A hard rock non-explosive continuous mining method and intelligent equipment

Abstract

This invention provides a method and intelligent equipment for continuous non-explosive mining of hard rock, applicable to the field of mining machinery and equipment. This method includes sensor deployment, data processing, automatic adjustment of operating parameters, and self-learning. By installing vibration sensors, torque sensors, speed sensors, and temperature sensors on the mining equipment and collecting time series data, the system uses feature extraction based on wavelet packet singular value decomposition, feature dimensionality reduction using the Relief algorithm, and an improved XGBoost model to identify rock formation hardness, achieving highly accurate hardness detection. Based on the identification results, the control center automatically adjusts operating parameters to ensure a high degree of match between the operating parameters and the rock formation hardness, achieving an efficient and stable mining process, reducing energy consumption while minimizing wear caused by vibration. By recording and analyzing data from actual operations, the XGBoost model is continuously optimized to enhance the system's adaptability and accuracy. By introducing intelligent, data-driven technologies, the system achieves adaptive capabilities in mining environments with complex and changing rock formation hardness.

Description

Technical Field

[0001] The present invention relates to the field of mining machinery and equipment, and in particular to a hard rock non-explosive continuous mining method and intelligent equipment. Background Art

[0002] Hard rock mining is a critical component of mining engineering and is widely used in coal, metal, and building stone mining. With the increasing depletion of mineral resources and the continuous increase in mining depth, hard rock mining technology faces higher safety, efficiency, and environmental protection requirements. Traditional hard rock mining methods mainly include explosive blasting and non-explosive mechanical mining. Explosive blasting is highly efficient, adaptable, and relatively low-cost, capable of rapidly breaking up large rocks. However, it also has disadvantages such as threats to operators and equipment during blasting, large amounts of dust, noise, vibration, and damage to the surrounding ecological environment.

[0003] While non-explosive mechanical mining methods offer advantages over explosive blasting in terms of safety and environmental friendliness, they still have significant drawbacks. Compared to explosive blasting, mechanical mining offers lower crushing efficiency and slower mining speeds. Mechanical equipment struggles to adapt to varying rock hardness, resulting in unstable mining efficiency. Furthermore, continuous mechanical operation consumes significant energy, and the high-intensity wear and tear of mechanical equipment leads to high maintenance and replacement costs.

[0004] In view of this, it is particularly important to develop a method and equipment for hard rock non-explosive continuous mining operations that can enhance mining efficiency, reduce energy consumption, and reduce equipment wear. Summary of the Invention

[0005] The purpose of this invention is to solve the problems of low mining efficiency, instability and high energy consumption of traditional non-blasting mechanical mining methods. A hard rock non-blasting continuous mining method and intelligent equipment are proposed, which overcomes the shortcomings of existing technologies in automation, monitoring, adaptability and cost control, and provides an efficient, safe and environmentally friendly mining solution, reduces unplanned downtime, improves overall operational efficiency, and can be widely used in the field of mining machinery and equipment.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A hard rock non-explosive continuous mining method specifically comprises the following steps:

[0008] S101, sensor deployment;

[0009] The sensor deployment includes installing a vibration sensor in the cutting section's pick drum, installing a torque sensor on the rotating shaft driving the pick drum, installing a first speed sensor on the transfer case's input shaft, installing a second speed sensor on the transfer case's cutting section output shaft, and installing a temperature sensor on the torque converter housing; the vibration sensor, torque sensor, first speed sensor, second speed sensor, and temperature sensor each collect time series data of a vibration variable V1, a torque variable V2, a speed variable V3, a speed variable V4, and a temperature variable V5 at their respective set frequencies, and transmit the data to a control center;

[0010] S102, data processing;

[0011] The data processing includes, after the control center receives the time series data, performing feature extraction based on wavelet packet singular value decomposition and feature data dimensionality reduction based on the Relief algorithm on the time series data of the vibration variable V1, the torque variable V2, the speed variable V3, the speed variable V4, and the temperature variable V5, and then generating high-order feature cross terms according to the feature vector of the vibration variable V1, the feature vector of the torque variable V2, the feature vector of the speed variable V3, the feature vector of the speed variable V4, and the feature vector of the temperature variable V5, using the high-order feature cross terms as the feature vector of the new variable V6, and standardizing the feature vectors of the vibration variable V1, the torque variable V2, the speed variable V3, the speed variable V4, the temperature variable V5, and the new variable V6 as overall input parameters, inputting them into the improved XGBoost model, and then performing rock formation hardness identification based on the improved XGBoost;

[0012] The feature extraction based on wavelet packet singular value decomposition includes using the wavelet packet algorithm to decompose the time series data layer by layer, decomposing the time series data into n nodes according to a preset number of decomposition layers, and obtaining the wavelet packet coefficient of each node; reconstructing the n nodes to obtain the reconstructed time series data of each node, and combining the reconstructed time series data of the n nodes to obtain a time-frequency matrix;

[0013] The feature data dimensionality reduction based on the Relief algorithm includes performing singular value decomposition on the time-frequency matrix, using the Relief algorithm to measure the classification ability of each order singular value through the correlation statistics corresponding to each order singular value, selecting the first k order singular values ​​of the correlation statistics from large to small, and using the k order singular values ​​as the feature vectors of the time series data;

[0014] The rock formation hardness identification based on the improved XGBoost includes using a Bayesian optimization algorithm combined with a Gaussian process proxy model to optimize the key hyperparameters of XGBoost. Then, based on the approximate optimal parameters found by Bayesian optimization, a grid search is performed to fine-tune the key hyperparameters. The improved XGBoost model obtains the probability distribution of each hardness level based on the overall input parameters, and selects the hardness level with the highest probability as the rock formation hardness identification result.

[0015] S103, automatic adjustment of working parameters;

[0016] The automatic adjustment of working parameters includes an improved XGBoost model trained and deployed locally on the intelligent equipment for hard rock non-explosive continuous mining. The parameter adjustment decision module built into the control center automatically adjusts the engine power, transmission gear, and main arm downforce to the most suitable level for the current rock hardness based on the hardness identification results.

[0017] S104, self-learning;

[0018] The self-learning includes recording the time series data and hardness recognition results collected each time. When the hardness recognition result does not match the actual situation and causes mining to stagnate, the engine power, transmission gear and main arm downforce are manually adjusted. When mining continues after the adjustment, the adjusted engine power, transmission gear and main arm downforce are recorded. Based on the adjusted engine power, transmission gear and main arm downforce, the correct hardness recognition result is evaluated, and then the sliding window technology is used to regularly introduce new data and eliminate old data to fine-tune and improve the XGBoost model.

[0019] As a preferred technical solution of the present invention, in step S102, the high-order feature cross term is obtained by weighted summation of the feature vector of any variable and the feature vector of another variable.

[0020]

[0021] Where, - high-order feature interactions, - the eigenvector of any variable, - eigenvector of another variable, t-weight.

[0022] As a preferred technical solution of the present invention, in step S103, the engine power, gearbox gear and main arm downforce that best match the current rock hardness grade are determined through field tests, and the best matching evaluation reference indicators are mining speed, energy consumption and equipment vibration.

[0023] As a preferred technical solution of the present invention, in step S102, the standardization process adopts Z-score standardization.

[0024]

[0025] Where f′ i,j -jth variable V j The i-th value in the eigenvector of j -jth variable V j The mean of the eigenvector, σ j -jth variable V j The standard deviation of the eigenvectors.

[0026] As a preferred technical solution of the present invention, in step S102, the hardness grade adopts fifteen grades of I, II, III, IIIa, IV, IVa, V, Va, VI, VIa, VII, VIIa, VIII, IX, and X classified according to the Proctor coefficient.

[0027] As a preferred technical solution of the present invention, in step S104, the method for judging the mining stagnation is that the output power of the engine remains basically unchanged, the oil temperature of the torque converter rises abnormally, and the value of the output shaft speed sensor of the cutting part of the transfer case decreases rapidly or even decreases to 0.

[0028] An intelligent equipment for continuous non-explosive mining of hard rock, used in a method for continuous non-explosive mining of hard rock, comprising a frame arranged on a walking crawler assembly, a main arm arranged at the front end of the frame, a cutting part located at the front end of the main arm, an engine, a gearbox, an electronic control system and a transmission system arranged at the rear end of the frame, wherein a hydraulic torque converter is provided between the engine and the gearbox, a transfer case is provided between the cutting part and the gearbox, power is transmitted between the transfer case and the gearbox via an input shaft of the transfer case, power is transmitted between the transfer case and the cutting part via an output shaft of the transfer case and a driven shaft of the cutting part, the cutting part comprises two pick rollers, a gearbox provided at The rotating shaft at the center axis of the two cutter drums and the transmission cutting chain arranged between the two cutter drums, one end of the transmission cutting chain is connected to the rotating shaft, and the other end is connected to the driven shaft of the cutting part. The cutter drum and the transmission cutting chain are both welded with cutters, driven by the engine, and the cutter drum is driven to rotate by the transmission cutting chain after passing through the transmission system. It is characterized in that: a torque sensor is provided on the rotating shaft of the driving cutter drum, a speed sensor is provided on the input shaft of the transfer case and the output shaft of the cutting part of the transfer case, and a temperature sensor is provided on the torque converter housing. An electronic control board with an integrated control center and a server with a locally deployed improved XG Boost model are added to the electronic control system.

[0029] As a preferred technical solution of the present invention, the control center receives the time series data collected by the sensor and performs feature extraction and dimensionality reduction processing on the feature data, and then inputs it into the improved XGBoost model to identify the hardness of the rock formation. Then, based on the hardness identification results, instructions are sent to adjust the engine power, transmission gear and main arm downforce.

[0030] As a preferred technical solution of the present invention, the control center records all collected time series data, hardness identification results, and adjusted engine power, transmission gear and main arm downforce in a storage device. When the storage device capacity is insufficient, a capacity warning is automatically issued to remind maintenance personnel to replace the storage device in time.

[0031] The beneficial effects of the present invention are as follows: by installing vibration sensors, torque sensors, speed sensors and temperature sensors on mining equipment and collecting time series data, rock hardness identification is performed using feature extraction based on wavelet packet singular value decomposition, feature dimensionality reduction of the Relief algorithm and an improved XGBoost model, thereby achieving high-accuracy hardness detection; the control center automatically adjusts the engine power, gearbox gear and main arm downforce according to the identification results to ensure that the mining parameters are highly matched with the rock hardness, thereby achieving high efficiency and stability in the mining process, reducing energy consumption while reducing wear caused by vibration; in addition, through a self-learning mechanism, data in actual operations are recorded and analyzed, the XGBoost model is continuously optimized, and the adaptability and accuracy of the system are improved; by introducing intelligent and data-driven technical means, the adaptive ability of the rock hardness mining environment with complex and changeable formations is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a flow chart of the hard rock non-explosive continuous mining method of the present invention;

[0033] Figure 2 This is a schematic structural diagram of the intelligent equipment for continuous non-explosive mining of hard rock according to the present invention;

[0034] Figure 3 This is a schematic diagram of the engine and transmission system of the intelligent equipment for continuous non-explosive mining of hard rock according to the present invention;

[0035] Reference numerals in the figure: 1- walking track assembly, 2- frame, 4- main arm, 5- cutting part, 51- cutting gear drum, 52- rotating shaft, 53- transmission cutting chain, 54- driven shaft, 6- transfer case, 61- input shaft, 6- cutting part output shaft, 7- gearbox, 8- torque converter, 9- engine. DETAILED DESCRIPTION

[0036] The following describes in detail specific embodiments of the present invention in conjunction with the accompanying drawings. It should be understood that the specific embodiments provided herein are intended only to illustrate and explain the present invention and are not intended to limit the present invention. It should be noted that many specific details are set forth in the following description to facilitate a full understanding of the present invention. However, the present invention may also have other embodiments and variations thereof. Therefore, the scope of protection of the present invention is not limited by the specific embodiments disclosed below.

[0037] In Example 1, the non-explosive continuous mining method for hard rock is as follows:

[0038] A hard rock non-explosive continuous mining method specifically comprises the following steps:

[0039] S101, sensor deployment;

[0040] A vibration sensor is installed in the pick drum 51 of the cutting section 5, a torque sensor is installed on the rotating shaft 52 that drives the pick drum 51, speed sensors are installed on the input shaft 61 of the transfer case 6 and the cutting section output shaft 62 of the transfer case 6, and a temperature sensor is installed on the housing of the torque converter 8. The vibration sensor, torque sensor, first speed sensor, second speed sensor, and temperature sensor respectively collect time series data of the vibration variable V1, torque variable V2, speed variable V3, speed variable V4, and temperature variable V5 at their respective set frequencies and transmit the data to the control center.

[0041] S102, data processing;

[0042] After receiving the time series data, the control center performs feature extraction based on wavelet packet singular value decomposition and feature data dimensionality reduction based on the Relief algorithm on the time series data of vibration variable V1, torque variable V2, speed variable V3, speed variable V4, and temperature variable V5;

[0043] The feature extraction based on wavelet packet singular value decomposition includes using a wavelet packet algorithm to decompose the time series data layer by layer, decomposing the time series data into n nodes according to a preset number of decomposition layers, and obtaining the wavelet packet coefficient of each node; reconstructing the n nodes to obtain the reconstructed time series data of each node, and combining the reconstructed time series data of the n nodes to obtain a time-frequency matrix;

[0044] The dimensionality reduction of characteristic data based on the Relief algorithm includes performing singular value decomposition on the time-frequency matrix, using the Relief algorithm to measure the classification ability of each order singular value through the correlation statistics corresponding to each order singular value, selecting the first k order singular values ​​of the correlation statistics from large to small, and using the k order singular values ​​as the characteristic vectors of the time series data; taking the time-frequency matrix of the vibration variable V1 as an example, given m samples for determining the hardness grade of the rock formation, recorded as {(x1,y1),(x2,y2),…,(x i,y i ),…,(x m ,y m )}, where y i is the corresponding sample x i The rock hardness grade, for each sample x i , and its r-order singular value is recorded as Find the sample with the same hardness grade as the sample rock layer and the closest distance as the negative neighbor x i,- , find the sample with different rock hardness grade and closest distance as the positive neighbor x i,+ , the correlation statistic is δ, and the correlation statistic component δ of its j-order singular value j for,

[0045]

[0046] Where, - Sample x i and x i,+ The distance on the j-th order singular value, - Sample x i and x i,- The distance on the j-th order singular value; if the sample x i Its positive neighbor x i,+ The distance on the j-th order singular value is greater than the sample x i and its negative neighbor x i,- The distance on the j-order singular value indicates that the j-order singular value is beneficial to distinguish different hardness grades of rock formations. The relevant statistical component δ of the j-order singular value j It will increase as the distance difference increases, and the j-order singular value has a stronger ability to distinguish different hardness grades of rock formations;

[0047] Then, high-order feature cross terms are generated based on the eigenvector of vibration variable V1, the eigenvector of torque variable V2, the eigenvector of speed variable V3, the eigenvector of speed variable V4, and the eigenvector of temperature variable V5. The high-order feature cross terms are used as the eigenvectors of new variable V6 to further enhance the model's ability to capture complex feature relationships. The high-order feature cross terms are obtained by weighted summation of the eigenvector of any variable and the eigenvector of another variable.

[0048]

[0049] Where, - high-order feature interactions, - the eigenvector of any variable, - eigenvector of another variable, t-weight;

[0050] The hardness grade is classified into fifteen grades, namely, I, II, III, IIIa, IV, IVa, V, Va, VI, VIa, VII, VIIa, VIII, IX, and X, based on the Proctor coefficient. Only the intelligent continuous mining in the relatively hard rock formations of grades II to Va is considered. Through field tests, the time series data of the vibration variable V1 of the pick drum 51, the time series data of the torque variable V2 of the pick drum shaft 52, the time series data of the speed variable V3 of the input shaft 61 of the transfer case 6, the time series data of the speed variable V4 of the output shaft 62 of the cutting part of the transfer case 6, and the time series data of the torque converter 8 under different hardness grades of grades II to Va are obtained. The time series data of the temperature variable V5 of the housing, the power of the engine 9, the gear position of the gearbox 7, the downward force of the main arm 4, the mining speed, the energy consumption and the equipment vibration data are obtained by the vibration sensor installed in the cab of the vehicle frame 2. The Pearson correlation coefficients of the vibration data of the pick drum 51, the torque data of the pick drum shaft 52, the speed data of the input shaft 61 of the transfer case 6, the speed data of the cutting part output shaft 62 of the transfer case 6, the temperature data of the torque converter 8 housing and the hardness grade are calculated and recorded as ρ1, ρ2, ρ3, ρ4 and ρ5 respectively. The importance ratio of the time series data of each variable to the hardness identification result is:

[0051]

[0052] Where, I i - the i-th variable V i The importance of the time series data to the hardness recognition result; further determine the weight t according to the importance,

[0053]

[0054] Where, I b -The importance of the time series data of any variable to the hardness identification results, I c - the importance of time series data of another variable on the hardness identification results;

[0055] The eigenvector of the vibration variable V1, the eigenvector of the torque variable V2, the eigenvector of the speed variable V3, the eigenvector of the speed variable V4, the eigenvector of the temperature variable V5 and the eigenvector of the new variable V6 are standardized and used as the overall input parameter X=[f′ i,j ], input into the improved XGBoost model, and then perform rock hardness identification based on the improved XGBoost model; the standardization process adopts Z-score standardization,

[0056]

[0057] Where f′ i,j -jth variable V jThe i-th value in the eigenvector of j -jth variable V j The mean of the eigenvector, σ j -jth variable V j The standard deviation of the eigenvectors;

[0058] The rock formation hardness identification based on improved XGBoost includes using the Bayesian optimization algorithm combined with the Gaussian process proxy model to optimize the key hyperparameters of XGBoost. XGBoost has multiple hyperparameters, and common key hyperparameters include the learning rate η, the number of trees n, and the number of trees. es , maximum depth d max , minimum subsample weight and cw min , column sampling ratio cs b , regularization parameters λ, α, define the hyperparameter space according to the maximum and minimum values ​​of each key hyperparameter,

[0059] Θ={η∈(0.01,0.3),n es ∈[100,1000],d max ∈[3,10],...} (6)

[0060] Select several initial hyperparameter combinations {θ1,θ2,…,θ m} and train the XGBoost model, record the cross-validation score g(θ i ), using the current hyperparameter combination and its corresponding score g(θ i ) Construct a Gaussian process agent model and select the next hyperparameter combination θ by optimizing the EI acquisition function m+1 ,In the framework of Gaussian process agent model, EI can be specifically expressed as,

[0061] EI(θ)=(μ(θ)-g(θ + ))Φ(Z)+σ(θ)φ(Z) (7)

[0062]

[0063] Where μ(θ) is the predicted mean of the Gaussian process at θ, σ(θ) is the predicted standard deviation of the Gaussian process at θ, and g(θ + )-the best cross-validation score currently observed, Φ(Z)-the cumulative distribution function of the standard normal distribution, φ(Z)-the probability density function of the standard normal distribution; select the hyperparameter combination θ that maximizes EI m+1 ,

[0064]

[0065] In θ m+1The model is trained, the cross-validation score is recorded, the Gaussian process proxy model is updated, and the above process is repeated until the approximately optimal hyperparameter combination θ is found. * ;

[0066] Then, based on the approximate optimal parameters found by Bayesian optimization, a grid search is performed to fine-tune the key hyperparameters. Based on the results of Bayesian optimization, the fine-tuning range of each hyperparameter is set to obtain the fine-tuned hyperparameter space Θ fine , set the step size of each hyperparameter according to the requirements, evaluate each hyperparameter combination within the fine-tuning range, and select the hyperparameter combination θ that gives the best cross-validation score fine ;

[0067]

[0068] Use the hyperparameter combination θ fine-tuned by Bayesian optimization and grid search fine , train the XGBoost model,

[0069] F(X) = XGBoost(X; θ fine ) (11)

[0070] Where, F(X) is the prediction score function. The softmax function is applied to the prediction score to obtain the probability distribution of each hardness level. The hardness level with the highest probability is selected as the hardness identification result of the rock formation.

[0071] S103, automatic adjustment of working parameters;

[0072] The parameter adjustment decision module built into the control center automatically adjusts the power of the engine 8, the gear position of the gearbox 7, and the downward force of the main arm 4 to the best match for the current rock formation hardness grade based on the hardness identification result. Through field tests, the power of the engine 8, the gear position of the gearbox 7, the downward force of the main arm 4, the mining speed, energy consumption, and equipment vibration data at different hardness grades from II to Va are obtained. The particle swarm optimization algorithm is used to find the best matching parameter combination for optimizing performance indicators under specific hardness grade conditions. Under the same hardness grade conditions, the power of the engine 8, the gear position of the gearbox 7, and the downward force of the main arm 4 corresponding to fast mining speed, low energy consumption, and small equipment vibration are used as the best matching evaluation criteria. The established objective function is:

[0073] obj=β(1-S)+γE+ξA (12)

[0074] Where obj is the objective function. The smaller the value, the better the matching of the parameter combination. β, γ, and ξ are weight coefficients, reflecting the importance of each indicator. The sum of the three is 1. In this embodiment, β = 0.5, γ = 0.3, and ξ = 0.2 are taken in the order of importance of mining speed, energy consumption, and equipment vibration, respectively.

[0075] S104, self-learning;

[0076] Record the time series data and hardness recognition results collected each time. When the hardness recognition result does not match the actual situation and causes mining to stagnate, the judgment method of mining stagnation is that the output power of engine 9 remains basically unchanged, the oil temperature of torque converter 8 rises abnormally, and causes the temperature of the outer shell of torque converter 8 to rise sharply, and the output shaft speed sensor value of the cutting part of transfer case 6 decreases rapidly or even decreases to 0. Then manually adjust the power of engine 8, the gear position of gearbox 7 and the downforce of main arm 4. After adjustment, when mining continues, record the adjusted power of engine 8, gear position of gearbox 7 and downforce of main arm 4. According to the adjusted power of engine 8, gear position of gearbox 7 and downforce of main arm 4, evaluate the correct hardness recognition result, and then use sliding window technology to regularly introduce new data and eliminate old data to fine-tune and improve the XGBoost model.

[0077] Embodiment 2, a hard rock non-explosive continuous mining intelligent equipment, including a frame 2 provided on a walking crawler assembly 1, a main arm 4 provided at the front end of the frame 2, a cutting part 5 located at the front end of the main arm 4, an engine 9 provided at the rear end of the frame 2, a gearbox 7, an electronic control system and a transmission system provided on the entire frame 2, wherein a hydraulic torque converter 8 is provided between the engine 9 and the gearbox 7, a transfer case 6 is provided between the cutting part 5 and the gearbox 7, power is transmitted between the transfer case 6 and the gearbox 7 through the input shaft 61 of the transfer case, power is transmitted between the transfer case 6 and the cutting part 5 through the cutting part output shaft 62 of the transfer case 6 and the driven shaft 54 ​​of the cutting part 5, the cutting part 5 includes two pick rollers 51, a gear roller 5 provided between the two pick rollers 5 1, and a transmission cutting chain 53 provided between the two pick drums 51. One end of the transmission cutting chain 53 is connected to the rotating shaft 52, and the other end is connected to the driven shaft 54 ​​of the cutting part 5. Picks are welded on the pick drum 51 and the transmission cutting chain 53. The pick drum 51 and the transmission cutting chain 53 are driven by the engine 9 and driven by the transmission system and the transmission cutting chain 53 to drive the pick drum 51 to rotate. A torque sensor is provided on the rotating shaft 52 of the driving pick drum 51, a speed sensor is provided on the input shaft 61 of the transfer case 6 and the cutting part output shaft 62 of the transfer case 6, and a temperature sensor is provided on the casing of the torque converter 8. An electronic control board integrating a control center and a server for locally deploying an improved XGBoost model are added to the electronic control system.

[0078] In this embodiment, the control center receives the time series data collected by the sensor and performs feature extraction and dimensionality reduction processing on the feature data, and then inputs it into the improved XGBoost model to identify the hardness of the rock formation. Then, based on the hardness identification results, instructions are sent to adjust the power of the engine 9, the gear position of the gearbox 7, and the downforce of the main arm 4.

[0079] In this embodiment, the control center records all collected time series data, hardness identification results, and the adjusted power of the engine 9, the gear position of the transmission 7, and the downforce of the main arm 4 in the storage device. When the storage device capacity is insufficient, a capacity warning is automatically issued to remind maintenance personnel to replace the storage device in time.

[0080] In summary, the hard rock non-explosive continuous mining method and intelligent equipment of the present invention have the characteristics of high efficiency and stability in the mining process, reduced equipment wear, and can intelligently adapt to the mining environment with complex and changeable rock hardness.

[0081] It should be understood that the above embodiments are one or more embodiments of the present invention, and there are many other embodiments and variations thereof based on the present invention; the variations and modifications made by ordinary technicians in this industry through the present invention without making groundbreaking innovations all fall within the scope of protection of the present invention.

Claims

1. A method for continuous non-explosive mining of hard rock, characterized in that The specific steps include: S101, sensor deployment; The sensor deployment includes installing a vibration sensor in the cutting section's pick drum, installing a torque sensor on the rotating shaft driving the pick drum, installing a first speed sensor on the transfer case's input shaft, installing a second speed sensor on the transfer case's cutting section output shaft, and installing a temperature sensor on the torque converter housing; the vibration sensor, torque sensor, first speed sensor, second speed sensor, and temperature sensor each collect time series data of a vibration variable V1, a torque variable V2, a speed variable V3, a speed variable V4, and a temperature variable V5 at their respective set frequencies, and transmit the data to a control center; S102, data processing; The data processing includes, after the control center receives the time series data, performing feature extraction based on wavelet packet singular value decomposition and feature data dimensionality reduction based on the Relief algorithm on the time series data of the vibration variable V1, the torque variable V2, the speed variable V3, the speed variable V4, and the temperature variable V5, and then generating high-order feature cross terms according to the feature vector of the vibration variable V1, the feature vector of the torque variable V2, the feature vector of the speed variable V3, the feature vector of the speed variable V4, and the feature vector of the temperature variable V5, using the high-order feature cross terms as the feature vector of the new variable V6, and standardizing the feature vectors of the vibration variable V1, the torque variable V2, the speed variable V3, the speed variable V4, the temperature variable V5, and the new variable V6 as overall input parameters, inputting them into the improved XGBoost model, and then performing rock formation hardness identification based on the improved XGBoost; The feature extraction based on wavelet packet singular value decomposition includes using the wavelet packet algorithm to decompose the time series data layer by layer, decomposing the time series data into n nodes according to a preset number of decomposition layers, and obtaining the wavelet packet coefficient of each node; reconstructing the n nodes to obtain the reconstructed time series data of each node, and combining the reconstructed time series data of the n nodes to obtain a time-frequency matrix; The feature data dimensionality reduction based on the Relief algorithm includes performing singular value decomposition on the time-frequency matrix, using the Relief algorithm to measure the classification ability of each order singular value through the correlation statistics corresponding to each order singular value, selecting the first k order singular values ​​of the correlation statistics from large to small, and using the k order singular values ​​as the feature vectors of the time series data; The rock formation hardness identification based on the improved XGBoost includes using a Bayesian optimization algorithm combined with a Gaussian process proxy model to optimize the key hyperparameters of XGBoost. Then, based on the approximate optimal parameters found by Bayesian optimization, a grid search is performed to fine-tune the key hyperparameters. The improved XGBoost model obtains the probability distribution of each hardness level based on the overall input parameters, and selects the hardness level with the highest probability as the rock formation hardness identification result. S103, automatic adjustment of working parameters; The automatic adjustment of working parameters includes an improved XGBoost model trained and deployed locally on the intelligent equipment for hard rock non-explosive continuous mining. The parameter adjustment decision module built into the control center automatically adjusts the engine power, transmission gear, and main arm downforce to the most suitable level for the current rock hardness based on the hardness identification results. S104, self-learning; The self-learning includes recording the time series data and hardness recognition results collected each time. When the hardness recognition result does not match the actual situation and causes mining to stagnate, the engine power, transmission gear and main arm downforce are manually adjusted. When mining continues after the adjustment, the adjusted engine power, transmission gear and main arm downforce are recorded. Based on the adjusted engine power, transmission gear and main arm downforce, the correct hardness recognition result is evaluated, and then the sliding window technology is used to regularly introduce new data and eliminate old data to fine-tune and improve the XGBoost model.

2. The hard rock non-explosive continuous mining method according to claim 1, characterized in that: In step S102, the high-order feature cross term is obtained by weighted summation of the feature vector of any variable and the feature vector of another variable. Where, - high-order feature interactions, - the eigenvector of any variable, - eigenvector of another variable, t-weight.

3. The hard rock non-explosive continuous mining method according to claim 1, characterized in that: In step S103, the engine power, gearbox position and main arm downforce that best match the current rock hardness grade are determined through field tests, and the best matching evaluation reference indicators are mining speed, energy consumption and equipment vibration.

4. The hard rock non-explosive continuous mining method according to claim 1, characterized in that: In step S102, the normalization process uses Z-score normalization. Where f′ i,j -jth variable V j The i-th value in the eigenvector of j -jth variable V j The mean of the eigenvector, σ j -jth variable V j The standard deviation of the eigenvectors.

5. The hard rock non-explosive continuous mining method according to claim 1, characterized in that: In step S102, the hardness grade adopts fifteen grades of I, II, III, IIIa, IV, IVa, V, Va, VI, VIa, VII, VIIa, VIII, IX, and X classified according to the Proctor coefficient.

6. The hard rock non-explosive continuous mining method according to claim 1, characterized in that: In step S104, the mining stagnation is determined in the following manner: the output power of the engine remains substantially unchanged, the oil temperature of the torque converter rises abnormally, and the value of the output shaft speed sensor of the cutting part of the transfer case decreases rapidly or even decreases to 0.

7. An intelligent equipment for continuous non-explosive mining of hard rock, used in the method for continuous non-explosive mining of hard rock according to any one of claims 1 to 6, comprising a vehicle frame mounted on a walking crawler assembly, a main arm mounted at the front end of the vehicle frame, a cutting unit located at the front end of the main arm, an engine mounted at the rear end of the vehicle frame, a gearbox, an electronic control system, and a transmission system mounted on the entire vehicle frame, wherein a torque converter is provided between the engine and the gearbox, a transfer case is provided between the cutting unit and the gearbox, and power is transmitted between the transfer case and the gearbox via an input shaft of the transfer case. The power is transmitted between the transfer case and the cutting part through the cutting part output shaft of the transfer case and the driven shaft of the cutting part. The cutting part includes two pick drums, a rotating shaft provided at the center axis of the two pick drums, and a transmission cutting chain provided between the two pick drums. One end of the transmission cutting chain is connected to the rotating shaft, and the other end is connected to the driven shaft of the cutting part. The pick drum and the transmission cutting chain are both welded with picks, driven by the engine, and the pick drum is driven to rotate through the transmission cutting chain after passing through the transmission system. It is characterized in that: A torque sensor is provided on the rotating shaft of the driving cutter drum, a speed sensor is provided on the input shaft of the transfer case and the output shaft of the cutting part of the transfer case, and a temperature sensor is provided on the torque converter housing. An electronic control board with an integrated control center and a server for locally deploying the improved XGBoost model are added to the electronic control system.

8. The intelligent equipment for continuous non-explosive mining of hard rock according to claim 7, characterized in that: The control center receives the time series data collected by the sensor and performs feature extraction and dimensionality reduction processing on the feature data, and then inputs it into the improved XGBoost model to identify the hardness of the rock formation. Then, based on the hardness identification results, the control center sends instructions to adjust the engine power, transmission gear and main arm downforce.

9. The intelligent equipment for continuous non-explosive mining of hard rock according to claim 7, characterized in that: The control center records all collected time series data, hardness identification results, and adjusted engine power, transmission gear, and main arm downforce in a storage device. When the storage device capacity is insufficient, a capacity warning is automatically issued to remind maintenance personnel to replace the storage device in time.

Images