Coal rock hardness identification method and system based on LDA-Adaboost
By applying the coal rock hardness identification method based on LDA-Adaboost in the boring machine, the problem that the boring machine is difficult to accurately identify the hardness of the rock wall is solved, and the accurate and rapid identification of the hardness of the rock wall is achieved, reducing the risk of tooth cutting damage and improving the The service life of the boring machine and the coal mining efficiency.
Patent Information
- Application Number
- CN202411557171.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-04
- Publication Date
- 2025-02-11
Smart Images
Figure CN119416029A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of roadheader pick protection, and more particularly to a coal and rock hardness identification method and system based on LDA-Adaboost. Background Art
[0002] As the main fossil energy, coal has a vital impact on social production and development. At present, the roadheader has become the main equipment for coal mining, and is highly favored for its wide range of application scenarios, safety and efficiency. However, the roadheader bears a large load during work, the construction environment is harsh, and it is often difficult for operators to accurately identify the current state of cutting the rock wall. When cutting harder rock walls at a higher speed, it is easy to cause damage to the cutting teeth, which seriously affects the cutting efficiency and coal mining efficiency. Therefore, a reliable rock wall hardness identification method is needed to prevent the roadheader cutting teeth from being damaged during operation.
[0003] The environment in which the tunnel boring machine is located is complex, and the vibration signal of the cutting arm has the characteristics of nonlinearity and non-stationarity. It is necessary to find a suitable signal processing method to process the vibration signal of the cutting arm of the tunnel boring machine to extract the hardness characteristics. The existing vibration signal extraction methods are mainly divided into the following categories: the first is to use harmonic wavelet packets to decompose the vibration signal, obtain the power spectrum of each frequency band to construct the feature vector, but this method ignores the time domain information of each frequency band; the second is to introduce singular value decomposition into the feature extraction of vibration signals, and use wavelet packet decomposition to construct the time-frequency feature matrix, but the decomposed feature matrix is not very explanatory and the model generalization ability is poor; the third is to use the empirical mode decomposition method to process the signal and identify the coal-rock interface according to the spectrum of each signal component, but the extracted signal components have modal aliasing and the decomposition effect is poor.
[0004] The Linear Discriminant Analysis (LDA) algorithm is relatively simple, with less computational effort than the above methods, and improves the computational speed. It can also reduce the dimension of the original data by finding the optimal projection direction, extract more stable and effective data features, and optimize the recognition effect.
[0005] Adaptive Boosting (Adaboost) is an ensemble learning technology. It is based on multiple weak classifiers and continuously iterates learning. During the iteration process, the weight of correctly classified samples is reduced, and the weight of incorrectly classified samples is increased. More attention is paid to incorrectly classified samples, and less attention is paid to correctly classified samples. The algorithm reduces the amount of calculation, improves the recognition speed and accuracy, and reduces the interference of the nonlinearity and non-stationarity of the cutting arm vibration signal on the recognition result. As an integrated classification algorithm, this algorithm can effectively solve the classification problem, but there is still little research on its application in the identification of the cutting hardness of the tunnel boring machine. The main difficulty in the identification of the rock wall hardness of the tunnel boring machine is to effectively combine the algorithm and the identification model.
[0006] Therefore, how to propose a coal and rock hardness identification method and system based on LDA-Adaboost, organically combine linear discriminant analysis and adaptive enhancement algorithm, overcome the interference of nonlinear and non-stationary vibration signals on the identification results, and realize accurate and rapid identification of rock wall hardness is a problem that technical personnel in this field urgently need to solve. Summary of the invention
[0007] In view of this, the present invention provides a method and system for identifying coal and rock hardness based on LDA-Adaboost, which organically combines linear discriminant analysis and adaptive enhancement algorithm to overcome the interference of nonlinear and non-stationary vibration signals on the identification results, and realize accurate and rapid identification of rock wall hardness. In order to achieve the above purpose, the present invention adopts the following technical solutions:
[0008] A coal rock hardness recognition method based on LDA-Adaboost, comprising:
[0009] Collect the vibration signal of the cutting arm underground and build a vibration signal data set;
[0010] Extraction of coal and rock hardness characteristics of vibration signals based on LDA algorithm;
[0011] Determine the number of heterogeneous classifiers and screen the recognition algorithms used by the heterogeneous classifiers;
[0012] Multiple recognition models are obtained by combining multiple determined heterogeneous classifiers;
[0013] Judging multiple recognition models to obtain the optimal recognition model;
[0014] The real-time vibration signal is input into the optimal recognition model to obtain the coal and rock hardness recognition results.
[0015] Optionally, collecting the vibration signal of the cutting arm downhole and constructing the vibration signal data set includes:
[0016] The data set was constructed based on the measured vibration signals of the cutting arm in coal mines. The experimental working condition was horizontal cutting, the coal rock hardness grade was m, and the vibration signal was collected by a vibration sensor installed on the cutting arm of the roadheader with a collection frequency of n kHz.
[0017] Optionally, the extraction of coal and rock hardness features of vibration signals based on the LDA algorithm includes:
[0018] Calculate the overall mean vector of all vibration signal data in the vibration signal data set;
[0019] Calculate the intra-class scatter matrix and inter-class scatter matrix of the vibration signal;
[0020] According to the two calculated matrices, a generalized eigenvalue problem is constructed to obtain the corresponding eigenvalues and eigenvectors, and the data of the vibration signal data set is projected onto the eigenvector with the largest eigenvalue to complete the extraction of data features.
[0021] Optionally, determining the number of heterogeneous classifiers includes: using a grey wolf optimization algorithm to find the optimal number of heterogeneous classifiers with the highest accuracy target.
[0022] Optionally, the population size of the grey wolf optimization algorithm is set to o, the maximum number of iterations is set to p, the upper and lower limits of the number of model layers are set to [1, 5], and after q rounds of optimization, the optimal number of layers r of the grey wolf algorithm is determined.
[0023] Optionally, the screening of the recognition algorithm used by the heterogeneous classifier includes:
[0024] Taking the features extracted by LDA algorithm as input, different classification algorithms are used to compare the recognition accuracy, and r recognition algorithms are selected for coal and rock hardness recognition.
[0025] Optionally, the combining of the determined multiple heterogeneous classifiers to obtain multiple recognition models includes:
[0026] Determine the r algorithms to be neural network, support vector machine and decision tree;
[0027] The number of layers of the recognition model was determined to be 3, and the neural network, support vector machine and decision tree were arranged in different layers to conduct recognition effect experiments.
[0028] Optionally, the multiple recognition models include: support vector machine + decision tree + neural network, support vector machine + neural network + decision tree, decision tree + support vector machine + neural network, decision tree + neural network + support vector machine, neural network + support vector machine + decision tree and neural network + decision tree + support vector machine.
[0029] Optionally, judging the multiple recognition models to obtain the optimal recognition model includes: weighing both accuracy and time when considering the recognition effect, screening the recognition models, and judging to obtain the optimal recognition model.
[0030] Optionally, a coal rock hardness recognition system based on LDA-Adaboost includes:
[0031] Acquisition module: used to collect the vibration signal of the cutting arm underground and build a vibration signal data set;
[0032] Feature extraction module: used to extract coal and rock hardness features of vibration signals based on LDA algorithm;
[0033] Algorithm screening module: used to determine the number of heterogeneous classifiers and screen the recognition algorithms used by the heterogeneous classifiers;
[0034] Combination module: used to obtain multiple recognition models based on the combination of multiple determined heterogeneous classifiers;
[0035] Judgment module: used to judge multiple recognition models to obtain the optimal recognition model;
[0036] Identification module: used to input real-time vibration signals into the optimal identification model to obtain coal and rock hardness identification results.
[0037] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a method and system for identifying coal and rock hardness based on LDA-Adaboost, which has the following beneficial effects:
[0038] The present invention proposes a coal rock hardness identification method based on LDA-Adaboost, including: collecting the vibration signal of the cutting arm underground, constructing a vibration signal data set; extracting the coal rock hardness characteristics of the vibration signal based on the LDA algorithm; determining the number of heterogeneous classifiers, and screening the recognition algorithms used by the heterogeneous classifiers; combining multiple heterogeneous classifiers to obtain multiple recognition models; judging multiple recognition models to obtain the optimal recognition model; inputting the real-time vibration signal into the optimal recognition model to obtain the coal rock hardness recognition result. The present invention introduces the LDA algorithm to extract the coal rock hardness characteristics of the data set, and uses the gray wolf algorithm to find the optimal number of model layers with the maximum accuracy target value, overcomes the interference of the vibration signal nonlinearity and non-stationarity on the recognition result, and selects different classification algorithms to compare the recognition effects, optimizes the model arrangement mode, and can realize the accurate and rapid recognition of the rock wall hardness, and provides important parameters for realizing the intelligent and stable control of the tunnel boring machine, which can effectively reduce the coal production reduction caused by the damage of the cutting teeth, thereby improving the service life of the tunnel boring machine. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0040] Figure 1 A schematic flow chart of a method for identifying coal and rock hardness based on LDA-Adaboost provided by the present invention.
[0041] Figure 2 This is a comparison chart of the accuracy of different numbers of layers provided by the present invention.
[0042] Figure 3 A comparison chart of training time for different numbers of layers provided by the present invention.
[0043] Figure 4 The recognition result diagram provided by the present invention. DETAILED DESCRIPTION
[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0045] The embodiment of the present invention discloses a method for identifying coal and rock hardness based on LDA-Adaboost, comprising:
[0046] Collect the vibration signal of the cutting arm underground and build a vibration signal data set;
[0047] Extraction of coal and rock hardness characteristics of vibration signals based on LDA algorithm;
[0048] Determine the number of heterogeneous classifiers and screen the recognition algorithms used by the heterogeneous classifiers;
[0049] Multiple recognition models are obtained by combining multiple determined heterogeneous classifiers;
[0050] Judging multiple recognition models to obtain the optimal recognition model;
[0051] The real-time vibration signal is input into the optimal recognition model to obtain the coal and rock hardness recognition results.
[0052] In a specific implementation, a coal rock hardness identification method based on LDA-Adaboost can achieve fast and accurate rock wall hardness identification on the premise of improving the recognition effect of the original Adaboost meta-algorithm. Figure 1 shown.
[0053] Step 1: construct a vibration signal data set based on the cutting arm vibration signal collected underground;
[0054] Step 2: Use LDA algorithm to extract the coal and rock hardness characteristics of the vibration signal;
[0055] Step 3: Use the Grey Wolf Optimization Algorithm to find the optimal number of heterogeneous classifiers with the highest accuracy.
[0056] Step 4: Compare the recognition effects of different classification algorithms and select the recognition algorithm used by the heterogeneous classifier;
[0057] Step 5: Compare the recognition effects of different combinations of models, combine recognition accuracy and speed, and build the optimal recognition model;
[0058] Step 6: Input the vibration signal into the optimal recognition model to obtain the coal rock hardness recognition result.
[0059] Furthermore, in step 1, a data set is constructed based on the vibration signal of the cutting arm measured underground in a coal mine. The experimental working condition is a horizontal cutting working condition, the coal rock hardness grade is 6, 7, and 8, and the vibration signal is collected by a vibration sensor installed on the cutting arm of the roadheader, and the collection frequency is 10kHz.
[0060] Furthermore, the step 2 extracts the hardness features in the data set and classifies the data features as follows:
[0061] 1) Calculate the overall mean vector of all data in the vibration signal data, Where N is the total number of signals, x i is the signal sample point;
[0062] 2) Calculate the intra-class scatter matrix and inter-class scatter matrix of the vibration signal. The calculation formula is as follows: C i is the sample set of the i-th category, x is the vibration signal sample vector, m i is the mean vector of the i-th category, and C is the total number of categories;
[0063] 3) Based on the two calculated matrices, construct a generalized eigenvalue problem to obtain the corresponding eigenvalues and eigenvectors. The calculation formula is S i -αS Bv = 0, find the non-zero eigenvalue α and the corresponding eigenvector v and project the data of each vibration signal data sample in the data set onto the eigenvector with the largest eigenvalue. The projected vector is a = v max b. Complete the extraction of data features.
[0064] Furthermore, in step 3, the population size of the gray wolf algorithm is first set to 10, the maximum number of iterations is set to 100, and considering the model construction requirements, the upper and lower limits of the model layers are set to [1, 5]; after 100 rounds of optimization, the optimal number of layers determined by the gray wolf algorithm is [3], and the model accuracy is improved from 43.1% to 98.8%.
[0065] Furthermore, in step 4, the features extracted by the LDA algorithm are used as input, and different classification algorithms are used to compare the recognition accuracy, and three recognition algorithms are selected for coal rock hardness recognition. The three algorithms are neural network, support vector machine and decision tree. The comparison of recognition accuracy of different classification algorithms is shown in Table 1.
[0066] Table 1
[0067] Identification method Recognition accuracy / % Neural Networks 70.1 Support Vector Machine 61.7 Decision Tree 56.6 Linear Classifier 41.2 Naive Bayes 37.5
[0068] Furthermore, the arrangement of different recognition methods in step 5 has different effects on the experimental results. In order to find a suitable arrangement, the following comparative experiment is arranged. The number of layers of the recognition model is determined to be 3. The neural network, support vector machine and decision tree are arranged in different layers to conduct recognition effect experiments. The different arrangements of the classification algorithms are shown in Table 2. The recognition accuracy results are shown in Table 2. Figure 2 When considering the recognition effect, we should consider both accuracy and time. The training time of the recognition model with different arrangements is shown in Figure 3 As shown, it can be seen that when the arrangement mode is the third one, the recognition speed is the fastest, and the recognition time is 4.48s. After comprehensively considering the accuracy and time, the present invention selects the arrangement mode 2 to construct the recognition model.
[0069] Table 2
[0070]
[0071]
[0072] Furthermore, in step 6, the present invention is used to perform a rock wall hardness identification experiment, and the identification result is as follows: Figure 4 As shown, data labels 6, 7, and 8 represent the hardness levels of the rock wall under horizontal cutting conditions, and the recognition accuracy of the above three hardness levels reaches 98.8%.
[0073] In a specific implementation, a coal rock hardness identification system based on LDA-Adaboost includes:
[0074] Acquisition module: used to collect the vibration signal of the cutting arm underground and build a vibration signal data set;
[0075] Feature extraction module: used to extract coal and rock hardness features of vibration signals based on LDA algorithm;
[0076] Algorithm screening module: used to determine the number of heterogeneous classifiers and screen the recognition algorithms used by the heterogeneous classifiers;
[0077] Combination module: used to obtain multiple recognition models based on the combination of multiple determined heterogeneous classifiers;
[0078] Judgment module: used to judge multiple recognition models to obtain the optimal recognition model;
[0079] Identification module: used to input real-time vibration signals into the optimal identification model to obtain coal and rock hardness identification results.
[0080] The present invention introduces the LDA algorithm to extract the coal and rock hardness characteristics of the data set, and adopts the Grey Wolf algorithm to find the optimal number of model layers with the maximum accuracy target value, thereby overcoming the interference of the nonlinearity and non-stationarity of the vibration signal on the recognition result. At the same time, different classification algorithms are selected to compare the recognition effects and optimize the model arrangement method. The present invention can realize accurate and rapid identification of rock wall hardness, provides important parameters for realizing intelligent and stable control of the roadheader, and can effectively reduce the coal production reduction caused by pick damage, thereby increasing the service life of the roadheader.
[0081] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
[0082] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A coal rock hardness identification method based on LDA-Adaboost, characterized in that: include: Collect the vibration signal of the cutting arm underground and build a vibration signal data set; Extraction of coal and rock hardness characteristics of vibration signals based on LDA algorithm; Determine the number of heterogeneous classifiers and screen the recognition algorithms used by the heterogeneous classifiers; Multiple recognition models are obtained by combining multiple determined heterogeneous classifiers; Judging multiple recognition models to obtain the optimal recognition model; The real-time vibration signal is input into the optimal recognition model to obtain the coal and rock hardness recognition results.
2. The method for identifying coal rock hardness based on LDA-Adaboost according to claim 1, characterized in that: The collecting of the vibration signal of the cutting arm in the well and constructing the vibration signal data set comprises: The data set was constructed based on the measured vibration signals of the cutting arm in coal mines. The experimental working condition was horizontal cutting, the coal rock hardness grade was m, and the vibration signal was collected by a vibration sensor installed on the cutting arm of the roadheader with a collection frequency of n kHz.
3. The method for identifying coal rock hardness based on LDA-Adaboost according to claim 1, characterized in that: The coal and rock hardness characteristics of the vibration signal extracted based on the LDA algorithm include: Calculate the overall mean vector of all vibration signal data in the vibration signal data set; Calculate the intra-class scatter matrix and inter-class scatter matrix of the vibration signal; According to the two calculated matrices, a generalized eigenvalue problem is constructed to obtain the corresponding eigenvalues and eigenvectors, and the data of the vibration signal data set is projected onto the eigenvector with the largest eigenvalue to complete the extraction of data features.
4. The method for identifying coal rock hardness based on LDA-Adaboost according to claim 1, characterized in that: Determining the number of heterogeneous classifiers includes: using a grey wolf optimization algorithm to find the best number of heterogeneous classifiers with the highest accuracy target.
5. The method for identifying coal rock hardness based on LDA-Adaboost according to claim 4 is characterized in that: The population size of the gray wolf optimization algorithm is set to o, the maximum number of iterations is set to p, the upper and lower limits of the model layer number are set to [1, 5], and after q rounds of optimization, the optimal number of layers r of the gray wolf algorithm is determined.
6. The method for identifying coal rock hardness based on LDA-Adaboost according to claim 1, characterized in that: The screening of the recognition algorithm used by the heterogeneous classifier includes: Taking the features extracted by LDA algorithm as input, different classification algorithms are used to compare the recognition accuracy, and r recognition algorithms are selected for coal and rock hardness recognition.
7. The method for identifying coal rock hardness based on LDA-Adaboost according to claim 1, characterized in that: The method of combining the determined multiple heterogeneous classifiers to obtain multiple recognition models includes: Determine the r algorithms to be neural network, support vector machine and decision tree; The number of layers of the recognition model was determined to be 3, and the neural network, support vector machine and decision tree were arranged in different layers to conduct recognition effect experiments.
8. The method for identifying coal rock hardness based on LDA-Adaboost according to claim 7, characterized in that: The multiple recognition models include: support vector machine + decision tree + neural network, support vector machine + neural network + decision tree, decision tree + support vector machine + neural network, decision tree + neural network + support vector machine, neural network + support vector machine + decision tree and neural network + decision tree + support vector machine.
9. The method for identifying coal rock hardness based on LDA-Adaboost according to claim 1, characterized in that: The determining of the optimal recognition model from the multiple recognition models includes: simultaneously weighing the two factors of accuracy and time, screening the recognition models, and determining the optimal recognition model.
10. A coal rock hardness identification system based on LDA-Adaboost, characterized in that: include: Acquisition module: used to collect the vibration signal of the cutting arm underground and build a vibration signal data set; Feature extraction module: used to extract coal and rock hardness features of vibration signals based on LDA algorithm; Algorithm screening module: used to determine the number of heterogeneous classifiers and screen the recognition algorithms used by the heterogeneous classifiers; Combination module: used to obtain multiple recognition models based on the combination of multiple determined heterogeneous classifiers; Judgment module: used to judge multiple recognition models to obtain the optimal recognition model; Identification module: used to input real-time vibration signals into the optimal identification model to obtain coal and rock hardness identification results.