A method of automatic height control for a plough

Abstract

The application discloses an automatic height adjusting control method of a coal plough, and belongs to the field of intelligent control of coal mining equipment, which comprises the following steps: adopting a top-bottom cutter group as a pick-type cutting tooth and a waist cutter group as a mixed arrangement of pick-type and knife-type cutting teeth, optimizing plough parameters through orthogonal test and joint simulation; collecting plough load, lower plough body stress and vibration multi-source signals, extracting time-frequency characteristics, and identifying coal-rock interface, gangue layer position and thickness based on SVM; taking the identification results, load fluctuation coefficient and ploughing parameters as inputs, dynamically calculating the height adjusting amount of the upper and middle plough groups through fuzzy control; and verifying and optimizing the control strategy through closed-loop iteration. The application realizes accurate identification of coal and rock, parameter collaborative optimization and dynamic load balance, controls the load fluctuation coefficient within 30%, and improves ploughing efficiency and tool life.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control technology for coal mining equipment, and particularly relates to an automatic height adjustment control method for a coal planer. Background Technology

[0002] As the core equipment for thin coal seam mining, the coal planer's working efficiency, tool life, and operational stability directly affect the economy and safety of coal mining. In actual planing operations, conditions such as uneven coal seam thickness, roof and floor undulations, sudden changes in coal and rock hardness, and random distribution of interbedded rock lead to uneven and drastic load distribution on the planer blades. This can easily result in problems such as excessively rapid tool wear and breakage, delayed response of the height adjustment mechanism, and insufficient positioning accuracy, severely restricting the efficient and stable operation of the coal planer.

[0003] Existing coal planer height control systems mostly employ fixed parameter control modes, enabling only simple height adjustments and failing to adjust control strategies in real time based on changes in coal and rock properties or fluctuations in planer blade load. Regarding blade arrangement, current solutions often use a single type of cutting tooth or a fixed-ratio mixed arrangement, without optimizing parameters to consider planing conditions and load balancing requirements, resulting in a trade-off between coal breaking efficiency and tool life.

[0004] In the field of coal and rock identification, existing research has attempted to identify coal and rock by collecting single or multiple types of signals using current sensors, vibration sensors, and acoustic emission sensors. Currently, publicly available technologies include methods for identifying coal-rock interfaces using current detection from a cutting motor, with current limits between the coal seam and rock strata recorded using a demonstration cutter as a criterion. Other proposed methods include coal and rock identification based on fuzzy neural network information fusion, which collects multiple signals such as vibration, current, and acoustic emission to construct an identification model, as well as coal-rock interface identification methods based on dual noise reduction and improved support vector machines. Regarding simulation studies of the influence of planing parameters on load characteristics, some scholars have also established simulation analysis models using software such as EDEM and ANSYS.

[0005] However, existing technologies lack a comprehensive, integrated technical solution that organically combines online coal and rock identification, cutter arrangement parameter optimization, dual-drive mechanism selection, and fuzzy dynamic height adjustment control. Current height adjustment control methods lack deep integration between coal and rock identification results and height adjustment control decisions. Identification accuracy is greatly affected by the environment, control rules largely rely on manual experience, and there is a lack of closed-loop iterative optimization mechanisms. This makes it difficult to achieve dynamic balance control of the cutter group load and efficient, safe, and energy-saving operation of the coal cutter.

[0006] Therefore, developing an automatic height adjustment control method for coal planers that can achieve online coal and rock identification, collaborative optimization of planer parameters, and dynamic load balancing has become an urgent technical challenge to be solved. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides an automatic height adjustment control method for a coal planer.

[0008] An automatic height adjustment control method for a coal planer includes the following steps:

[0009] Planer blade group layout and parameter optimization: The planer blade group is arranged in sections, with the waist blade group using a mixed arrangement of pick-type and blade-type cutting teeth. The planer blade layout parameters and height response speed are optimized through orthogonal experiments and joint simulation.

[0010] Multi-source signal sensing and coal and rock identification: Collect multi-source signals during the planing process, extract signal feature vectors, and identify coal and rock properties online based on a support vector machine model;

[0011] Fuzzy dynamic height adjustment and control execution: Taking the coal and rock identification results, load fluctuation coefficient and planing parameters as input, the real-time height adjustment of the planing tool group is calculated through fuzzy control and output to the height adjustment execution mechanism;

[0012] Closed-loop iterative verification: The control strategy is applied to the test platform, and the model and parameters are corrected through feedback from the test data, iterating until the preset target is met.

[0013] In the partitioned arrangement of the planer blade group, both the top blade group and the bottom blade group adopt pick-type cutting teeth, the proportion of pick-type cutting teeth in the waist blade group is set to 60% to 80%, and the mixed spacing of the planer blades is set to 60mm to 100mm.

[0014] The parameter optimization specifically includes: using L9 orthogonal experiments to screen for better parameter combinations, and using the EDEM-ANSYS co-simulation platform to correct the proportion of pick-type cutting teeth in the waist knife group, the mixed spacing of the planer blades, and the height response speed to determine the optimized parameter set.

[0015] The multi-source signals include planer load signal, lower planer force signal, and mechanism vibration acceleration signal; the signal feature vector includes time domain features and frequency domain features; the preprocessing of the multi-source signals includes wavelet decomposition, improved soft threshold denoising, and low-pass filtering.

[0016] The time-domain features include mean, variance, peak value, and kurtosis, and the frequency-domain features include dominant frequency, spectral energy, and frequency band distribution.

[0017] The support vector machine model adopts a multi-output architecture, predicting the coal-rock interface location, interbedded rock layer position, and interbedded rock thickness through three parallel sub-models. It uses a radial basis kernel function and optimizes the penalty coefficient C and kernel function parameter g through a grid search method.

[0018] In the fuzzy control, the Mamdani minimum method is used to calculate the activation intensity of the control rule, and the centroid method is used to defuzzify the fuzzy output into a precise adjustment amount, so that the load fluctuation coefficient of the planer group is stabilized within the target range of ≤30%.

[0019] The real-time height adjustment is calculated only for the upper and middle planer groups, while the bottom planer group remains unchanged.

[0020] The preset targets for the closed-loop iterative verification include: the load fluctuation coefficient of the cutter group ≤30% and the coal breaking energy consumption ≤0.4kW·h / m³.

[0021] The method also includes a height adjustment mechanism selection and adaptation step: adopting a combination drive scheme of ball screw and servo motor or a combination drive scheme of hydraulic cylinder and hydraulic servo system, and cooperating with dovetail precision guide rail, and determining the optimal drive form through response speed, positioning accuracy, load-bearing capacity and energy consumption tests.

[0022] By employing the above technical solution, the present invention has at least the following beneficial effects:

[0023] (1) The automatic height adjustment control method of the coal planer provided by the present invention can realize online accurate identification of coal and rock: by collecting multi-source signals through strain gauges and triaxial acceleration sensors, combined with SVM model, it can identify the coal and rock interface, interbedded rock layer and thickness change in real time, which solves the problem of blindness in the existing height adjustment control and provides accurate working condition basis for dynamic height adjustment;

[0024] (2) The automatic height adjustment control method of the coal planer provided by the present invention can achieve parameter synergistic optimization: through orthogonal experiment and EDEM-ANSYS joint simulation, the planer blade arrangement parameters and height adjustment response speed are synergistically optimized, taking into account coal breaking efficiency, load balance and tool life, and avoiding performance shortcomings caused by single parameter optimization.

[0025] (3) The automatic height adjustment control method of the coal planer provided by the present invention can realize dynamic load balance control: based on the fuzzy control algorithm, combined with the coal and rock identification results and real-time load fluctuation, the height adjustment amount is dynamically calculated, and the load fluctuation coefficient of the planer group is controlled within 30%, which effectively reduces the risk of planer wear and breakage and extends the tool life.

[0026] (4) The automatic height adjustment control method of the coal planer provided by the present invention can adapt to complex working conditions: by selecting a dual-drive height adjustment mechanism, it takes into account both response speed and positioning accuracy, adapts to complex planing working conditions such as roof and floor undulations and uneven coal and rock hardness, and improves the stability and reliability of the coal planer operation.

[0027] (5) The automatic height adjustment control method of the coal planer provided by the present invention is energy-saving and efficient: by optimizing the planer blade arrangement and height adjustment control strategy, the coal breaking ratio energy consumption is reduced and the planing efficiency is improved, providing technical support for efficient, energy-saving and safe mining of thin coal seams. Attached Figure Description

[0028] Figure 1 The overall design flowchart of the automatic height adjustment control method for a coal planer provided by the present invention is shown below.

[0029] Figure 2 A schematic diagram of the partitioned arrangement of the planer blade group;

[0030] Figure 3 A schematic diagram of the coupled architecture for EDEM-ANSYS co-simulation;

[0031] Figure 4 A schematic diagram showing the installation location of the strain gauge for the planer tool;

[0032] Figure 5 This is a schematic diagram showing the installation location of the triaxial accelerometer.

[0033] Figure 6 This is a schematic diagram showing the installation position of the strain gauge on the lower planer body;

[0034] Figure 7 This is a flowchart of the online identification process for coal and rock properties based on Support Vector Machine (SVM).

[0035] Figure 8 A flowchart for closed-loop iterative verification and parameter optimization;

[0036] In the picture:

[0037] 1. Blade-shaped cutting teeth; 2. Pick-shaped cutting teeth; 3. Top cutter assembly; 4. Bottom cutter assembly; 5. Strain gauge; 6. Lower planer body; 7. Roof plate; 8. Coal body; 9. Accelerometer; 10. Guide chain frame. Detailed Implementation

[0038] To better explain and facilitate understanding of the present invention, the technical solution and effects of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0039] Combination Figures 1-8 As shown in the figure, the automatic height adjustment control method for a coal planer provided in this embodiment specifically includes the following steps:

[0040] S1: Basic parameter presets and mixed planer blade arrangement.

[0041] Based on the results of previous tests on the coal breaking characteristics of single-blade and pick-type cutting teeth, and combined with the derivation of coal and rock fracture mechanics theory, a zoned layout scheme for the cutting blade group was determined, and its specific layout is as follows:

[0042] S1.1: Arrangement of top cutter group 3 and bottom cutter group 4: Both top cutter group 3 and bottom cutter group 4 adopt pick-type cutting teeth 2. Because the installation direction of the pick-type cutting teeth 2 is close to the planing tangent direction, it has the characteristics of strong impact resistance and good adaptability to complex working conditions of top and bottom plates, and can maintain good cutting performance when encountering complex media such as hard rock and gangue during the planing process.

[0043] S1.2: Hybrid Arrangement of the Waist Cutter Group: The waist cutter group adopts a mixed arrangement of pick-type cutting teeth 2 and knife-type cutting teeth 1. The proportion of pick-type cutting teeth 2 is set at 60%-80%, with the specific proportion to be optimized based on subsequent orthogonal experiments; knife-type cutting teeth 1 supplement the shearing and coal breaking efficiency. Knife-type cutting teeth 1 are installed along the planing direction, cutting coal by a cutting method, with a cutting depth greater than that of pick-type cutting teeth 2, which can improve the traction speed; pick-type cutting teeth 2 break the coal body 8 by a wedge-cutting method, with a large cutting depth, and can automatically sharpen during operation, resulting in lower energy consumption and longer service life than knife-type cutting teeth 1. The two complement each other, taking into account both hard point crushing and efficient coal body cutting.

[0044] S1.3: Planer blade mixing spacing setting: The planer blade mixing spacing is set to 60 mm ~ 100 mm. The determination of this spacing range is based on the theoretical calculation results of load balance of the planer blade group. If the spacing is too large, it will lead to incomplete coal breaking and coal dust accumulation. If the spacing is too small, it will lead to concentrated load on the planer blades and accelerated wear.

[0045] After the above arrangement plan, the following was formed: Figure 2 The illustrated planer blade group zoning scheme provides a foundation for subsequent parameter optimization and control strategy implementation.

[0046] S2: Orthogonal experiments are used to screen for optimal parameter combinations.

[0047] This step involves an orthogonal experimental design based on three core influencing factors: the proportion of pick-type cutting teeth in the waist knife group, the mixed spacing of the planer blades, and the increase in response speed.

[0048] S2.1: Factor and Level Settings: Set the proportion of the pick-type cutting teeth 2 of the waist cutter group as factor A, with three levels: 60%, 70%, and 80%. Set the planer blade mixing distance as factor B, with three levels: 60mm, 80mm, and 100mm. Set the response speed adjustment as factor C, with three levels: 0.3s, 0.5s, and 0.7s.

[0049] S2.2: Determination of Evaluation Indicators: The three core evaluation indicators are the load fluctuation coefficient of the planer blade group, the coal breaking ratio energy consumption, and the wear uniformity. The load fluctuation coefficient reflects the uniformity of the load distribution of the planer blade group, the coal breaking ratio energy consumption reflects the energy efficiency of the planing operation, and the wear uniformity reflects the stability of the tool life.

[0050] S2.3: Orthogonal Experiment: According to L9 (3 4 An orthogonal experimental design was used to conduct nine planing tests, as shown in Table 1. Data for the three core evaluation indicators were collected for each test. For example, in the first test, the proportion of pick-type cutting teeth 2 in the planer group was 60%, the mixed spacing of the planer blades was 60mm, and the height adjustment response speed was 0.3s; in the second test, the proportion was 60%, the spacing was 80mm, and the speed was 0.5s; and so on to complete all nine test groups.

[0051] S2.4: Range Analysis and Significance Test: Statistical analysis is performed on the evaluation index data recorded in each group of experiments. Range analysis is used to calculate the degree of influence of each factor on each evaluation index and determine the order of importance of different factors; significance test is used to determine the statistical significance of each factor on the evaluation index.

[0052] S2.5: Selection of Optimal Parameter Combinations: Taking into account the optimization requirements of the three evaluation indicators, and considering the requirements of balanced load on the cutter group, high coal breaking efficiency, and uniform tool wear, the optimal parameter combinations were selected. The parameter combinations that achieve the following results are considered optimal: load fluctuation coefficient ≤ 25%, coal breaking specific energy consumption ≤ 0.35 kW·h / m³, and wear uniformity index above 0.85. These combinations provide initial parameters for subsequent simulation optimization.

[0053] Table 1. Schematic diagram of factors and levels in orthogonal experiments

[0054] 1 A1 B1 C1 2 A1 B2 C2 3 A1 B3 C3 4 A2 B1 C2 5 A2 B2 C3 6 A2 B3 C1 7 A3 B1 C3 8 A3 B2 C1 9 A3 B3 C2

[0055] Among them, Factor A is the proportion of pick-type cutting teeth 2 in the waist knife group, with horizontal values ​​of A1: 60%, A2: 70%, and A3: 80%; Factor B is the mixed spacing of the planer blades, with horizontal values ​​of B1: 60mm, B2: 80mm, and B3: 100mm; and Factor C is the adjustment of the response speed, with horizontal values ​​of C1: 0.3s, C2: 0.5s, and C3: 0.7s. The evaluation indicators are the load fluctuation coefficient of the planer blade group, the coal breaking energy consumption, and the uniformity of planer blade wear.

[0056] Step 3: Refine the parameters through co-simulation.

[0057] This step utilizes the EDEM-ANSYS co-simulation platform to construct a coupled simulation model for coal and rock breaking and cutter stress analysis. Details are as follows:

[0058] Step 3.1: Model Construction: Input the optimal parameter combination selected in Step 2 into the simulation model, establish a coal and rock particle model in EDEM software to simulate the crushing effect of coal and rock and the change of cutting resistance during the planing process; establish a planer finite element model in ANSYS software, apply the load spectrum calculated by EDEM as the boundary condition to the ANSYS model to simulate the stress distribution characteristics, strain and fatigue life of the planer.

[0059] Step 3.2: Simulation Process: Run EDEM-ANSYS coupled simulation to obtain coal crushing characteristic data such as coal and rock fragmentation morphology and coal dust discharge efficiency. Simultaneously, obtain the stress distribution cloud map, strain cloud map, and fatigue life prediction results for the planer blade under different cycle numbers. Identify stress concentration areas and easily worn parts of the planer blade through simulation.

[0060] Step 3.3: Parameter Correction: Based on the simulation results, analyze the influence of each parameter on the load fluctuation, stress distribution, and fatigue life of the planer cutter, and optimize the ratio of pick-type cutting teeth 2 and cutter-type cutting teeth 1, the mixing distance, and the response speed. For example, if the simulation finds that the proportion of pick-type cutting teeth 2 in the planer cutter group is too high, leading to load concentration, then appropriately reduce this proportion; if the planer cutter mixing distance is found to be too large, causing poor coal dust discharge and resulting in additional load, then appropriately reduce the distance. By specifically correcting the proportion of cutting teeth in the planer cutter group, the mixing distance, and increasing the response speed, the problems of load concentration, excessive stress, and uneven wear that occurred during the simulation are solved, and the preliminary optimized parameter set is determined. The coupling process diagram of the EDEM-ANSYS co-simulation is shown below. Figure 3 As shown.

[0061] Step 4: Select and adapt the dual-drive height adjustment mechanism.

[0062] This step involves building test prototypes of two height adjustment drive schemes for comparison and selection.

[0063] Step 4.1: Construction of two drive schemes: Both schemes are equipped with dovetail-shaped precision guide rails to ensure the guiding accuracy and operational stability of the height adjustment mechanism.

[0064] Option 1: A drive scheme that combines a ball screw and a servo motor. The servo motor outputs power, which is converted into linear motion through the ball screw pair, thereby driving the planer head to rise and fall.

[0065] Option 2: A drive scheme that combines a hydraulic cylinder and a hydraulic servo system, where the hydraulic servo system controls the extension and retraction of the hydraulic cylinder piston rod, and the height adjustment action is achieved through hydraulic drive.

[0066] Step 4.2: Static and Dynamic Performance Testing: The following tests were performed on the prototypes of the two drive schemes:

[0067] Step 4.2.1: Response speed test: Test the action completion time under different conditions such as height adjustment of 10mm, 20mm, 30mm and 50mm, and record the response delay.

[0068] Step 4.2.2: Positioning accuracy test: Test the deviation between the actual displacement and the set displacement of the height adjustment mechanism, and measure the steady-state positioning error and repeatability accuracy.

[0069] Step 4.2.3: Load-bearing capacity test: Test the operational stability of the mechanism under rated load conditions, including smooth start-stop and speed uniformity.

[0070] Step 4.2.4: Energy consumption test: Under the same planing conditions, test the overall energy consumption of the two schemes and the energy consumption of a single height adjustment action.

[0071] Step 4.3: Adaptation Decision: By comprehensively comparing the test results across four dimensions—response speed, positioning accuracy, load-bearing capacity, and energy consumption—the optimal height adjustment drive form is determined. Specifically, for scenarios such as thin coal seams, the optimal height adjustment drive form is determined through comprehensive comparison of test data. For example, in one specific application scenario, based on a comprehensive comparison of test results across these four dimensions, a hydraulic servo drive is selected due to the prominent need for high-speed response. In another specific application scenario, based on a comprehensive comparison of test results across these four dimensions, a ball screw solution is selected due to high requirements for positioning accuracy and energy conservation. The specific form is selected based on the specific circumstances, and this selection result provides suitable actuator support for the subsequent implementation of the automatic height adjustment control strategy.

[0072] Step 5: Multi-source signal acquisition and feature extraction.

[0073] In this step, on the prototype with the optimal height adjustment drive form determined in step 4, sensors are installed and multi-source signals are acquired, and feature extraction of the signals is completed.

[0074] Step 5.1: Sensor Installation and Arrangement: Strain gauges 5 are installed in the stress concentration areas of the pick-shaped cutting teeth 2 and the blade-shaped cutting teeth 1 at typical positions on the top and waist of the planer to collect the real-time load signal F(t) of the planer during the planing process, such as... Figure 4 As shown; a triaxial accelerometer 9 is installed on the base of the height adjustment mechanism to collect the real-time vibration signal G(t) of the mechanism, as shown. Figure 5 As shown; strain gauges 5 are installed at the contact positions between the left and right lower planer bodies 6 and the guide frame 10 to collect collision signals or force signals F_body(t) of the lower planer body 6 during planing. Figure 6 As shown.

[0075] Step 5.2: Multi-condition signal acquisition: Simulate planing operations under different coal and rock conditions, including but not limited to pure coal, coal interbedded with gangue, and hard rock. Simultaneously acquire planer load signals, lower planer body 6 force signals, and mechanism vibration signals under each condition. Collect no less than 1024 sampling points for each condition to ensure the amount of data and statistical reliability for subsequent feature extraction.

[0076] Step 5.3: Signal Preprocessing: Perform the following processing on the acquired raw signals in sequence:

[0077] Step 5.3.1: Wavelet Decomposition: The original signal x(t), containing the load F(t), the force F_body(t) on the lower excavator 6, and the vibration acceleration G(t), is subjected to N-level wavelet decomposition, where N is 3 levels, resulting in one low-frequency approximation coefficient A_N and three high-frequency detail coefficients D1, D2, and D3. The three-level decomposition scheme balances denoising effect and computational efficiency, meeting the requirements of real-time online processing. Here, x(t) = A3 + D1 + D2 + D3, where A3 is the core effective component of the signal, representing the main information of coal and rock cutting; D1, D2, and D3 are the noisy high-frequency components, with D1 having the strongest noise and D3 containing some effective impact components.

[0078] Step 5.3.2: Threshold Calculation and High-Frequency Coefficient Processing: An improved soft thresholding function is used to process D1, D2, and D3 respectively. First, the adaptive threshold T is calculated based on the noise statistical characteristics:

[0079] ;

[0080] ;

[0081] Where: n is the number of original signal sampling points (taken as 1024); σ is the median absolute value of the high-frequency coefficients of the first layer; σ is the estimated noise standard deviation.

[0082] The improved soft thresholding formula is as follows:

[0083] , ;

[0084] Where i = 1, 2, 3; exp( This is used to alleviate the constant deviation problem of soft threshold, making the processed signal closer to the true value.

[0085] Step 5.3.3: Signal reconstruction: Reconstruct the processed high-frequency coefficients D1', D2', and D3' with the low-frequency coefficient A3 to obtain the denoised signal x'(t): x'(t) = A3 + D1' + D2' + D3'.

[0086] Step 5.3.4: Low-pass filter design:

[0087] Design a fourth-order Butterworth low-pass filter for secondary filtering, with a cutoff frequency f. c Select 500Hz.

[0088] The transfer function (in the s-domain) is: ,in: Angular cutoff frequency (f) c At 500Hz, w c≈3141.6 rad / s).

[0089] Step 5.3.5: Discretization Implementation (Adapting to Real-Time Acquisition Systems):

[0090] The transfer function in the s-domain is transformed to the z-domain using the bilinear transformation method to avoid frequency aliasing. The transformation formula is as follows: Where: Ts=1 / fs is the sampling period (fs=1kHz, so Ts=1ms).

[0091] Substituting H(s) into the equation yields the z-domain transfer function H(z), which is then transformed into a difference equation: ,in Let y(k) be the signal after wavelet denoising, and y(k) be the signal after low-pass filtering. These are the coefficients of a 4th-order Butterworth low-pass filter.

[0092] The filtering operation is performed in real time using a DSP chip. The signal x'(t) after wavelet denoising is calculated point by point according to the difference equation. The processing time for a single frame signal is less than 0.1ms, which meets the real-time requirements of online recognition.

[0093] Step 5.4: Feature Extraction:

[0094] Step 5.4.1: Time-domain feature extraction: Based on the clean signal y(t) after low-pass filtering, extract four core time-domain features: mean μ, variance σ², peak value P, and kurtosis K. For each signal (planer load F, lower planer body 6 force F), extract the following features: mean μ, variance σ², peak value P, and kurtosis K. body Vibration G y Each of the four time-domain features is extracted, resulting in a total of 12-dimensional time-domain feature vectors.

[0095] Step 5.4.2: Frequency domain feature extraction: The preprocessed time-domain clean signal is converted into a frequency domain signal by Fast Fourier Transform (FFT), and three core frequency domain features are extracted: main frequency f0, spectral energy E, and frequency band distribution R.

[0096] Among them, the dominant frequency reflects the strongest vibration frequency of the signal: ;

[0097] Spectral energy reflects the total intensity of the signal in the frequency domain: ;

[0098] Frequency band distribution reflects the proportion of high-frequency impact energy: .

[0099] Three frequency domain features are extracted for each type of signal, resulting in a total of nine-dimensional frequency domain feature vectors.

[0100] Step 5.5: Signal Feature Vector Construction: Combine the above time-domain and frequency-domain features to form a 21-dimensional signal feature vector, which serves as the input to the SVM coal and rock identification model. .

[0101] This 21-dimensional feature vector serves as the input to the subsequent SVM coal and rock identification model, providing data support for online identification of coal and rock properties and height control decisions.

[0102] Step 6: Online identification of coal and rock properties.

[0103] This step uses the 21-dimensional signal feature vector extracted in step 5 as the input variable, and the location of the coal-rock interface, the location of the interbedded rock layer, and its thickness as the output labels to construct a support vector machine (SVM) coal-rock property identification model. Details are as follows:

[0104] Step 6.1: Sample Set Construction: Select experimental sample data under different coal and rock conditions, use a 21-dimensional signal feature vector as input, and use the coal-rock interface location, interbedded rock layer position, and interbedded rock thickness as output labels to construct a complete sample set. The total number of samples should not be less than 200 sets, with the training set and test set divided in a ratio of approximately 7:3.

[0105] Step 6.2: Model Architecture Design: A multi-output SVM architecture is adopted, using three parallel SVM sub-models to predict the coal-rock interface location (regression model), interbedded rock layer position (classification model), and interbedded rock thickness (regression model), respectively. The core theory is based on statistical learning theory, achieving the mapping between feature vectors and output labels by finding the optimal classification hyperplane.

[0106] Optimal classification hyperplane formula (applicable to rock interbedded layer classification): ,in: is the 21-dimensional hyperplane normal vector (with the same dimension as the input feature vector), x is the 21-dimensional signal feature vector extracted in step 5, and b is the bias term.

[0107] Model objective function (minimize classification error and model complexity): Where: C is the penalty coefficient (controlling the trade-off between classification error and model complexity). is a slack variable (allowing a small number of samples for classification error to improve the model's generalization ability), and N is the total number of samples.

[0108] Step 6.3: Kernel Function Selection (Adapting to High-Dimensional Feature Vectors): The Radial Basis Function (RBF) is used to handle the nonlinear mapping problem of high-dimensional feature vectors. The kernel function formula is: ,in: denoted as 21-dimensional feature vectors for different samples, and g is the kernel function parameter (controlling the influence range of the kernel function). Let be the Euclidean distance between the feature vectors of the two samples.

[0109] Step 6.4: Model Optimization Training: With 5-fold cross-validation accuracy as the optimization objective, the penalty coefficient C and kernel function parameter g are optimized using a grid search method to ensure the model's recognition accuracy and generalization ability. Search range (adapting to high-dimensional features and coal and rock identification needs): , Calculate the accuracy of the 5-fold cross-validation for all parameter combinations, and select the parameter combination C and g with the highest accuracy.

[0110] Optimize the objective function (accuracy of 5-fold cross-validation): ;

[0111] Optimal parameter solution: By traversing all parameter combinations, calculating the corresponding cross-validation accuracy, and selecting the parameter combination with the highest accuracy: , .

[0112] Training process: Input the training set into the SVM model, substitute the optimal parameters C∗ and g∗, map the 21-dimensional feature vectors to a high-dimensional feature space through kernel function mapping, solve for the optimal hyperplane, and complete the model training. Training convergence condition: During training, the convergence condition is set to ensure that the sum of slack variables is controllable, thereby avoiding overfitting and ensuring generalization ability.

[0113] Step 6.5: Model Validation and Deployment: Input the test set feature vectors into the trained model, calculate the recognition accuracy, and validate the model's precision. Validation formula:

[0114] .

[0115] The optimized SVM model is deployed to the DSP controller, and online real-time identification of coal and rock conditions is achieved through a process of real-time acquisition, preprocessing, feature extraction, model input, and classification decision. Specifically, the 21-dimensional feature vector extracted in step 5 is input into the model, and after mapping by a kernel function, a classification hyperplane made a classification decision, outputting the coal and rock type and interface offset. The complete process of online identification of coal and rock properties is as follows: Figure 7 As shown.

[0116] Specifically, this is achieved through the following kernel function mapping: ;

[0117] Classification Decision: ;

[0118] in: For support vector coefficients, Labels for the training set samples;

[0119] Output solution:

[0120] Coal and rock type: according to Output (1 = pure coal, 3 = interbedded with gangue, 5 = hard rock; other coal and rock types can be extended with reference to this classification system, without exceeding the scope of protection of this invention).

[0121] Coal-rock interface location: The interface offset is calculated based on the correspondence between the peak position in the feature vector and the sensor installation position.

[0122] Step 7: Calculate the dynamic height adjustment amount using fuzzy control.

[0123] This step constructs a fuzzy control model, using the coal and rock property identification results from step 6 and the real-time acquisition signals from step 5 as inputs, to dynamically calculate the adjustment amount. Details are as follows:

[0124] Step 7.1: Define control variables: Construct a fuzzy control model with 4 input variables and 2 output variables.

[0125] Input variables: Coal and rock condition identification result X1, which comes from step 6 and includes the location of the coal and rock interface, the position and thickness of the interbedded rock layer, represented by quantization values ​​1-5, corresponding to pure coal, thin interbedded rock, medium interbedded rock, thick interbedded rock, and hard rock respectively; Cutting tool group load fluctuation coefficient X2, which comes from step 5 and is represented by the ratio of the real-time acquired signal to the mean value within the 50% range; Cutting depth X3, which ranges from 0.1 m to 0.3 m; Cutting speed X4, which ranges from 1 m / s to 3 m / s.

[0126] Output variables: Upper planer group height adjustment U1, range -10~+50mm; Middle planer group height adjustment U2, range -10~+30mm. Bottom planer group remains unchanged.

[0127] Control logic: The input is fuzzified → rule reasoning → defuzzified, and the outputs are continuously adjustable U1 and U2, which drive the actuator to achieve dynamic adjustment, so that the load fluctuation coefficient is stabilized within the preset threshold (e.g., ≤30%).

[0128] Step 7.2: Definition of membership function: The definitions of the universe of discourse and fuzzy subsets of each variable are shown in Table 2.

[0129] Table 2. Definitions of the universe of discourse and fuzzy subsets of fuzzy control variables

[0130] <![CDATA[X1 Coal Rock Working Condition]]> 1~5 1~5 Coal (1), interbedded gangue (3), hard rock (5) <![CDATA[X2 Load fluctuation coefficient]]> 0~50% 0~50 Small (S), Medium (M), Large (L) <![CDATA[X3 Depth of cutting]]> 0.1~0.3 m 0.1~0.3 Light (Q), Medium (Z), Dark (S) <![CDATA[X4 Planing Speed]]> 1~3 m / s 1~3 Slow (M), Medium (Z), Fast (K) <![CDATA[U1 boost amount]]> -10~+50 mm -10~+50 Decrease (D), Remain unchanged (Z), Slight increase (XU), Significant increase (DU) <![CDATA[U2 boost volume]]> -10~+50 mm -10~+30 Decrease (D), Remain unchanged (Z), Slight increase (XU), Significant increase (DU)

[0131] The mapping from precise input to fuzzy linguistic values ​​is achieved using a triangular membership function:

[0132] ,

[0133] Where a, b, and c are the left boundary, peak center, and right boundary of the fuzzy subset, respectively. For any input x, the membership degree μ∈[0,1] of each fuzzy subset can be calculated.

[0134] Step 7.3: Establishment of Fuzzy Control Rules: Based on the data analysis conclusions from the orthogonal experiment in Step 2 and the joint simulation in Step 3, and combined with engineering practice experience, a core fuzzy control rule base is constructed, as shown in Table 3. Table 3 presents five typical control rules. The complete rule base contains the aforementioned rules, covering the combinations of various input variables.

[0135] Rule 1: If X1 = coal and X2 = small, then U1 = remain and U2 = remain. This rule applies to stable operating conditions with pure coal and requires no significant adjustment.

[0136] Rule 2: If X1 = interbedded with rock and X2 = in the middle, then U1 = slightly increased (+15mm) and U2 = significantly increased (+25mm). This rule applies to the interbedded rock condition in the middle, prioritizing raising the middle planer to balance the load.

[0137] Rule 3: If X1 = hard rock and X2 = large, then U1 = significantly increased (+35mm) and U2 = moderately increased (+15mm). This rule applies to the hard rock condition of the roof plate 7, where the upper planer should be raised first to avoid impact.

[0138] Rule 4: If X3 = deep and X4 = fast, then U1 = medium upward adjustment (+20mm) and U2 = large upward adjustment (+25mm). This rule applies to deep planing and high-speed conditions, requiring the tool to be lifted earlier to prevent impact.

[0139] Rule 5: If X2 = small and X1 = coal, then U1 = hold and U2 = hold. This rule corresponds to a continuously stable load state, locking the parameter increase to maintain operational stability.

[0140] Table 3 Core Fuzzy Control Rules

[0141] 1 <![CDATA[IF X1 = coal AND X2 = small THEN U1 = hold, U2 = hold]]> Stable operating conditions for pure coal, with no need for significant adjustments. (0,0) 2 <![CDATA[IF X1=Intercalated Gangue AND X2=Medium THEN U1=Small Increase, U2=Large Increase]]> In the case of interbedded rock, priority should be given to raising the central balancing load. (+15,+25) 3 <![CDATA[IF X1 = Hard Rock AND X2 = Large THEN U1 = Substantially Increased, U2 = Moderately Increased]]> In hard rock conditions with a roof, priority should be given to lifting the upper part to avoid impact. (+35, +15) 4 <![CDATA[IF X3 = Deep AND X4 = Fast THEN U1 = Medium Increase, U2 = Large Increase]]> Deep planing + high-speed operation, early tool lifting to prevent impact. (+20, +25) 5 <![CDATA[IF X2 = small AND X1 = coal THEN U1 = hold, U2 = hold]]> The load remained stable, and the lock-up height parameter was adjusted. (0, 0)

[0142] Step 7.4: Fuzzy Inference: Calculate the activation strength of each rule using the Mamdani minimum method. In the formula: Let be the activation weight of the i-th rule, with a value ranging from [0,1]. All activation rules are synthesized, and then the fuzzy output is defuzzified into a precise output value using the centroid method. Where: U is the final solution U1 or U2, The center value of the output fuzzy subset corresponding to this rule (e.g., keep = 0, slightly increase = +15, significantly increase = +35, etc.); n is the number of currently active valid rules.

[0143] Step 7.5: Control strategy output: Output the calculated height adjustment amounts U1 and U2 to the height adjustment drive mechanism, drive the servo motor (or hydraulic servo system) to drive the planer head to achieve real-time dynamic adjustment, so that the load fluctuation coefficient of the planer group is stabilized within the preset threshold (i.e., load fluctuation coefficient ≤ 30%).

[0144] Step 8: Closed-loop iterative verification and parameter optimization.

[0145] This step involves constructing an integrated planing test platform combining a hybrid planer blade group and an optimal height adjustment mechanism, applying the fuzzy control dynamic height adjustment strategy from step 7 to the test system. Details are as follows:

[0146] Step 8.1: Test system setup: Install the planer group according to the mixed planer arrangement scheme in Step 1, construct the height adjustment actuator according to the optimal height adjustment drive form determined in Step 4, configure the sensor system described in Step 5, and deploy the fuzzy control algorithm in Step 7 to the controller to form a complete test system.

[0147] Step 8.2: Working Condition Simulation Test: Simulate the actual thin coal seam planing conditions, including complex conditions such as roof and floor undulations of 0-50mm and coal-rock interbedded distribution (rock thickness varies from 10-200mm). Run the test system to conduct continuous planing tests. During the test, record data such as the planer load fluctuation coefficient, coal breaking energy consumption, and tool wear (characterized by periodically measuring the wear of the planer cutting edge) in real time.

[0148] Step 8.3: Target Determination: Compare the collected data with preset targets. The preset targets are: load fluctuation coefficient of the planer group ≤ 30%, coal breaking energy consumption ≤ 0.4 kW·h / m³, and single-cycle wear of the tool ≤ preset threshold. If the load fluctuation coefficient exceeds 30% or other targets are not met, the current control strategy is deemed to have failed.

[0149] Step 8.4: Closed-loop feedback correction: If the preset target is not achieved, the experimental data is fed back to the co-simulation platform to correct the simulation model parameters or control strategy parameters (including SVM identification model parameters, fuzzy control rule thresholds, etc.). After correction, the data is re-entered into the experiment for verification, and the process is repeated iteratively until all indicators meet the preset target.

[0150] Step 8.5: Final Output: Through a closed-loop iterative process of initial parameter input, control strategy execution, experimental data acquisition, target determination, and model correction, as follows... Figure 8 As shown, the SVM coal and rock identification model, fuzzy control rules, and cutter arrangement parameters were continuously optimized to ultimately stabilize the load fluctuation coefficient of the cutter group within the target range of ≤30%. When the target requirements are met in three consecutive tests, the current parameter combination and control strategy are output as the optimal control strategy and parameter configuration.

Claims

1. An automatic height adjustment control method for a coal planer, characterized in that, Includes the following steps: Planer blade group layout and parameter optimization: The planer blade group is arranged in sections, with the waist blade group using a mixed arrangement of pick-type and blade-type cutting teeth. The planer blade layout parameters and height response speed are optimized through orthogonal experiments and joint simulation. Multi-source signal sensing and coal and rock identification: Collect multi-source signals during the planing process, extract signal feature vectors, and identify coal and rock properties online based on a support vector machine model; Fuzzy dynamic height adjustment and control execution: Taking the coal and rock identification results, load fluctuation coefficient and planing parameters as input, the real-time height adjustment of the planing tool group is calculated through fuzzy control and output to the height adjustment execution mechanism; Closed-loop iterative verification: The control strategy is applied to the test platform, and the model and parameters are corrected through feedback from the test data, iterating until the preset target is met.

2. The automatic height adjustment control method for a coal planer according to claim 1, characterized in that, In the partitioned arrangement of the planer blade group, both the top blade group and the bottom blade group adopt pick-type cutting teeth, the proportion of pick-type cutting teeth in the waist blade group is set to 60% to 80%, and the mixed spacing of the planer blades is set to 60mm to 100mm.

3. The automatic height adjustment control method for a coal planer according to claim 1, characterized in that, The parameter optimization specifically includes: using L9 orthogonal experiments to screen for better parameter combinations, and using the EDEM-ANSYS co-simulation platform to correct the proportion of pick-type cutting teeth in the waist knife group, the mixed spacing of the planer blades, and the height response speed to determine the optimized parameter set.

4. The automatic height adjustment control method for a coal planer according to claim 1, characterized in that, The multi-source signals include planer load signal, lower planer body force signal, and mechanism vibration acceleration signal; the signal feature vector includes time domain features and frequency domain features. Preprocessing of multi-source signals includes wavelet decomposition, improved soft threshold denoising, and low-pass filtering.

5. The automatic height adjustment control method for a coal planer according to claim 4, characterized in that, The time-domain features include mean, variance, peak value, and kurtosis, and the frequency-domain features include dominant frequency, spectral energy, and frequency band distribution.

6. The automatic height adjustment control method for a coal planer according to claim 1, characterized in that, The support vector machine model adopts a multi-output architecture, predicting the coal-rock interface location, interbedded rock layer position, and interbedded rock thickness through three parallel sub-models. It uses a radial basis kernel function and optimizes the penalty coefficient C and kernel function parameter g through a grid search method.

7. The automatic height adjustment control method for a coal planer according to claim 1, characterized in that, In the fuzzy control, the Mamdani minimum method is used to calculate the activation intensity of the control rule, and the centroid method is used to defuzzify the fuzzy output into a precise adjustment amount, so that the load fluctuation coefficient of the planer group is stabilized within the target range of ≤30%.

8. The automatic height adjustment control method for a coal planer according to claim 1, characterized in that, The real-time height adjustment is calculated only for the upper and middle planer groups, while the bottom planer group remains unchanged.

9. The automatic height adjustment control method for a coal planer according to claim 1, characterized in that, The preset targets for the closed-loop iterative verification include: the load fluctuation coefficient of the cutter group ≤30% and the coal breaking energy consumption ≤0.4kW·h / m³.

10. The automatic height adjustment control method for a coal planer according to claim 1, characterized in that, The method also includes a height adjustment mechanism selection and adaptation step: adopting a combination drive scheme of ball screw and servo motor or a combination drive scheme of hydraulic cylinder and hydraulic servo system, and cooperating with dovetail precision guide rail, and determining the optimal drive form through response speed, positioning accuracy, load-bearing capacity and energy consumption tests.

Images