An adaptive control method for a shearer based on the recognition of coal and rock by monitoring the force on pick

Abstract

The present invention belongs to the technical field of lithology identification, and discloses a self - adaptive control method for a shearer based on coal - rock identification by monitoring the forces on pick teeth. Based on the force data of pick teeth and combined with the operating parameters of the shearer, it realizes accurate and efficient identification of coal and rock and self - adaptive control of the shearer. The self - adaptive control method for the shearer based on coal - rock identification by monitoring the forces on pick teeth obtains the original force data of pick teeth in real - time through pressure sensors arranged at the root of each pick tooth. Combining with the operating parameters of the shearer, it uses data processing and analysis algorithms to accurately and efficiently identify the coal - rock interface, and accordingly dynamically adjusts the cutting parameters of the shearer, realizing safe, efficient and high - precision coal - rock identification and self - adaptive control of the shearer, improving the intelligence of coal - mining operations, increasing the coal - mining efficiency, reducing equipment wear, and ensuring the safety and stability of mining operations.

Description

Technical Field

[0001] The present invention belongs to the technical field of lithology identification, and particularly relates to a self-adaptive control method for a shearer based on pick force monitoring for coal and rock identification. Background Art

[0002] Gangue is inevitably mixed in raw coal. During the coal mining process, when mining gangue, it will cause problems such as accelerated wear of equipment such as shearers, increased transportation costs and energy consumption, increased washing costs, and reduced coal quality.

[0003] At present, for different working conditions and coal and rock types, different coal and rock identification methods have different limitations due to different focuses of their identification technical principles. Infrared thermal imaging technology is based on the identification of temperature differences during the cutting of the shearer drum to determine the cutting object, that is, when the drum cuts the rock stratum, the temperature is higher than that when cutting the coal seam. Through this temperature comparison, it can be judged whether the object cut by the drum is a coal seam or a rock stratum. However, its workload is large and time-consuming, and it is difficult to accurately capture the temperature at the contact between the pick and the coal and rock in a complex mining environment, so it is severely limited in practical applications. Vibration monitoring technology judges whether the coal and rock interface is cut by collecting vibration signals of the shearer picks, rocker arms, rotating shafts and the fuselage. The application premise of this technology is that there are significant differences in the Prandtl coefficients of coal and rock strata. Only when the shearer cuts to the rock stratum can the coal and rock interface be detected. Vibration signals are easily attenuated during propagation, resulting in a short detection distance and increased difficulty, affecting the accuracy and practicality of identification. Image recognition technology uses visible light to obtain coal and rock images, and identifies coal and rock according to the differences in the color, texture, shape, fracture, luster, etc. of the coal and rock interface between the roof and floor. It is still in the experimental stage and has not reached the maturity of wide promotion and application, and still requires further technological breakthroughs and optimizations. The γ-ray detection technology uses a ray sensor to receive the intensity of γ-rays emitted by the coal and rock strata, so as to calculate the coal seam thickness. However, this technology is more harmful and will endanger the health of miners after long-term use.

[0004] Therefore, there is an urgent need to propose a safe, efficient and high-precision coal and rock identification and shearer self-adaptive control method. Summary of the Invention

[0005] The purpose of the present invention is to propose a self-adaptive control method for a shearer based on pick force monitoring for coal and rock identification, so as to accurately and efficiently identify coal and rock based on pick force data and in combination with the operating parameters of the shearer, and realize the self-adaptive control of the shearer.

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

[0007] A self - adaptive control method for a shearer based on the recognition of coal and rock by monitoring the force on pick teeth. A pressure sensor is set at the root of each pick tooth of the shearer to which the method is applied. The signal of the pressure sensor is connected to a data acquisition unit, the signal of the data acquisition unit is connected to a shearer controller, and the signal of the shearer controller is connected to a host computer;

[0008] The method includes the following steps:

[0009] S1. Obtain the original data of the force on the pick teeth collected by each pressure sensor and the cutting parameters of the shearer under different coal and rock properties during cutting;

[0010] S2. Pre - process the original data of the force on the pick teeth;

[0011] S3. Train a machine learning model. The machine learning model selects a random forest model for training and classification recognition of coal and rock properties;

[0012] S4. Verify the machine learning model;

[0013] S5. Dynamically adjust the cutting parameters of the shearer according to the results output by the optimized model.

[0014] Based on the above - mentioned self - adaptive control method for a shearer based on the recognition of coal and rock by monitoring the force on pick teeth, the present invention also proposes a computer device. The computer device includes a memory and one or more processors. An executable code is stored in the memory. When the processor executes the executable code, it is used to implement the above - mentioned self - adaptive control method for a shearer based on the recognition of coal and rock by monitoring the force on pick teeth.

[0015] Based on the above - mentioned self - adaptive control method for a shearer based on the recognition of coal and rock by monitoring the force on pick teeth, the present invention also proposes a computer - readable storage medium, on which a program is stored. When the program is executed by a processor, it is used to implement the above - mentioned self - adaptive control method for a shearer based on the recognition of coal and rock by monitoring the force on pick teeth.

[0016] The present invention has the following advantages:

[0017] The present invention can obtain the original data of the force on the pick teeth in real - time through the pressure sensors set at the roots of each pick tooth, combine with the operating parameters of the shearer, use data processing and analysis algorithms to accurately and efficiently identify the coal - rock interface, and accordingly dynamically adjust the cutting parameters of the shearer, realizing safe, efficient and high - precision coal - rock recognition and self - adaptive control of the shearer, improving the intelligence of coal mining operations, increasing the coal mining efficiency, reducing equipment wear, and ensuring the safety and stability of mining operations. Description of the Drawings

[0018] Figure 1It is a flowchart of the shearer adaptive control method for coal and rock identification based on pick force monitoring in the embodiments of the present invention;

[0019] Figure 2 It is a schematic diagram of coal mining operation in the embodiments of the present invention. Specific embodiments

[0020] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0021] Embodiment 1

[0022] As Figure 1 、 2 shown, in this embodiment, a shearer adaptive control method for coal and rock identification based on pick force monitoring is proposed. A pressure sensor is provided at the root of each pick of the shearer to which the method is applied. The pressure sensor is signal-connected to a data acquisition unit, and the data acquisition unit is signal-connected to the shearer controller via a wireless transmission module. The shearer controller is signal-connected to a host computer via a wireless transmission module.

[0023] The pressure sensor is a miniature pressure sensor with high sensitivity, high temperature resistance, impact resistance and can adapt to harsh underground working conditions. Its range is set according to the estimated maximum force when the pick cuts the rock to ensure that the subtle changes in the pick force can be accurately captured. At the root of each pick of the shearer, the pressure sensor is firmly installed through an embedded installation structure to ensure that the pressure sensor is in close contact with the pick and can measure the axial, normal and tangential forces received by the pick during cutting operations in real time and accurately. The embedded installation structure is made of high-strength alloy material and has a shock absorption and buffering function to protect the pressure sensor from the interference of complex vibrations and impacts underground. The data acquisition unit integrates a multi-channel A / D conversion chip to collect the data of each pick pressure sensor at high speed and synchronously. The processed data is transmitted to the shearer controller in real time through the wireless transmission module to ensure high-efficiency, stable and low-latency data transmission. The shearer controller continues to transmit the data to the host computer through the wireless transmission module. The host computer is a computer set on the ground for data storage and processing.

[0024] The method includes the following steps:

[0025] S1. Obtain the original data of the pick force collected by each pressure sensor and the cutting parameters of the shearer under different coal and rock properties during cutting.

[0026] S1 is specifically:

[0027] S11. During the operation of the shearer cutting coal and rock, the data acquisition unit real-time collects the original data of the forces on the pick monitored by each pressure sensor throughout the cutting process. Herein, the entire cutting process includes the process of cutting into the coal and rock, the process of stably cutting the coal and rock, and the process of cutting out from the coal and rock; the original data of the forces on the pick includes the axial, normal, and tangential forces on the pick; the data acquisition unit uploads the original data of the forces on the pick to the shearer controller; meanwhile, the shearer controller records the cutting parameters of the shearer, wherein the cutting parameters of the shearer include the rotational speed of the cutting drum, the cutting height, and the traction speed.

[0028] For each measurement point (i.e., the pressure sensor on each pick), continuously collect data to ensure obtaining a long enough time series data to reflect the force change of the pick during the entire cutting process, including the force characteristics during the process of cutting into the coal and rock, the process of stably cutting the coal and rock, and the process of cutting out from the coal and rock.

[0029] S12. The shearer controller uploads the cutting parameters of the shearer and the original data of the forces on the pick to the host computer, and uploads the different coal and rock properties corresponding to the shearer cutting to the host computer.

[0030] S2. Preprocess the original data of the forces on the pick.

[0031] Specifically, S2 is as follows:

[0032] S21. The host computer sets a data quality monitoring program to real-time monitor whether the transmitted data is abnormal. Such as data loss, mutation, exceeding the reasonable range, etc. are abnormal data, and mark and record the abnormal data.

[0033] S22. Denoise, smooth, and normalize the original data of the forces on the pick.

[0034] Adopt a filtering algorithm to remove the noise of the original data of the forces on the pick, and the calculation formula is:

[0035] ; Equation (1)

[0036] Wherein, is the output signal; is the input signal; is the impulse response function; t is time.

[0037] Use the moving average method to reduce data fluctuations to smooth the signal, and the definition is as follows:

[0038] ; Equation (2)

[0039] Wherein, is the moving average value; n is the window size; is the delay value of the input signal; t is time.

[0040] Normalize data with different dimensions to eliminate the influence of dimensions and convert the data into a distribution with a mean of 0 and a standard deviation of 1. The calculation formula is as follows:

[0041] ; Equation (3)

[0042] where is its value after standardization; is the feature mean; is the standard deviation.

[0043] S23. Data annotation.

[0044] Annotate the original data of the pick force after normalization according to different cutting processes. The annotated data is used as the true value sample data set for training the machine learning model.

[0045] S24. Use 85% of the true value sample data set as the training set to train the random forest model; use the remaining 15% of the true value sample data set as the test set to test the accuracy of the machine learning model and evaluate the generalization ability and performance indicators of the machine learning model.

[0046] S3. Train the machine learning model. The machine learning model selects the random forest model for training and classification recognition of coal and rock properties.

[0047] Specifically, S3 is as follows:

[0048] S31. Extract time-domain features and frequency-domain features from the training set, and form a comprehensive feature vector from the time-domain features and frequency-domain features.

[0049] The time-domain features in S31 include mean, variance, peak value, peak-to-peak value, and root mean square; the frequency-domain features in S31 include fast Fourier transform, spectral amplitude, spectral energy, power spectral density, and frequency-domain centroid.

[0050] ① Time-domain features

[0051] The mean (Mean) reflects the average level of a certain feature in the time series of the data set. The calculation formula is as follows:

[0052] ;

[0053] where is the mean; is the number of data points; is the th data point.

[0054] Variance measures the degree of dispersion of data around the mean, reflecting the stability or volatility of the data. The calculation formula is as follows:

[0055] ;

[0056] where, is the variance; is the number of data points; is the th data point; is the mean. Here, is used for unbiased estimation of the population variance.

[0057] The peak value is the point with the largest value in a given data sequence.

[0058] ;

[0059] where, is the peak value; represents the th data point. The maximum value is found by comparing all data points.

[0060] The peak-to-peak value is the difference between the maximum and minimum values in a data sequence, reflecting the fluctuation range of the data.

[0061] ;

[0062] where, is the peak-to-peak value; represents the th data point. The peak-to-peak value is obtained by finding the maximum and minimum values and calculating their difference.

[0063] The root mean square (RMS) takes the square root of the mean of the squares of the data and is commonly used to measure the effective strength or energy of a signal. The calculation formula is as follows:

[0064] ;

[0065] where, is the root mean square; is the number of data points, represents the th data point. First, square each data point, find the mean, and then take the square root.

[0066] ② Frequency domain characteristics

[0067] The Fast Fourier Transform (FFT) is to convert a time-domain signal into a frequency-domain representation to reveal the amplitude and phase information of different frequency components in the signal. The calculation formula is as follows:

[0068] ;

[0069] where, is the th frequency-domain coefficient; , is the number of time-domain data points; is the th time-domain data point; is the complex exponential term used to decompose the time-domain signal into different frequencies.

[0070] The Spectrum Amplitude is the amplitude of the frequency-domain coefficient after Fourier transform, reflecting the relative intensity of different frequency components in the signal. The calculation formula is as follows:

[0071] ;

[0072] where, is the frequency-domain coefficient after Fourier transform; is 's real part; is 's imaginary part. The spectrum amplitude is obtained by calculating the square root of the sum of the squares of the real part and the imaginary part.

[0073] The Spectral Energy is the sum of the energies of all frequency components of the signal in the frequency domain, used to measure the overall intensity of the signal in the frequency domain. The calculation formula is as follows:

[0074] ;

[0075] where, is the spectral energy; , is the number of frequency-domain data points (the same as the number of time-domain data points); is the th frequency-domain coefficient; is the spectrum amplitude. The spectral energy is obtained by summing the squares of the amplitudes of all frequency-domain coefficients.

[0076] The Power Spectral Density (PSD) is the sum of the energies of all frequency components of the signal in the frequency domain, used to measure the overall intensity of the signal in the frequency domain. The calculation formula is as follows:

[0077] ;

[0078] ;

[0079] Among them, is the estimated value of the power spectral density of the signal at frequency ; , is the number of time-domain data points; are the frequency-domain coefficients after Fourier transform; is the th frequency point; is the time-domain sampling interval. This formula estimates the power spectral density by squaring the spectral amplitude and dividing by the number of data points .

[0080] The centroid frequency is the "center of gravity" frequency of the power spectral density function, reflecting the frequency position where the signal energy is concentrated. For the power spectral density , the centroid frequency is:

[0081] ;

[0082] For the discrete power spectrum estimation , the centroid frequency is approximately:

[0083] ;

[0084] Among them, is the centroid frequency. In the continuous case, is the power spectral density function, is the frequency variable. In the discrete case, is the number of frequency-domain data points, is the th discrete frequency point, is the value of the estimated power spectral density at . The centroid frequency is calculated by weighted average, with the weight being the power spectral density.

[0085] In this embodiment, 5 time-domain features and 5 frequency-domain features are extracted for each pressure sensor, and there are a total of 5 × 5 = 25 features, forming a comprehensive feature vector.

[0086] S32. For each decision tree, the random forest randomly samples N samples from the training set with replacement to form the training subset D m :

[0087] ; Equation (4)

[0088] Among them, D mRepresents the training subset of the m-th tree; x i is the i-th data point in the dataset; N is the size of the dataset.

[0089] S33. Feature subset extraction.

[0090] Adopt the "dynamic feature subset extraction" strategy of random forest. Assume that the training set has m features. During the construction of each decision tree, when node splitting is required, randomly select k features, where, .

[0091] The extraction process uses random sampling without replacement to ensure that the feature subsets extracted each time are not repeated; the selected k features form the candidate feature subset of the current node, and the model calculates the best split point only based on these k features.

[0092] In the coal-rock identification scenario of a shearer, the original data of the pick forces collected by multiple pressure sensors has high dimensionality and low correlation. The time-domain and frequency-domain features (such as mean, variance, spectral energy, etc.) of the original data of each pick force form a multi-dimensional feature vector. The correlation between different features is weak, and some features may carry redundant or noisy information. Due to the low correlation between features, random extraction can maximize the diversity of feature subsets and avoid information redundancy. Through multiple random samplings, redundant features are dispersed into different decision trees, reducing their impact on the overall model.

[0093] S34. Select the best split feature and split point.

[0094] Among the selected k features, select a best feature and its split point according to the Gini impurity splitting criterion to maximize the purity of the subset after splitting on the target variable.

[0095] The specific calculation formula is as follows:

[0096] ; Equation (5)

[0097] where, is the Gini index; is the dataset of the current node; G is the total number of categories. There are two categories, coal and rock, and G takes 2; p i is the category in the node proportion.

[0098] In a random forest, the importance of a feature is measured by the average value of the reduction in Gini impurity brought by this feature in all decision trees. The calculation formula is as follows:

[0099] ; Equation (6)

[0100] where, is the importance of the j-th feature; M is the total number of decision trees; is the decrease in the Gini index brought by feature j in the m-th tree.

[0101] For each selected feature, traverse all possible split points, calculate the Gini impurity before and after splitting, and select the feature and split point that minimize the Gini impurity; according to the selected best feature and split point, divide the dataset D of the current node into two subsets and .

[0102] ; Equation (7)

[0103] where D is the dataset of the current node; A is the selected feature; split_point is the split point.

[0104] S35. Recursively construct subtrees.

[0105] For subsets and recursively execute steps S42 to S43 until the stopping condition is met.

[0106] The stopping conditions include: reaching the maximum tree depth; the number of samples in the node is less than the minimum sample split number; the impurity of the node is lower than the preset threshold; no further splitting is possible.

[0107] When the stopping condition is met, mark the node as a leaf node, and the category is determined by the majority category of the samples in the node; the calculation formula is as follows:

[0108] ; Equation (8)

[0109] where is the parameter function for taking the maximum value; is the indicator function; y i is the label of sample i; c is the current category; N is the number of samples; when it is 1, otherwise it is 0.

[0110] S36. Integrate multiple decision trees.

[0111] A random forest consists of multiple decision trees. Each tree randomly selects a feature subset and a sample subset during training; the final classification result is determined by the majority vote of all decision trees; for the m-th tree, the classification result is expressed as:

[0112] ; Equation (9)

[0113] where DecisionTree is the decision tree; is the prediction result of the m-th tree for the sample x; are the parameters of the m-th tree.

[0114] For classification tasks, the prediction results of all decision trees determine the final classification result through majority voting. The calculation formula is as follows:

[0115] ; Equation (10)

[0116] where is the final prediction result; is the mode function, which is used to select the value with the highest frequency from a set of values; is the prediction result of the m-th tree for the sample x; M is the total number of decision trees.

[0117] By integrating multiple decision trees, the random forest can effectively reduce the variance of a single tree and improve the overall generalization ability and stability.

[0118] S37. Hyperparameter tuning.

[0119] Optimize the classification performance of the model by adjusting hyperparameters.

[0120] Hyperparameters include: the number of trees, that is, the number of decision trees in the random forest; the maximum tree depth, that is, the maximum depth of each decision tree; the minimum number of samples for splitting, that is, the minimum number of samples required for internal node re-partitioning; the minimum number of samples in leaf nodes, that is, the minimum number of samples on leaf nodes; whether to use the bootstrap method, that is, whether to use bootstrap sampling when constructing trees.

[0121] Adopt the method of grid search combined with cross-validation to systematically search for the best hyperparameter combination:

[0122] Define the hyperparameter grid , where each is a different value of a hyperparameter. The goal is to find the hyperparameter combination that makes the performance of the validation set optimal. The calculation formula is as follows:

[0123] ; Equation (11)

[0124] where h* is the optimal hyperparameter combination; is the accuracy evaluation index of the hyperparameter combination h; H is the hyperparameter grid.

[0125] Adopt k-fold cross-validation. Divide the training set into k subsets, train the model with k - 1 subsets in turn, and use the remaining subset to verify the model performance. Finally, take the average of the k verification results as the model performance evaluation. The calculation formula is as follows:

[0126] ; Equation (12)

[0127] Among them, is the cross-validation score of the hyperparameter combination h; k is the number of subsets; represents the validation set of the i-th fold.

[0128] S4. Validate the machine learning model.

[0129] Specifically, S4 is as follows:

[0130] S41. Test set validation.

[0131] Use the test set to evaluate the accuracy, precision, recall, and F1-score of the machine learning model to ensure the generalization ability of the machine learning model on unseen data.

[0132] Accuracy refers to the proportion of the number of samples correctly predicted by the model to the total number of samples, which measures the overall prediction correctness of the model. The calculation formula is as follows:

[0133] ;

[0134] Among them, TP is the true positive example, that is, the number of samples that are actually positive and are correctly predicted as positive by the model; TN is the true negative example, that is, the number of samples that are actually negative and are correctly predicted as negative by the model; FP is the false positive example, that is, the number of samples that are actually negative but are wrongly predicted as positive by the model; FN is the false negative example, that is, the number of samples that are actually positive but are wrongly predicted as negative by the model.

[0135] Precision refers to the proportion of the samples actually positive among the samples predicted as positive by the model. It reflects the accuracy of the model in predicting positive classes. The calculation formula is as follows:

[0136] .

[0137] Recall refers to the proportion of the samples actually positive that are correctly predicted as positive by the model. It measures the ability of the model to find all positive samples. The calculation formula is as follows:

[0138] .

[0139] The F1-score is the harmonic mean of precision and recall:

[0140] .

[0141] S42. Cross-validation.

[0142] Adopt k-fold cross-validation to evaluate the stability and generalization ability of the machine learning model.

[0143] ; Equation (13)

[0144] Among them, is the cross-validation score of the machine learning model; k is the number of subsets; is the model object used for evaluation; represents the validation set of the i-th fold.

[0145] S43. If the requirements of the predicted set accuracy are met, the model training is completed and the prediction result is output; if the set accuracy requirements are not met, return to step S3, adjust the hyperparameters of the model, and perform model training again.

[0146] S5. Dynamically adjust the cutting parameters of the shearer according to the results output by the optimized model.

[0147] S5 specifically is:

[0148] S51. Formulate the basis and rules for adjusting the cutting parameters of the shearer.

[0149] S511. Construct the feature vector.

[0150] Combine all the extracted time-domain and frequency-domain features into a feature vector F:

[0151] ; Equation (14)

[0152] Among them, F is the combined feature vector; F a , F t , F n are the spectral amplitude feature, spectral energy feature, and power spectral density feature; M, V are the mean and variance; Peak, Peak-to-Peak are the peak and peak-to-peak value; RMS is the root mean square value; Dominant Frequency is the main frequency.

[0153] S512. Model prediction.

[0154] Input the feature vector F into the trained random forest model to obtain the prediction result :

[0155] ; Equation (15)

[0156] Among them, .

[0157] The prediction result , among which, represents the real-time feature extraction function.

[0158] If = coal, increase the rotational speed of the cutting drum, and / or, increase the cutting height, and / or, increase the traction speed.

[0159] If = rock, reduce the rotational speed of the cutting drum, and / or, reduce the cutting height, and / or, reduce the traction speed.

[0160] S52. The shearer controller dynamically adjusts the shearer cutting parameters.

[0161] The formula for adjusting the rotational speed of the cutting drum is:

[0162] ; Equation (16)

[0163] ; Equation (17)

[0164] Wherein, is the target rotational speed after adjustment, revolutions per minute, rpm; is the rotational speed of the current cutting drum, revolutions per minute, rpm; is the rotational speed adjustment ratio; T is the adjustment time constant, s, controlling the speed at which the rotational speed transitions from the current value to the target value; is the real-time adjusted rotational speed at time t.

[0165] The formula for adjusting the cutting height is:

[0166] ; Equation (18)

[0167] ; Equation (19)

[0168] Wherein, is the cutting height after adjustment, m, is the height of the current cutting drum, m; is the height adjustment ratio; T is the adjustment time constant, s, controlling the height at which the height transitions from the current value to the target value; is the real-time adjusted height at time t.

[0169] The formula for adjusting the traction speed is:

[0170] ; Equation (20)

[0171] ; Equation (21)

[0172] Wherein, is the shearer traction speed after adjustment, m / min; is the current shearer traction speed, m / min; is the traction speed adjustment ratio; T is the adjustment time constant, in s, for controlling the traction speed to transition from the current value to the target value; is the real-time adjusted traction speed at time t.

[0173] Implementation of the dynamic adjustment logic of the shearer controller for the shearer cutting parameters:

[0174] Receive adjustment instructions: The shearer controller receives new set values of rotational speed, traction speed, and cutting height.

[0175] Execute parameter adjustment: Use a variable frequency drive (VFD) to adjust the speed of the drum motor. The VFD precisely controls the output speed of the motor by changing the frequency and voltage of the power supply. During startup and adjustment, the VFD makes the motor speed change smoothly according to the preset acceleration and deceleration time parameters, avoiding mechanical shock and electrical overload to the equipment caused by sudden speed changes.

[0176] Traction speed adjustment: With the help of the drive controller of the traction motor, precisely control the speed of the traction motor by changing the power supply frequency and voltage of the motor. During the adjustment process, make the speed change process smoother according to the preset acceleration and deceleration curves, avoiding impact on the equipment.

[0177] Cutting height adjustment: Use the pressure oil provided by the hydraulic pump station to drive the boom lifting cylinder. Control the flow direction of the hydraulic oil through an electromagnetic directional valve, so that the piston rod of the cylinder extends or retracts, and then drives the boom to swing up and down around its rotation center to achieve the adjustment of the boom height. During the adjustment process, the boom height is monitored in real time by an angle sensor installed on the boom. When the set height is reached, the control system automatically stops the action of the cylinder to ensure the precise positioning of the boom height.

[0178] The shearer controller has fault self-diagnosis and safety redundancy design. In case of abnormalities such as communication interruption and sensor failure, it can ensure the stable shutdown of the shearer according to the preset safety mode (such as maintaining the lowest safety operation parameters) to prevent accidents.

[0179] Embodiment 2

[0180] This Embodiment 2 describes a computer device, which includes a memory and one or more processors. Executable code is stored in the memory. When the processor executes the executable code, it is used to implement the steps of the shearer adaptive control method based on pick force monitoring for coal and rock identification as in Embodiment 1 above.

[0181] Embodiment 3

[0182] Embodiment 3 describes a computer-readable storage medium with a program stored thereon. When the program is executed by a processor, it is used to implement the steps of the shearer adaptive control method for coal and rock identification based on pick force monitoring as described in Embodiment 1 above.

[0183] The computer-readable storage medium can be an internal storage unit of any device or apparatus with data processing capabilities, such as a hard disk or memory, or an external storage device of any device with data processing capabilities, such as a plug-in hard disk, a Smart Media Card (SMC), an SD card, a Flash Card, etc. equipped on the device.

[0184] So far, this embodiment has been described in detail with reference to the accompanying drawings. Based on the above description, those skilled in the art should have a clear understanding of the present invention. The shearer adaptive control method for coal and rock identification based on pick force monitoring of the present invention combines the operating parameters of the shearer, uses data processing and analysis algorithms to accurately and efficiently identify the coal-rock interface, and dynamically adjusts the cutting parameters of the shearer accordingly, realizing safe, efficient and high-precision coal and rock identification and the adaptive control of the shearer, improving the intelligence of coal mining operations, increasing the coal mining efficiency, reducing equipment wear, and ensuring the safety and stability of mining operations.

[0185] Of course, the above description is only a preferred embodiment of the present invention. The present invention is not limited to listing the above embodiments. It should be noted that all equivalent substitutions and obvious deformation forms made by any person skilled in the art under the teaching of this specification fall within the substantial scope of this specification and should be protected by the present invention.

Claims

1. An adaptive control method for a shearer based on the recognition of coal and rock by monitoring the forces on pick teeth, characterized in that pressure sensors are provided at the root of each pick tooth of the shearer to which the method is applied. The pressure sensors are signal-connected to a data acquisition unit, the data acquisition unit is signal-connected to a shearer controller, and the shearer controller is signal-connected to a host computer; The method comprises the following steps: S1. Obtain the original pick tooth force data collected by each pressure sensor and the cutting parameters of the shearer under different coal and rock properties during cutting; S2. Preprocess the original pick tooth force data; The specific content of S2 is as follows: S21. Denoise, smooth and normalize the original pick tooth force data; A filtering algorithm is used to remove the noise of the original pick tooth force data, and the calculation formula is: where y(t) is the output signal; x(τ) is the input signal; h(t - τ) is the impulse response function; t is time; The moving average method is used to reduce data fluctuations to smooth the signal, which is defined as follows: Among them, MA n (t) is the moving average value; n is the window size; x(t - i) is the delayed value of the input signal; t is the time; Data with different dimensions is normalized to eliminate the influence of dimensions, and the data is converted into a distribution with a mean of 0 and a standard deviation of 1. The calculation formula is: where x' i is its value after standardization; μ i is the mean of feature x i ; σ i is the standard deviation; S22. Data annotation; The normalized original pick tooth force data is annotated according to different cutting processes, and the annotated data is used as the true value sample data set for training the machine learning model; One part of the true value sample data set is used as the training set to train the random forest model; the remaining other part of the true value sample data set is used as the test set to test the accuracy of the machine learning model and evaluate the generalization ability and performance indicators of the machine learning model; S3. Train the machine learning model, and the machine learning model selects the random forest model for training and classification recognition of coal and rock properties; S4. Verify the machine learning model; S5. Dynamically adjust the cutting parameters of the shearer according to the results output by the optimized model.

2. The shearer adaptive control method based on shearer pick force monitoring for coal and rock identification according to claim 1, wherein, The specific content of S1 is as follows: S11. During the coal and rock cutting operation of the shearer, the data acquisition unit real-time collects the original pick tooth force data monitored by each pressure sensor during the entire cutting process. The entire cutting process includes the process of cutting into coal and rock, the process of stable cutting of coal and rock, and the process of cutting out from coal and rock; the original pick tooth force data includes the axial, normal and tangential forces on the pick teeth. The data acquisition unit uploads the original pick tooth force data to the shearer controller; at the same time, the shearer controller records the cutting parameters of the shearer, where the cutting parameters of the shearer include the rotational speed of the cutting drum, the cutting height and the traction speed; S12. The shearer controller uploads the cutting parameters of the shearer and the original pick tooth force data to the host computer, and uploads the different coal and rock properties corresponding to the shearer cutting to the host computer.

3. The shearer adaptive control method based on shearer pick force monitoring for coal and rock identification according to claim 1, wherein The specific content of S3 is as follows: S31. Extract time-domain features and frequency-domain features from the training set, and form a comprehensive feature vector from the time-domain features and frequency-domain features; S32. For each decision tree, the random forest randomly draws N samples from the training set with replacement to form the training subset D of the decision tree m : D m = {x1, x2,... x i ,... x N}; Equation (4) Among them, D m represents the training subset of the m-th tree, and x i is the i-th data point in the dataset, and N is the size of the dataset; S33. Feature subset extraction; Adopt the "dynamic extraction of feature subsets" strategy of random forest. Assume that the training set has m features. During the construction of each decision tree, when node splitting is required, randomly select k features, where The extraction process uses random sampling without replacement to ensure that the feature subsets extracted each time are not repeated; the selected k features form the candidate feature subset of the current node; S34. Select the best splitting feature and splitting point; Among the selected k features, a best feature and its splitting point are selected according to the Gini impurity splitting criterion, so that the purity of the subsets after splitting is maximized on the target variable; The specific calculation formula is as follows: Among them, Gini(D) is the Gini index; D is the dataset of the current node; G is the total number of categories. There are two categories, coal and rock, and G takes 2; p i is the proportion of category i in node D; In a random forest, the importance of a feature is measured by the average of the reduction in Gini impurity brought by this feature in all decision trees. The calculation formula is as follows: Among them, Feature Importance j is the importance of the j-th feature; M is the total number of decision trees; ΔGini j,m is the reduction in Gini index brought by feature j in the m-th tree; For each selected feature, traverse all possible split points, calculate the Gini impurity before and after the split, and select the feature and split point that minimize the Gini impurity; according to the selected best feature and split point, divide the dataset D of the current node into two subsets D left and D right ; D left ,D right = Split(D, A, split_point); Equation (7) where D is the dataset of the current node; A is the selected feature; split_point is the splitting point; S35. Recursively construct subtrees; For subset D left and D right Recursively execute steps S42 to S43 until the stop condition is satisfied; The stopping conditions include: reaching the maximum tree depth; the number of samples in the node is less than the minimum sample splitting number; the impurity of the node is lower than the preset threshold; no further splitting is possible; When the stopping condition is met, mark this node as a leaf node, and the category is determined by the majority category of the samples in this node; the calculation formula is as follows: Among them, is a parameter function for taking the maximum value; Π is an indicator function; y i is the label of sample i; c is the current category; N is the number of samples; when y i = c, the value is 1, otherwise it is 0; S36. Integrate multiple decision trees; A random forest consists of multiple decision trees. Each tree randomly selects a subset of features and a subset of samples during training; the final classification result is determined by the majority vote of all decision trees; for the m-th tree, the classification result is expressed as: h m (x) = DecisionTree(x; Θ m ); Equation (9) Among them, DecisionTree is the decision tree; h m (x) is the prediction result of the m-th tree for the sample x; Θ m is the parameter of the m-th tree; For a classification task, the prediction results of all decision trees are used to determine the final classification result by majority vote. The calculation formula is as follows: Among them, is the final prediction result; mode is the mode function used to select the value with the highest frequency from a set of values; h m (x) is the prediction result of the m-th tree for the sample x; M is the total number of decision trees; S37. Hyperparameter tuning; Optimize the classification performance of the model by adjusting hyperparameters; Hyperparameters include: the number of trees, that is, the number of decision trees in the random forest; the maximum tree depth, that is, the maximum depth of each decision tree; the minimum sample splitting number, that is, the minimum number of samples required for internal node re-division; the minimum number of samples in leaf nodes, that is, the minimum number of samples on leaf nodes; whether to use the bootstrap method, that is, whether to use bootstrap sampling when constructing trees; Adopt a method combining grid search and cross-validation to systematically search for the best hyperparameter combination: Define the hyperparameter grid $H = \{h_1, h_2, \ldots, h$ n $\}$, where each $h$ i is a different value of a hyperparameter. The goal is to find the hyperparameter combination that maximizes the performance on the validation set, and the calculation formula is as follows: where h* is the optimal hyperparameter combination; Performance(h) is the accuracy evaluation index of the hyperparameter combination h; H is the hyperparameter grid; Adopt k-fold cross-validation. Divide the training set into k subsets. Use k - 1 subsets to train the model in turn, and use the remaining subset to verify the model performance. Finally, take the average of the k verification results as the model performance evaluation. The calculation formula is as follows: Among them, CV tune (h) is the cross-validation score of the hyperparameter combination h; k is the number of subsets; represents the validation set of the i-th fold.

4. The shearer adaptive control method based on shearer pick force monitoring for coal and rock identification according to claim 3, wherein The time-domain features in S31 include mean, variance, peak value, peak-to-peak value, and root mean square; the frequency-domain features in S31 include fast Fourier transform, spectral amplitude, spectral energy, power spectral density, and frequency-domain centroid.

5. The shearer adaptive control method based on shearer pick force monitoring for coal and rock identification according to claim 3, wherein, The specific content of S4 is as follows: S41. Test set verification; Use the test set to evaluate the accuracy, precision, recall rate, and F1 score of the machine learning model to ensure the generalization ability of the machine learning model on unseen data; S42. Cross-validation; Adopt k-fold cross-validation to evaluate the stability and generalization ability of the machine learning model; Among them, CV final (h) is the cross-validation score of the machine learning model; k is the number of subsets; h is the model object used for evaluation; represents the validation set of the i-th fold; S43. If the requirements of the predicted set accuracy are met, the model training is completed and the prediction result is output; if the set accuracy requirements are not met, return to step S3, adjust the hyperparameters of the model, and perform model training again.

6. The shearer adaptive control method based on shearer pick force monitoring for coal and rock identification according to claim 5, characterized in that The specific content of S5 is as follows: S51. Formulate the basis and rules for adjusting the cutting parameters of the shearer; S511. Feature vector construction; Combine all the extracted time-domain and frequency-domain features into a feature vector F: F = [F a , F t , F n , M, V, Peak, Peak-to-Peak, RMS, Dominant Frequency,...]; Equation (14) Among them, F is the combined feature vector; F a , F t , F n are the spectral amplitude feature, spectral energy feature, and power spectral density feature; M and V are the mean and variance; Peak and Peak-to-Peak are the peak and peak-to-peak value; RMS is the root mean square value; DominantFrequency is the main frequency; S512. Model prediction; Input the feature vector F into the trained random forest model to obtain the prediction result Among them, Prediction result where f(x real-time ) represents the real-time feature extraction function; If increase the rotational speed of the cutting drum, and / or increase the cutting height, and / or increase the traction speed; If reduce the rotational speed of the cutting drum, and / or, reduce the cutting height, and / or, reduce the traction speed; S52. The shearer controller dynamically adjusts the cutting parameters of the shearer; The rotation speed adjustment formula for the cutting drum is: RPM new = RPM current × (1 + ΔRPM); Equation (16) Among them, RPM new is the adjusted target rotational speed; RPM current is the rotational speed of the current cutting drum; ΔRPM is the rotational speed adjustment ratio; T is the adjustment time constant, which controls the speed at which the rotational speed transitions from the current value to the target value; RPM adjust (t) is the real-time adjusted rotational speed at time t; The cutting height adjustment formula is: H new = H current × (1 + ΔH); Equation (18) Among them, H new is the adjusted cutting height; H current is the height of the current cutting drum; ΔH is the height adjustment ratio; T is the adjustment time constant, which controls the height transition from the current value to the target value; H adjust (t) is the real-time adjusted height at time t; The traction speed adjustment formula is: TS new = TS current × (1 + ΔTS); Equation (20) Among them, TS new is the adjusted traction speed of the shearer; TS current is the current traction speed of the shearer; ΔTS is the traction speed adjustment ratio; T is the adjustment time constant, which controls the traction speed from the current value to the target value; TS adjust is the real-time adjusted traction speed at time t.

7. The shearer adaptive control method based on shearer pick force monitoring for coal and rock identification according to claim 1, characterized in that, The data acquisition unit is signal-connected to the shearer controller via a wireless transmission module.

8. A computer device, including a memory and one or more processors, wherein executable code is stored in the memory, characterized in that, When the processor executes the executable code, it is used to implement the shearer adaptive control method based on coal and rock identification by monitoring the force on pick teeth as described in any one of claims 1 to 7.

9. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements the shearer adaptive control method based on coal and rock identification by monitoring the force on pick teeth as described in any one of claims 1 to 7.

Images