Coal cutter cutting unit transmission system fault diagnosis method based on multi-source vibration signals

Abstract

The invention provides a coal mining machine cutting part transmission system fault diagnosis method based on a multi-source vibration signal, and the method comprises the steps: S1, installing a plurality of sensors at a plurality of key positions of a coal mining machine rocker arm cutting part transmission system, and carrying out the collection of a multi-channel vibration signal; s2, processing the multi-channel vibration signals to obtain a data set; s3, constructing a deep learning model, and training to obtain a matched fault recognition model so as to output a corresponding preliminary diagnosis result; s4, calculating an error distance among the plurality of preliminary diagnosis results, and mapping the error distance into a trust function by using a Gaussian membership function; and S5, evidence conflict detection is carried out based on the trust function in combination with other related parameters, and fusion diagnosis is carried out by adopting a D-S evidence theory synthesis rule to obtain a diagnosis result. The coal mining machine transmission system is analyzed, the positions of the sensors are reasonably arranged, an omnibearing fault diagnosis system is constructed from the three aspects of mechanical, electrical and hydraulic, and multifunctional fault monitoring and alarming are achieved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of coal mining machine, and particularly relates to a coal mining machine cutting part transmission system fault diagnosis method based on multi-source vibration signals. BACKGROUND

[0002] Coal is the main energy material in China, and occupies a pivotal position in the national energy structure and layout. With the development of science and technology and the progress of society, the requirements for coal mining are continuously improved, and therefore, improving the efficiency and intelligent level of coal mining and building green mines have become key tasks. Compared with foreign countries, the actual annual coal mining work capacity and service life of foreign equipment can reach 2-3 times of that of domestic equipment of the same type, and the starting rate of foreign fully mechanized mining equipment can reach more than 95%, but the starting rate of domestic equipment is only 35%. The main reason is that the fully mechanized mining equipment in coal mines in China has a high failure rate and poor reliability. Therefore, enhancing the information sensing ability, fault pre-diagnosis ability and overall reliability of the equipment has become one of the problems to be solved in the development of large mining mechanical and electrical equipment in China.

[0003] The coal mining machine is the core equipment of the fully mechanized working face in coal mines, and mainly undertakes the tasks of cutting, loading and unloading and transporting the coal seam in the working face. The rocker arm cutting part transmission system is the most important component of the coal mining machine, responsible for the tasks of cutting the coal seam and transmission, and has a relatively complex structure, including electrical, oil and mechanical parts. During the cutting process of the rocker arm of the coal mining machine, due to the impact load and long transmission chain of the gear box, and due to the extremely harsh environment in the coal mine, the narrow space, the wet and dusty working face and other influences, the shaft, gear and bearing parts in the gear box often fail, causing the shutdown of the fully mechanized working face. The distinctive coal mine environment and the complex structure of the rocker arm cutting part transmission system of the coal mining machine make the fault diagnosis technology of the rocker arm cutting part transmission system of the coal mining machine different from the fault diagnosis technology of ordinary mechanical equipment.

[0004] The shearer rocker arm cutting part transmission system is a complex system integrating machine, electricity and liquid, and different types of failures often occur due to the poor working environment. The commonly used monitoring method is to monitor the electrical signal, temperature and vibration signal of the shearer rocker arm, and a single sensor is usually used to detect certain information, and then a fault warning is performed. In order to eliminate the defects of one-sidedness and insufficiency of the information obtained by a single sensor, multi-sensor information fusion uses a reasonable and effective fusion algorithm to perform mathematical analysis on the data information collected by multiple sensors through computer technology, so as to eliminate redundant data and obtain more effective object information. Compared with single sensor sensing, multi-sensor information fusion has the advantages of high detection accuracy, wide sensing dimension, processing information in a short time, being suitable for various application environments, low information acquisition cost and good system fault tolerance. Common multi-sensor information fusion methods include evidence theory (D-S) method, clustering analysis, neural network, expert system and the like. The D-S evidence theory is an extension of the Bayesian theory, and can solve the uncertainty distribution problem of the reasoning model based on human; the clustering analysis uses a set of heuristic algorithms in biological science and social science to divide them into several natural groups or sets, and then associates them with the type of the expected object; the input vector of the neural network can realize nonlinear transformation, and the output of the vector can also achieve effective classification of neural network data, so that the artificial neural network can be used to comprehensively describe the data of multiple sensors. Although this method is similar to the clustering analysis method in actual application, the effect of this method is more significant in the presence of noise; the expert system has the characteristics of relying on the expression of the main knowledge, so it has great flexibility and can be expressed by numbers, symbols and reasoning characteristics. As can be seen from the above, according to the characteristics of the shearer rocker arm cutting part transmission system, the data collected by multiple sensors are analyzed and fused, and a multi-sensor fusion monitoring system is constructed, which has high value for improving the fault diagnosis accuracy, expanding the fault diagnosis dimension and improving the fault diagnosis efficiency. SUMMARY

[0005] The present application aims at the deficiencies of the prior art, and provides a shearer cutting part transmission system fault diagnosis method based on multi-source vibration signals. Through in-depth research on key technologies such as multi-source sensor feature extraction, data enhancement, deep information mining and fault diagnosis system construction of the shearer rocker arm cutting part transmission system sensor signal, the running state of the shearer rocker arm cutting part transmission system can be accurately evaluated in real time.

[0006] The application provides a coal cutter cutting part transmission system fault diagnosis method based on multi-source vibration signals, comprising the following steps: S1, a plurality of sensors are installed at a plurality of key positions of a coal cutter rocker arm cutting part transmission system to collect multi-channel vibration signals; S2, the collected multi-channel vibration signals are processed to obtain a data set; S3, a deep learning model is constructed, and a multi-channel matched fault recognition model is trained based on the data set to output corresponding preliminary diagnosis results; S4, error distances between a plurality of the preliminary diagnosis results are calculated, and the error distances are mapped into trust functions by using a Gaussian membership function; S5, evidence conflict detection is performed based on the trust functions and in combination with other related parameters, and fusion diagnosis is performed by using a D-S evidence theory synthesis rule to obtain a diagnosis result.

[0007] Preferably, in the step S1, a plurality of sensors with different collection frequency ranges are installed at each position to collect multi-channel vibration signals.

[0008] Preferably, the step S2 comprises: S21, sample pretreatment is performed on the collected vibration signals, and one-dimensional data sets are expanded by overlapping sampling or irregular sampling; S22, the one-dimensional data sets are converted into two-dimensional data sets to highlight vibration signal characteristics.

[0009] Preferably, the step S2 further comprises: S23, the converted two-dimensional data sets are expanded again by generating data sets by using a generative adversarial network to increase training samples of the data sets.

[0010] Preferably, the step S2 further comprises: S24, signal super-resolution processing is simultaneously performed on the two-dimensional data sets to highlight fault characteristics.

[0011] Preferably, the step S3 comprises: S31, fault feature extraction and classifier training are performed based on training samples of the data sets; S32, a deep learning model is constructed and trained and tested to obtain a trained fault recognition model.

[0012] Preferably, in the step S32, the deep learning fault recognition model is optimized based on at least residual connection, attention mechanism and LSTM optimization algorithm.

[0013] Preferably, the process of the step S5 comprises: the trust functions and other related parameters are taken as a plurality of evidence bodies to perform conflict degree threshold judgment, if the conflict is low, fusion diagnosis is directly performed by using a D-S evidence theory synthesis rule, if the conflict is high, the evidence bodies are corrected, and fusion diagnosis is performed according to the D-S evidence theory synthesis rule.

[0014] Preferably, the modification processing of the evidence bodies comprises the following steps: calculating the similarity between each evidence body; calculating the support degree of each evidence body and performing normalization processing; calculating the trust degree of each evidence body and modifying the corresponding evidence body.

[0015] Preferably, in the step S5, the other related parameters at least include current, voltage and temperature data of the electrical part of the cutting unit transmission system of the coal mining machine and moisture, viscosity, abrasive particle content and oil temperature data of the oil part.

[0016] On the basis of common knowledge in the art, the above-mentioned preferred conditions can be combined arbitrarily, i.e., to obtain each preferred example of the present application.

[0017] The positive progress effect of the present application is that:

[0018] According to the coal mining machine cutting unit transmission system fault diagnosis method based on multi-source vibration signals, a plurality of sensors are installed at a plurality of key positions of the rocker arm cutting unit transmission system of the coal mining machine to collect multi-channel vibration signals; the collected multi-channel vibration signals are further processed to obtain a data set; then, a deep learning model is constructed and a multi-channel matching fault recognition model is trained based on the data set to output a corresponding preliminary diagnosis result; then, the error distance between a plurality of preliminary diagnosis results is calculated, and the error distance is mapped to a trust function by using a Gaussian membership function; finally, based on the trust function and in combination with other related parameters, evidence conflict detection is performed, and fusion diagnosis is performed by using the D-S evidence theory synthesis rule to obtain a diagnosis result.

[0019] The present application aims at the problems of the traditional monitoring method of the coal mining machine transmission system, such as insufficient monitoring content, insufficient fault type monitoring and low accuracy, etc., by analyzing the coal mining machine transmission system, reasonably arranging the sensor positions according to the working characteristics of different key components, and using the D-S evidence theory to fuse and diagnose the multi-channel signals, realizing multi-functional fault monitoring and alarm. Specifically, by multi-source sensor feature extraction, data enhancement, deep information mining and fault diagnosis system construction on the sensor signals of the rocker arm cutting unit transmission system of the coal mining machine, on the basis of the diagnosis results of the multi-channel vibration signals, other related parameters such as current, voltage, temperature data of the electrical part and moisture, viscosity, abrasive particle content and oil temperature data of the oil part are comprehensively considered, so as to ensure that the running state of the rocker arm cutting unit transmission system of the coal mining machine (especially the transmission system fault) is accurately evaluated in real time.

[0020] The coal mining machine cutting part transmission system fault diagnosis method based on multi-source vibration signals of the present application, on the basis of existing mine equipment fault diagnosis research, aiming at the rocker arm cutting part transmission system of the coal mining machine, studies the formation mechanism of various fault damages of typical components, meets the development needs of the coal mining machine in the coal mine, improves the information perception ability and fault pre-diagnosis ability of the coal mining machine, lays a theoretical and experimental foundation for accurate diagnosis, reliable improvement and efficient application in the field of the rocker arm cutting part transmission system fault of the coal mining machine, and also provides reference for the research of multi-source information perception, data enhancement and fault diagnosis system construction of such machine, electric and hydraulic integrated large mine equipment. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 is a flowchart of the coal mining machine cutting part transmission system fault diagnosis method based on multi-source vibration signals in the embodiments of the present application.

[0022] Figure 2 is a block diagram of the coal mining machine cutting part transmission system fault diagnosis method based on multi-source vibration signals in the embodiments of the present application.

[0023] Figure 3 is a multi-channel signal processing process schematic diagram in the embodiments of the present application.

[0024] Figure 4 is a deep learning-based fault diagnosis algorithm process schematic diagram in the embodiments of the present application. DETAILED DESCRIPTION

[0025] The present application will be further described below in conjunction with the embodiments shown in the accompanying drawings.

[0026] <EMBODIMENT>

[0027] As shown in the accompanying Figure 1 , the present embodiment discloses a coal mining machine cutting part transmission system fault diagnosis method based on multi-source vibration signals, comprising steps S1 to S5. Specifically,

[0028] Step S1, a plurality of sensors are installed at a plurality of key positions of the rocker arm cutting part transmission system of the coal mining machine to collect multi-channel vibration signals.

[0029] In this step, due to the complex structure of the rocker arm cutting part transmission system of the coal mining machine, and also in order to overcome the defects that the information obtained by a single sensor is insufficient in information quantity, resulting in insufficient comprehensiveness and accuracy of diagnosis. Therefore, as Figure 2As shown, a plurality of different types of sensors need to be arranged to meet the comprehensive monitoring. Specifically, according to the structural characteristics of the shearer rocker arm cutting part transmission system, the sensors are arranged from the aspects of electricity, oil and machinery. The real-time monitoring is adopted for the electricity and oil parts, and the vibration signal analysis is mainly performed for the machinery part.

[0030] As a specific embodiment, in step S1, a plurality of sensors with different acquisition frequency ranges are installed at each key position of the shearer rocker arm cutting part transmission system, each sensor can acquire corresponding vibration signals, and thus multi-channel vibration signals, i.e. vibration signals of many different frequencies, can be acquired. For example, Figure 2 The vibration signals acquired by the plurality of sensors are shown in channel 1 to channel n.

[0031] Step S2, processing the acquired multi-channel vibration signals to obtain a data set.

[0032] The traditional vibration fault signal analysis method is to manually extract fault features and then use a classifier to identify faults. In the field of deep learning fault diagnosis, it is not necessary to manually extract fault features, but a large amount of data is needed for model training. Therefore,

[0033] First, for the problem of redundant data in the data information collected by the plurality of sensors, invalid data needs to be removed to obtain more effective object information. The plurality of sensor information needs to be considered comprehensively to obtain effective and certain fault information.

[0034] Secondly, in the application of deep learning theory, for the fault diagnosis of the shearer cutting part transmission system, there are problems such as less fault data samples and not obvious fault features, which affect the diagnosis accuracy. Therefore, the multi-channel vibration signals collected need to be processed from multiple angles to obtain a data set that is richer and has more obvious fault features.

[0035] As a specific embodiment, as shown in Figure 3 The data set can be expanded in step S2. Specifically, the signal processing in step S2 can include steps S21 and S22.

[0036] That is,

[0037] Step S21, sample pre-processing is performed on the acquired vibration signals, and the one-dimensional data set is expanded by overlapping sampling or irregular sampling.

[0038] Step S22, converting the one-dimensional data set into a two-dimensional data set to highlight the vibration signal features.

[0039] Based on steps S21 and S22, as shown inFigure 4 As shown, the data set is expanded by using the method of overlapping sampling and irregular sampling in one-dimensional angle; in addition, the one-dimensional signal sample after expansion is converted into two-dimensional, and the fault features of the highlight signal are realized.

[0040] In another embodiment, as Figure 3 shown, the data processing of step S2 can also include step S23. That is,

[0041] Step S23, the converted two-dimensional data set is expanded again by generating a data set through a generative adversarial network, so as to increase the training samples of the data set.

[0042] Based on step S23, as Figure 4 shown, the data set is further enriched by using a generative adversarial network to generate a data set in two-dimensional angle.

[0043] In another embodiment, as Figure 3 shown, the data processing of step S2 can also include step S23. That is,

[0044] Step S24, the two-dimensional data set is simultaneously subjected to signal super-resolution processing to highlight the fault features.

[0045] Based on step S24, as Figure 4 shown, by performing super-resolution processing on the two-dimensional signal, the clarity of the image color points is improved, the interference of the background factors is reduced, and the like, thereby realizing the highlighting of the fault features.

[0046] Step S3, a deep learning model is constructed, and a multi-channel matched fault recognition model is trained based on the data set to output the corresponding preliminary diagnosis result.

[0047] Step S3 includes steps S31 and S32. Specifically,

[0048] Step S31, fault feature extraction and classifier training are performed based on the training samples of the data set.

[0049] Step S32, a deep learning model is constructed and trained and tested, thereby obtaining the trained fault recognition model.

[0050] In steps S31 and S32, the conventional fault diagnosis method generally needs to perform feature extraction on the signal, which requires the staff to have certain experience, and the deep learning can realize end-to-end fault recognition, thereby not needing manual signal feature extraction. In addition, combined with the vibration signal characteristics of the key components of the rocker arm transmission system of the coal mining machine, a suitable deep learning algorithm is selected, a deep learning model is built, and after training and testing, a trained fault diagnosis model is obtained, which can realize the preliminary diagnosis of the corresponding vibration signal fault.

[0051] In another embodiment, during the construction of the deep learning model, the model is optimized by residual connection and the like to reduce the gradient disappearance phenomenon in the training process. The features extracted by the convolutional neural network are further extracted by using the LSTM algorithm and the attention mechanism. Moreover, the neurons are optimized by the Dropout layer, the training data of each batch is randomly selected, and the Softmax is used for fault classification. In addition, for the fault images converted in step S2, the model is trained and recognized by optimization, so as to complete the construction of the vibration signal fault diagnosis model of the rocker arm cutting part transmission system of the coal mining machine, improve the training speed of the model, and reduce the generalization error of the model.

[0052] Step S4, calculate the error distance between a plurality of preliminary diagnosis results, and map the error distance to a trust function by using a Gaussian membership function.

[0053] Step S5, based on the trust function and in combination with other related parameters, evidence conflict detection is performed, and fusion diagnosis is performed by using the D-S evidence theory synthesis rule to obtain a diagnosis result.

[0054] Specifically, as shown in Figure 2 The process of step S5 includes: taking the trust function and other related parameters as a plurality of evidence bodies to judge the conflict degree threshold value, if the conflict is low, the D-S evidence theory synthesis rule is directly used for fusion diagnosis, if the conflict is high, the evidence bodies are modified, and then the D-S evidence theory synthesis rule is used for fusion diagnosis.

[0055] The modification of the evidence bodies includes the following steps: calculating the similarity between the evidence bodies; calculating the support degree of each evidence body and performing normalization processing; calculating the trust degree of each evidence body and modifying the corresponding evidence body.

[0056] In another embodiment, the other related parameters in step S5 at least include current, voltage and temperature data of the electrical part of the cutting part transmission system of the coal mining machine and moisture, viscosity, abrasive particle content and oil temperature data of the oil part, which are obtained by real-time acquisition through corresponding sensors.

[0057] In summary, the coal mining machine cutting part transmission system fault diagnosis method based on multi-source vibration signals according to the embodiment analyzes the characteristics of the coal mining machine transmission system from the aspects of electricity, oil and machinery, reasonably arranges the sensor positions, and comprehensively monitors a plurality of fault types.

[0058] The multiple sensors are installed at multiple key positions of the cutting part transmission system of the shearer rocker arm to collect multiple-channel vibration signals; the collected multiple-channel vibration signals are further processed to obtain a data set; then, a deep learning model is constructed and a multi-channel matching fault recognition model is trained based on the data set to output corresponding preliminary diagnosis results; then, error distances between the multiple preliminary diagnosis results are calculated, and the error distances are mapped into trust functions by using a Gaussian membership function; finally, evidence conflict detection is performed based on the trust functions and in combination with other related parameters, and fusion diagnosis is performed by using a D-S evidence theory synthesis rule to obtain a diagnosis result.

[0059] Specifically, the multi-channel vibration signals are collected, a deep learning algorithm model is used to recognize faults of each channel signal, error distances are calculated, and the error distances are mapped into trust functions by using a Gaussian membership function; then, evidence conflict detection is performed, and fusion diagnosis is performed by using a D-S evidence theory synthesis rule; finally, a comprehensive fault diagnosis system of the cutting part transmission system of the shearer rocker arm is constructed from the aspects of electromechanical and hydraulic.

[0060] The diagnosis method of the embodiment is aimed at the problems of the traditional monitoring method of the shearer transmission system, such as insufficient monitoring content, insufficient fault type, and low accuracy, analyzes the shearer transmission system, reasonably arranges the sensor positions according to the working characteristics of different key components, and performs fusion diagnosis on the multi-channel signals by using a D-S evidence theory to realize multifunctional fault monitoring and alarm. Specifically, multi-source sensor feature extraction, data enhancement, deep information mining, and fault diagnosis system construction are performed on the sensor signals of the shearer rocker arm cutting part transmission system, other related parameters such as current, voltage, and temperature data of the electrical part and moisture, annual, abrasive content, and oil temperature data of the oil part are comprehensively considered based on the diagnosis results of the multi-channel vibration signals, and the running state of the shearer rocker arm cutting part transmission system (especially the transmission system fault) is ensured to be accurately evaluated in real time.

[0061] For example, for the fault diagnosis of the shearer transmission system gear box, the collected vibration signals are analyzed to realize fault diagnosis, and the accurate diagnosis of the fault can be realized by combining factors such as electricity and oil.

[0062] The coal cutter cutting part transmission system fault diagnosis method based on the multi-source vibration signal of the embodiment is based on the existing mine equipment fault diagnosis research, aims at the rocker arm cutting part transmission system of the coal cutter, researches the formation mechanism of various fault damages of the typical components, meets the development needs of the coal mine underground coal cutter, improves the information perception ability and fault pre-diagnosis ability of the coal cutter, lays the theoretical and experimental foundation for the accurate diagnosis of the rocker arm cutting part transmission system fault of the coal cutter, the effective improvement of the reliability and the efficient application in the field engineering, simultaneously provides the reference for the researches such as the multi-source information perception, data enhancement and fault diagnosis system construction of the machine, electric and liquid fusion large mine equipment.

[0063] The scope of the present application is not limited to the above-mentioned embodiments, and it is obvious that those skilled in the art can make various modifications and changes to the present application without departing from the scope and spirit of the present application. If these modifications and changes belong to the scope of the claims of the present application and its equivalent technologies, the intention of the present application also includes these modifications and changes.

Claims

1. A fault diagnosis method for the transmission system of a coal mining machine's cutting section based on multi-source vibration signals, characterized in that, Includes the following steps: S1. Install multiple sensors at multiple key locations in the transmission system of the rocker arm cutting section of the coal mining machine to collect multi-channel vibration signals; S2. Process the acquired multi-channel vibration signals to obtain a dataset; S3. Construct a deep learning model and train it based on the dataset to obtain a multi-channel matching fault identification model, so as to output the corresponding preliminary diagnosis results; S4. Calculate the error distance between multiple preliminary diagnostic results, and map the error distance to a trust function using a Gaussian membership function; S5. Based on the trust function and combined with other relevant parameters, evidence conflict detection is performed, and fusion diagnosis is performed using the DS evidence theory synthesis rules to obtain the diagnostic result.

2. The method according to claim 1, characterized in that, In step S1, multiple sensors with different acquisition frequency ranges are installed at each location to acquire multi-channel vibration signals.

3. The method according to claim 1, characterized in that, Step S2 includes: S21. Perform sample preprocessing on the collected vibration signals and expand the one-dimensional dataset through overlapping sampling or irregular sampling. S22. Convert the one-dimensional dataset into a two-dimensional dataset to highlight the characteristics of the vibration signal.

4. The method according to claim 3, characterized in that, Step S2 further includes: S23. The transformed two-dimensional dataset is further expanded by generating a dataset using a generative adversarial network to increase the number of training samples in the dataset.

5. The method according to claim 4, characterized in that, Step S2 further includes: S24. Simultaneously, perform signal super-resolution processing on the two-dimensional dataset to highlight fault characteristics.

6. The method according to claim 5, characterized in that, Step S3 includes: S31. Based on the training samples of the dataset, perform fault feature extraction and classifier training; S32. Construct a deep learning model and train and test it to obtain a trained fault identification model.

7. The method according to claim 6, characterized in that, In step S32, the deep learning fault identification model is optimized based at least on residual connections, attention mechanisms, and LSTM optimization algorithms.

8. The method according to claim 6, characterized in that, The process of step S5 includes: The trust function and other relevant parameters are used as multiple evidence bodies to determine the degree of conflict. If the conflict is low, the DS evidence theory synthesis rules are directly used for fusion diagnosis. If the conflict is high, the evidence bodies are modified and then fusion diagnosis is performed according to the DS evidence theory synthesis rules.

9. The method according to claim 8, characterized in that, The correction process for the evidence body includes the following steps: Calculate the similarity between the various pieces of evidence; Calculate the support level of each piece of evidence and normalize it; Calculate the confidence level of each piece of evidence and make corrections to the corresponding pieces of evidence.

10. The method according to claim 1, characterized in that, In step S5, other relevant parameters include at least the current, voltage, and temperature data of the electrical components of the transmission system of the coal mining machine's cutting section, as well as the moisture, viscosity, abrasive content, and oil temperature data of the oil components.

Images