of the feature fusion module described in this preferred embodiment is: the two sets of features extracted by the feature extraction module are initially fused by feature addition, and further fused by two convolution layers. By analyzing and fusing the collected multiple types of data, different types of data are effectively used to directly connect for fault diagnosis, so as to more accurately diagnose equipment faults.

[0051] Specifically, Figure 2 As shown, taking electrical parameters, temperature parameters, amplitude and vibration frequency, four digital signals that can characterize the operating status of mining equipment as examples, the electrical parameters and temperature parameters are used as a set of input data for the twin neural network, and the amplitude and vibration frequency are used as another set of input data for the twin neural network. After normalization, they are input into the twin neural network for feature extraction.

[0052] Furthermore, if Figure 3As shown, the features extracted by the feature extraction module from the two sets of input data are recorded as feature one and feature two respectively, and the feature fusion module splices and fuses the features extracted by the feature extraction module, specifically including the following steps: first, feature one and feature two are spliced, and the splicing method can be a simple connection, or a complex operation such as merging or superposition, and feature one and feature two are unified into the same feature space through a 3×3 convolution; then, two different downsampling methods, average downsampling and maximum sampling, and convolutions with receptive fields of 3×3 and 7×7 are used in turn to extract features, and feature three and feature four are obtained; then, the extracted features three and feature four are spliced, and features three and feature four are unified into the same feature space through a 3×3 convolution. Through the above feature splicing and fusion processing, the discriminability of the features can be effectively enhanced, thereby improving the accuracy of the fault diagnosis model.

[0053] S6. The fault diagnosis module analyzes the spliced ​​and fused features and makes a fault diagnosis. At the same time, the fault information is generated according to the fault diagnosis result and a fault alarm is issued.

[0054] The basic principle of the present invention is to use a neural network to analyze the extracted feature vectors. At the same time, the neural network learns the mapping relationship between the normal state and the fault state of mining equipment through training, and uses the input feature vectors to perform fault diagnosis, determine whether the equipment has a fault and its type, and then output fault information according to the diagnosis results, including fault location, fault type, severity, etc.

[0055] The fault diagnosis module described in this preferred embodiment includes a classifier, which can input the features fused by the feature fusion module into the classifier for classification through the SoftMax function. The classification process of the SoftMax function is expressed by the following formula (1):

[0056]

[0057] In formula (1), K represents the number of elements of feature z, where the size of K is equal to the number of fault categories; z i represents the value of the i-th element of the feature; j represents the output of the j-th neuron in a certain layer of the neural network, which is a natural constant; e is a natural constant; p i represents the probability of belonging to the i-th category.

[0058] S7. Optimize the parameters of the fault diagnosis model according to the fault diagnosis results and the training loss of the fault diagnosis model, and repeat steps S2-S6 to monitor the mining equipment in real time.

[0059] Specifically, in the parameter optimization process of the fault diagnosis model, the model network structure is cut according to the training loss to reduce the network parameters that have a smaller impact on the training loss. The training loss is calculated by the cross entropy loss function, and the cross entropy loss function is expressed by the following formula (2):

[0060]

[0061] In formula (2), K represents the number of fault categories; y ab is the characteristic information of the operating status of mining equipment in a certain period of time, indicating the probability that sample a belongs to class b, y ab =1 means it belongs to, y ab =0 means not belonging; Indicates the predicted category.

[0062] The cross entropy loss function is a commonly used loss function in classification problems. It can measure the difference between the predicted probability distribution and the true distribution. It is used to optimize parameters in model training. The smaller the cross entropy value, the better the model prediction effect. It is often used in conjunction with the SoftMax function.

[0063] In this preferred embodiment, the cross entropy loss function is used to train 300 generations in the process of training and optimizing the fault diagnosis model, wherein the learning rate of the first 200 generations is 10 -4 , the learning rate for the next 100 generations is 10 -5 .

[0064] The above-described embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A mining equipment fault diagnosis method based on multi-parameter change monitoring, characterized in that: The following steps are involved: S1. A general data set containing mapping relationships between the status of mining equipment and fault categories with multiple parameters is learned through a neural network to build a fault diagnosis model; The fault diagnosis model includes a data preprocessing module, a feature extraction module based on a twin neural network, a feature fusion module and a fault diagnosis module; S2, collect analog signals of the operating status of mining equipment with different parameters in real time through a number of sensors, and convert them into digital signal output; S3, receiving the digital signal output by the sensor through a data preprocessing module and performing preprocessing; S4, dividing the preprocessed digital signals into two groups, normalizing the two groups of digital signals through the data preprocessing module, and unifying them into the same dimensional space as input data of the twin neural network; S5, extracting features that can characterize the operating status of the equipment through the feature extraction module, and then splicing and fusing the extracted features through the feature fusion module; S6. Analyze the spliced ​​and fused features through the fault diagnosis module and make a fault diagnosis. At the same time, generate fault information and issue a fault alarm according to the fault diagnosis result. S7. Optimize the parameters of the fault diagnosis model according to the fault diagnosis results and the training loss of the fault diagnosis model, and repeat steps S2-S6 to monitor the mining equipment in real time.

2. According to claim 1, a method for fault diagnosis of mining equipment based on multi-parameter change monitoring is characterized in that: The sensor is arranged on the mining equipment and is used to collect vibration, temperature, pressure or other analog signals of the mining equipment that can characterize the operating status of the mining equipment. The sensor is equipped with an analog-to-digital converter that can convert the analog signal into a digital signal.

3. The method for fault diagnosis of mining equipment based on multi-parameter change monitoring according to claim 1 is characterized in that: The data preprocessing module preprocesses the digital signal by filtering, denoising and amplifying the digital signal based on Fourier transform and wavelet transform, and performs dimensionality reduction on the digital signal based on principal component analysis.

4. A mining equipment fault diagnosis method based on multi-parameter change monitoring according to claim 3, characterized in that: The data preprocessing module performs normalization processing on the two groups of digital signals, including: performing data cleaning on the two groups of digital signals through a binarization method, filtering noise in the data, and unifying the data into the same dimensional space through segmentation and zero-filling processing methods.

5. The method for fault diagnosis of mining equipment based on multi-parameter change monitoring according to claim 1 is characterized in that: The feature extraction module based on the twin neural network adopts the Transformer network structure. Each network in the twin network builds two neural networks containing 6 layers of encoders and decoders, and uses SwiGLU as the activation function to extract features from the input data.

6. A mining equipment fault diagnosis method based on multi-parameter change monitoring according to claim 5, characterized in that: The fusion mode of the feature fusion module is: the two sets of features extracted by the feature extraction module are initially fused by feature addition, and further fused by two convolution layers.

7. A mining equipment fault diagnosis method based on multi-parameter change monitoring according to claim 6, characterized in that: The fault diagnosis module includes a classifier, which can input the features fused by the feature fusion module into the classifier for classification through a SoftMax function. The classification process of the SoftMax function is represented by the following formula (1): In formula (1), K represents the number of elements of feature z, where the size of K is equal to the number of fault categories; z i represents the value of the i-th element of the feature; j represents the output of the j-th neuron in a certain layer of the neural network, which is a natural constant; e is a natural constant; p i represents the probability of belonging to the i-th category.

8. The method for fault diagnosis of mining equipment based on multi-parameter change monitoring according to claim 1 is characterized in that: In the parameter optimization process of the fault diagnosis model, the model network structure is cut according to the training loss to reduce the network parameters that have a small impact on the training loss. The training loss is calculated by the cross entropy loss function, and the cross entropy loss function is expressed by the following formula (2): In formula (2), K represents the number of fault categories; y ab is the characteristic information of the operating status of mining equipment in a certain period of time, indicating the probability that sample a belongs to class b, y ab =1 means it belongs to, y ab =0 means not belonging; Indicates the predicted category.

Images