Fault judgment and detection system based on double-inlet and double-outlet coal mill pulverizing system

Abstract

The invention discloses a fault judgment and detection system based on a double-inlet and double-outlet coal mill pulverizing system, and relates to the technical field of coal mill fault detection systems. The data acquisition module is used for acquiring operation parameters of the pulverizing system of the double-inlet and double-outlet coal mill in real time based on each sensor; the data transmission module is used for transmitting the signals acquired by the data acquisition module to the central processing unit in a wired or wireless mode; the signal preprocessing module is used for denoising, filtering and amplifying the collected original signal; and the fault judgment module is used for carrying out deep analysis and processing on the preprocessed signals by adopting a neural network or a support vector machine. Based on wavelet transform, a neural network, a support vector machine and the like, the fault of the coal pulverizing system can be quickly and accurately judged, and the accuracy and response speed of fault detection are improved; the system adopts the intelligent analysis module to carry out fault judgment, can automatically learn and identify fault features, and reduces the demand of manual intervention.

Description

Technical Field

[0001] The invention relates to the technical field of coal mill fault detection systems, and in particular to a fault determination and detection system based on a double-inlet and double-outlet coal mill pulverizing system. Background Art

[0002] The coal pulverizing system is a system used to grind coal into fine powder. It is widely used in thermal power plants and other industrial fields that require coal powder as fuel or raw material. In actual operation, various faults often occur in the double-inlet and double-outlet coal pulverizing system, affecting the stable operation and production efficiency of the unit.

[0003] Traditional fault detection methods mainly rely on manual experience and simple sensor monitoring, which have problems such as low detection accuracy, slow response speed, misjudgment and missed judgment. Therefore, a fault judgment and detection system based on the double-inlet and double-outlet coal mill pulverizing system is urgently needed.

[0004] After searching, the application scheme of Chinese patent application number CN202010194055.6 discloses a method for identifying coal mill operation faults, including the following steps: collecting historical operating parameters and corresponding fault types of coal mill operation faults; standardizing the historical operating parameters; constructing a PCA model and a corresponding T2 statistic model for the historical operating parameters corresponding to each fault type, and determining the early warning threshold of the fault; substituting the parameters to be identified into the T2 statistic model corresponding to different faults, if the early warning threshold is exceeded, the corresponding fault type exists. The present invention also provides a coal mill operation detection system applying this method, which quickly determines the fault and cause by calculating the real-time operating parameters and comparing them with the early warning threshold. The fault identification method in the above patent has the following shortcomings: it is highly dependent on the quality and quantity of historical operating parameters; if the historical data is insufficient or there is noise, it may affect the accuracy of the PCA model and the effectiveness of the T2 statistic, and the real-time performance of the scheme is still insufficient and needs to be improved. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings existing in the prior art and to propose a fault determination and detection system based on a double-inlet and double-outlet coal mill pulverizing system.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] The fault determination and detection system based on the double-inlet and double-outlet coal mill pulverizing system includes:

[0008] Data acquisition module: Based on various sensors, real-time data is collected on the operating parameters of the double-inlet and double-outlet coal mill pulverizing system;

[0009] Data transmission module: transmits the signals collected by the data acquisition module to the central processing unit by wired or wireless means;

[0010] Signal preprocessing module: denoising, filtering and amplifying the collected original signal;

[0011] Fault judgment module: Use neural network or support vector machine to deeply analyze and process the pre-processed signal to achieve accurate identification and judgment of unknown data;

[0012] Display module: Based on the display interface, the running status, fault judgment results and early warning information of the milling system are displayed in real time;

[0013] Data storage and backup module: responsible for storing the original data collected by the data acquisition module and the analysis results of the fault judgment module.

[0014] Preferably, the fault determination module performs fault determination, specifically comprising the following steps:

[0015] S1: Feature extraction: Use wavelet transform to perform frequency decomposition on the signal and extract the feature vector reflecting the operating status of the coal mill;

[0016] S2: Intelligent analysis: Use neural networks, support vector machines or decision trees to learn and classify the extracted feature vectors; by training a large number of historical data samples, establish a mapping relationship between fault features and fault types, thereby achieving accurate identification and judgment of unknown data;

[0017] S3: Fault warning: According to the fault judgment result, early warning information is issued in time to prompt the operating personnel to take corresponding measures to deal with it.

[0018] Preferably: the feature extraction specifically includes:

[0019] S11: Wavelet transform: Use wavelet transform technology to perform frequency decomposition on the pre-processed signal and decompose the signal into different frequency bands to highlight the mutation points and abnormal patterns in the signal;

[0020] S12: Feature vector extraction: extract the feature vector reflecting the operating status of the milling system from the decomposed signal.

[0021] Preferably: the intelligent analysis specifically includes:

[0022] S21: Data preparation: Collect historical data samples, including milling system operating parameters under normal and fault conditions;

[0023] S22: Model training: Use the collected data samples to train the intelligent algorithm model and establish a mapping relationship between fault characteristics and fault types;

[0024] S23: Model evaluation and testing: Use the validation set or test set to evaluate and test the trained model to verify the classification performance and accuracy of the model;

[0025] S24: Real-time judgment: The feature vector collected in real time is input into the trained intelligent algorithm model, and the model automatically outputs the fault judgment result according to the input feature vector.

[0026] Preferably: the fault warning specifically includes:

[0027] S31: Setting of warning threshold: According to the actual situation of the milling system and historical experience data, the warning threshold of different fault types is set; when the characteristic parameter exceeds the preset threshold, the system will automatically trigger the warning mechanism;

[0028] S32: Early warning information release: When the early warning mechanism is triggered, the system immediately releases early warning information to the operator through sound and light alarms and SMS notifications;

[0029] S33: Emergency response: After receiving the warning information, the operating personnel will handle it according to the predetermined emergency plan.

[0030] Preferably: the wavelet transform is specifically:

[0031] First, a mother wavelet is selected, and then a series of wavelet basis functions are generated through scaling and translation operations;

[0032] Next, the signal to be processed is convolved with these wavelet basis functions to obtain the coefficients in each frequency band.

[0033] Preferably, the wavelet transform is specifically implemented as follows:

[0034] The wavelet transform technology is used to perform multi-scale analysis on the pre-processed signal, extract the energy distribution and singular point characteristics in different frequency bands, and capture the mutation information in the signal. The specific formula is as follows:

[0035]

[0036] in, represents the wavelet coefficients, is the original signal, is the wavelet basis function, is the scale factor, is the translation factor.

[0037] Preferably, when the intelligent analysis uses a support vector machine as the core model of the fault determination module:

[0038] For the linearly separable case:

[0039] There are two types of data sample sets ,in represents the category label, ; Based on linearly separable SVM, find an optimal hyperplane , so that the data points of different categories are located on both sides of the hyperplane, and the interval between the two categories is maximized; the optimization problem can be expressed as:

[0040]

[0041]

[0042] in, is the normal vector of the hyperplane, is the offset, is a regularization term used to control the maximization of the classification interval;

[0043] For nonlinear separable cases: Introducing kernel function Map the data to a high-dimensional space to make the data linearly separable; the optimization problem becomes:

[0044]

[0045]

[0046] in, is the Lagrange multiplier.

[0047] Preferably, when the intelligent analysis uses a neural network as the core model of the fault determination module, it specifically includes:

[0048] Feedforward neural network: It consists of multiple layers of neurons, each layer of neurons is connected by weights; suppose the input layer has nodes, and the hidden layer has nodes, and the output layer has nodes; for the Layer of neurons , its output is expressed as:

[0049]

[0050] in, It is from Layer Neuron to Layer The weight of the neuron, is the bias term, is the activation function;

[0051] Backpropagation algorithm: used to train neural networks and update weights by calculating the gradient of the loss function for each weight; the details are as follows:

[0052]

[0053] Where N is the number of samples, is the true value, is the predicted value; the gradient is calculated and the weights are updated using the chain rule:

[0054]

[0055] in, is the learning rate.

[0056] Preferably, when the intelligent analysis uses a decision tree as the core model of the fault determination module, it specifically includes:

[0057] Information gain: used to select the optimal split attribute; assuming the data set According to the characteristics The values ​​of can be divided into Subset , then the information gain is defined as:

[0058]

[0059] in, It is a dataset The entropy of represents the uncertainty of the data; Is a subset The entropy of .

[0060] The beneficial effects of the present invention are:

[0061] 1. The present invention is based on wavelet transform, neural network, support vector machine, etc., which can realize rapid and accurate judgment of milling system faults, and improve the accuracy and response speed of fault detection.

[0062] 2. The system of the present invention adopts an intelligent analysis module for fault determination, which can automatically learn and identify fault characteristics, reduce the need for manual intervention, and reduce operation and maintenance costs and labor intensity.

[0063] 3. According to the fault judgment result, the system of the present invention can issue early warning information in time to prompt the operating personnel to take corresponding measures to deal with it, which effectively avoids the further expansion of the accident and enhances the safe and stable operation capability of the milling system. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is a fault detection processing flow chart of the fault judgment detection system based on the double-inlet and double-outlet coal mill pulverizing system proposed by the present invention. DETAILED DESCRIPTION

[0065] The technical solution of the present invention is further described in detail below in conjunction with specific implementation methods.

[0066] Embodiment 1:

[0067] The fault determination and detection system based on the double-inlet and double-outlet coal mill pulverizing system includes:

[0068] Data acquisition module: Based on various sensors, real-time data is collected on the operating parameters of the double-inlet and double-outlet coal mill pulverizing system, including but not limited to the inlet and outlet temperature, pressure, flow, current, vibration and other signals of the coal mill;

[0069] Data transmission module: transmits the signals collected by the data acquisition module to the central processing unit by wired or wireless means;

[0070] Signal preprocessing module: performs denoising, filtering, amplification and other processing on the collected original signals to improve signal quality;

[0071] Fault judgment module: Use neural network or support vector machine to deeply analyze and process the pre-processed signal to achieve accurate identification and judgment of unknown data;

[0072] Display module: Based on the display interface, the running status, fault judgment results and early warning information of the milling system are displayed in real time;

[0073] Data storage and backup module: responsible for storing the original data collected by the data acquisition module and the analysis results of the fault judgment module.

[0074] The fault detection and processing flow is as follows:

[0075] Each sensor collects the operating parameters of the double-inlet and double-outlet coal pulverizing system in real time;

[0076] The signals collected by the data acquisition module are transmitted to the central processing unit by wired or wireless means;

[0077] Perform denoising, filtering, amplification and other processing on the collected original signals to improve signal quality;

[0078] Conduct in-depth analysis and processing of pre-processed signals to achieve accurate identification and judgment of unknown data;

[0079] When the early warning mechanism is triggered, the system immediately issues early warning information to the operator through sound and light alarms, SMS notifications, etc.

[0080] After receiving the early warning information, the operating personnel will immediately handle it according to the predetermined emergency plan.

[0081] The fault determination module performs fault determination, specifically comprising the following steps:

[0082] S1: Feature extraction: Use wavelet transform to perform frequency decomposition on the signal and extract the feature vector reflecting the operating status of the coal mill;

[0083] S2: Intelligent analysis: Use neural networks, support vector machines or decision trees to learn and classify the extracted feature vectors; by training a large number of historical data samples, establish a mapping relationship between fault features and fault types, thereby achieving accurate identification and judgment of unknown data;

[0084] S3: Fault warning: According to the fault judgment result, early warning information is issued in time to prompt the operating personnel to take corresponding measures to deal with it.

[0085] The feature extraction specifically includes:

[0086] S11: Wavelet transform: Use wavelet transform technology to perform frequency decomposition on the pre-processed signal and decompose the signal into different frequency bands to highlight the mutation points and abnormal patterns in the signal;

[0087] S12: Feature vector extraction: extract the feature vector reflecting the operating status of the milling system from the decomposed signal.

[0088] The intelligent analysis specifically includes:

[0089] S21: Data preparation: Collect historical data samples, including milling system operating parameters under normal and fault conditions;

[0090] S22: Model training: Use the collected data samples to train the intelligent algorithm model and establish the mapping relationship between fault characteristics and fault types. During the training process, the generalization ability and classification accuracy of the model are optimized by continuously adjusting the model parameters.

[0091] S23: Model evaluation and testing: Use the validation set or test set to evaluate and test the trained model to verify the classification performance and accuracy of the model;

[0092] S24: Real-time judgment: The feature vector collected in real time is input into the trained intelligent algorithm model, and the model automatically outputs the fault judgment result according to the input feature vector.

[0093] The fault warning specifically includes:

[0094] S31: Warning threshold setting: According to the actual situation of the milling system and historical experience data, set the warning thresholds for different fault types; when a characteristic parameter exceeds the preset threshold, the system will automatically trigger the warning mechanism;

[0095] S32: Early warning information release: When the early warning mechanism is triggered, the system immediately releases early warning information to the operator through sound and light alarms, SMS notifications, etc. The early warning information includes the fault type, possible location, severity, and recommended treatment measures;

[0096] S33: Emergency response: After receiving the warning information, the operating personnel will immediately handle it according to the predetermined emergency plan.

[0097] The wavelet transform is specifically:

[0098] First, a mother wavelet is selected, and then a series of wavelet basis functions are generated through scaling and translation operations;

[0099] Next, the signal to be processed is convolved with these wavelet basis functions to obtain the coefficients in each frequency band.

[0100] Wherein, the wavelet transform is specifically implemented as follows:

[0101] The wavelet transform technology is used to perform multi-scale analysis on the pre-processed signal, extract the energy distribution, singular points and other features in different frequency bands, and capture the mutation information in the signal. The specific formula is as follows:

[0102]

[0103] in, represents the wavelet coefficients, is the original signal, is the wavelet basis function, is the scale factor, is the translation factor.

[0104] Wherein, when the intelligent analysis adopts the support vector machine as the core model of the fault judgment module:

[0105] For the linearly separable case:

[0106] There are two types of data sample sets ,in represents the category label, ; Based on linearly separable SVM, find an optimal hyperplane , so that the data points of different categories are located on both sides of the hyperplane, and the interval between the two categories is maximized; the optimization problem can be expressed as:

[0107]

[0108]

[0109] in, is the normal vector of the hyperplane, is the offset, is a regularization term used to control the maximization of the classification interval;

[0110] For nonlinear separable cases: Introducing kernel function Map the data to a high-dimensional space to make the data linearly separable; the optimization problem becomes:

[0111]

[0112]

[0113] in, is the Lagrange multiplier.

[0114] Wherein, the intelligent analysis, when using a neural network as the core model of the fault determination module, specifically includes:

[0115] Feedforward neural network: It consists of multiple layers of neurons, each layer of neurons is connected by weights; suppose the input layer has nodes, and the hidden layer has nodes, and the output layer has nodes; for the Layer of neurons , its output is expressed as:

[0116]

[0117] in, It is from Layer Neuron to Layer The weight of the neuron, is the bias term, is the activation function;

[0118] Backpropagation algorithm: used to train neural networks and update weights by calculating the gradient of the loss function for each weight; the details are as follows:

[0119]

[0120] Where N is the number of samples, is the true value, is the predicted value; the gradient is calculated and the weights are updated using the chain rule:

[0121]

[0122] in, is the learning rate.

[0123] Wherein, the intelligent analysis, when using a decision tree as the core model of the fault determination module, specifically includes:

[0124] Information gain: used to select the optimal split attribute; assuming the data set According to the characteristics The values ​​of can be divided into Subset , then the information gain is defined as:

[0125]

[0126] in, It is a dataset The entropy of represents the uncertainty of the data; Is a subset The entropy of .

[0127] Based on wavelet transform, neural network, support vector machine, etc., it can realize fast and accurate judgment of milling system faults, improve the accuracy of fault detection and response speed; the system adopts intelligent analysis module for fault judgment, which can automatically learn and identify fault characteristics, reduce the need for manual intervention, and reduce operation and maintenance costs and labor intensity.

[0128] Based on the fault judgment results, the system can issue early warning information in a timely manner to prompt the operating personnel to take corresponding measures to deal with it, effectively avoiding the further expansion of the accident and enhancing the safe and stable operation capabilities of the milling system.

[0129] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A fault determination and detection system based on a double-inlet and double-outlet coal mill pulverizing system, characterized in that: include: Data acquisition module: Based on various sensors, real-time data is collected on the operating parameters of the double-inlet and double-outlet coal mill pulverizing system; Data transmission module: transmits the signals collected by the data acquisition module to the central processing unit by wired or wireless means; Signal preprocessing module: denoising, filtering and amplifying the collected original signal; Fault judgment module: Use neural network or support vector machine to deeply analyze and process the pre-processed signal to achieve accurate identification and judgment of unknown data; Display module: Based on the display interface, the running status, fault judgment results and early warning information of the milling system are displayed in real time; Data storage and backup module: responsible for storing the original data collected by the data acquisition module and the analysis results of the fault judgment module.

2. The fault determination and detection system based on the double-inlet and double-outlet coal mill pulverizing system according to claim 1 is characterized in that: The fault determination module performs fault determination, specifically comprising the following steps: S1: Feature extraction: Use wavelet transform to perform frequency decomposition on the signal and extract the feature vector reflecting the operating status of the coal mill; S2: Intelligent analysis: Use neural networks, support vector machines or decision trees to learn and classify the extracted feature vectors; by training a large number of historical data samples, establish a mapping relationship between fault features and fault types, thereby achieving accurate identification and judgment of unknown data; S3: Fault warning: According to the fault judgment result, early warning information is issued in time to prompt the operating personnel to take corresponding measures to deal with it.

3. The fault determination and detection system based on the double-inlet and double-outlet coal mill pulverizing system according to claim 2 is characterized in that: The feature extraction specifically includes: S11: Wavelet transform: Use wavelet transform technology to perform frequency decomposition on the pre-processed signal and decompose the signal into different frequency bands to highlight the mutation points and abnormal patterns in the signal; S12: Feature vector extraction: extract the feature vector reflecting the operating status of the milling system from the decomposed signal.

4. The fault determination and detection system based on the double-inlet and double-outlet coal mill pulverizing system according to claim 2 is characterized in that: The intelligent analysis specifically includes: S21: Data preparation: Collect historical data samples, including milling system operating parameters under normal and fault conditions; S22: Model training: Use the collected data samples to train the intelligent algorithm model and establish a mapping relationship between fault characteristics and fault types; S23: Model evaluation and testing: Use the validation set or test set to evaluate and test the trained model to verify the classification performance and accuracy of the model; S24: Real-time judgment: The feature vector collected in real time is input into the trained intelligent algorithm model, and the model automatically outputs the fault judgment result according to the input feature vector.

5. The fault determination and detection system based on the double-inlet and double-outlet coal mill pulverizing system according to claim 2 is characterized in that: The fault warning specifically includes: S31: Setting of warning threshold: According to the actual situation of the milling system and historical experience data, the warning threshold of different fault types is set; when the characteristic parameter exceeds the preset threshold, the system will automatically trigger the warning mechanism; S32: Early warning information release: When the early warning mechanism is triggered, the system immediately releases early warning information to the operator through sound and light alarms and SMS notifications; S33: Emergency response: After receiving the warning information, the operating personnel will handle it according to the predetermined emergency plan.

6. The fault determination and detection system based on the double-inlet and double-outlet coal mill pulverizing system according to claim 3 is characterized in that: The wavelet transform is specifically: First, a mother wavelet is selected, and then a series of wavelet basis functions are generated through scaling and translation operations; Next, the signal to be processed is convolved with these wavelet basis functions to obtain the coefficients in each frequency band.

7. The fault determination and detection system based on the double-inlet and double-outlet coal pulverizing system according to claim 6 is characterized in that: The wavelet transform is specifically implemented as follows: The wavelet transform technology is used to perform multi-scale analysis on the pre-processed signal, extract the energy distribution and singular point characteristics in different frequency bands, and capture the mutation information in the signal. The specific formula is as follows: in, represents the wavelet coefficients, is the original signal, is the wavelet basis function, is the scale factor, is the translation factor.

8. The fault determination and detection system based on the double-inlet and double-outlet coal mill pulverizing system according to claim 2 is characterized in that: When the intelligent analysis uses the support vector machine as the core model of the fault judgment module: For the linearly separable case: There are two types of data sample sets ,in represents the category label, ; Find an optimal hyperplane based on linearly separable SVM , so that the data points of different categories are located on both sides of the hyperplane, and the interval between the two categories is maximized; the optimization problem can be expressed as: in, is the normal vector of the hyperplane, is the offset, is a regularization term used to control the maximization of the classification interval; For nonlinear separable cases: Introducing kernel function Map the data to a high-dimensional space to make the data linearly separable; the optimization problem becomes: in, is the Lagrange multiplier.

9. The fault determination and detection system based on the double-inlet and double-outlet coal mill pulverizing system according to claim 2 is characterized in that: When the intelligent analysis uses a neural network as the core model of the fault determination module, it specifically includes: Feedforward neural network: It consists of multiple layers of neurons, each layer of neurons is connected by weights; suppose the input layer has nodes, and the hidden layer has nodes, and the output layer has nodes; for the Layer of neurons , its output is expressed as: in, It is from Layer Neuron to Layer The weight of a neuron, is the bias term, is the activation function; Backpropagation algorithm: used to train neural networks and update weights by calculating the gradient of the loss function for each weight; the details are as follows: Where N is the number of samples, is the true value, is the predicted value; the gradient is calculated and the weights are updated using the chain rule: in, is the learning rate.

10. The fault determination and detection system based on the double-inlet and double-outlet coal mill pulverizing system according to claim 2 is characterized in that: The intelligent analysis, when using a decision tree as the core model of the fault determination module, specifically includes: Information gain: used to select the optimal split attribute; assuming the data set According to the characteristics The values ​​of can be divided into Subset , then the information gain is defined as: in, It is a dataset The entropy of represents the uncertainty of the data; Is a subset The entropy of .

Images