Intelligent speed control method for high-wear-resistance scraper conveyor

Abstract

The application discloses a kind of high wear-resistant scraper conveyor intelligent speed regulation control methods, it is related to mine machinery intelligent control technical field, including the real-time operating state data of acquisition scraper conveyor drive motor current, scraper chain tension, machine head and tail speed difference, and the coal flow characteristic data of instantaneous coal quantity, coal block granularity distribution, gangue content ratio.Extract fault characteristics to operating state data, generate health degree index including chain wear state, groove wear rate, drive system efficiency.According to the nonlinear relationship of historical coal flow data and working face advancing speed, an improved prediction model is constructed to predict the change of coal load. Fusion health index and load prediction results to develop speed regulation strategy, combined with real-time operating deviation to introduce feedback regulation to correct speed and liquid volume curve, closed-loop control command is issued. The method can reduce the wear rate of scraper conveyor components, improve the speed regulation response accuracy, and adapt to the high wear condition requirements of equipment operation.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control technology for mining machinery, specifically a method for intelligent speed regulation control of a high wear-resistant scraper conveyor. Background Technology

[0002] Existing scraper conveyor speed control primarily relies on real-time coal quantity or drive motor current as the core basis for regulation. Some solutions only monitor a single operating parameter for speed adjustment. Coal flow load prediction often uses linear data fitting, relying solely on historical coal flow data for load forecasting without incorporating the face advance speed into the prediction model. Equipment wear condition monitoring is mostly independent of the speed control system, lacking a correlation mechanism between wear parameters and speed control strategies. The regulation of hydraulic coupling filling volume and main drive motor speed is often based on temporary adjustments according to real-time load, without closed-loop feedback regulation logic.

[0003] Single-parameter control modes cannot fully reflect the wear status of the scraper conveyor chain and trough, as well as the actual efficiency of the drive system. Under high wear-resistant conditions, the wear rate of components continues to accelerate, and the speed regulation action is poorly adapted to the equipment wear status. The linear coal flow prediction model cannot match the nonlinear correlation between historical coal flow and working face advance speed, resulting in a large deviation in coal load prediction. The speed regulation response lags behind changes in coal flow, further aggravating the frictional wear of the scraper chain and trough. The control of drive system speed and liquid volume lacks multi-dimensional data support, and real-time operational deviations cannot be corrected in a timely manner, making it difficult for speed regulation strategies to meet the operational requirements of high wear-resistant conditions.

[0004] Fault features need to be extracted based on various types of real-time operating status data to form quantitative indicators of core wear parameters and drive efficiency of the corresponding equipment. A predictive model that fits the nonlinear relationship between historical coal flow and working face advance speed needs to be constructed to achieve accurate prediction of coal load changes. Based on the equipment health status and load prediction results, a speed regulation strategy adapted to high wear resistance constraints should be formulated. The speed regulation parameters should be dynamically corrected through feedback adjustment to form a closed-loop control. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art;

[0006] Therefore, this invention proposes an intelligent speed control method for a high wear-resistant scraper conveyor, comprising:

[0007] The system collects real-time operating status data of the scraper conveyor and coal flow characteristic data of the working face. The real-time operating status data includes drive motor current, scraper chain tension, and speed difference between the head and tail of the conveyor. The coal flow characteristic data of the working face includes instantaneous coal quantity, coal particle size distribution, and gangue content ratio.

[0008] Fault features are extracted from the real-time operating status data to generate a scraper conveyor health index, which includes chain wear status, trough wear rate, and drive system efficiency.

[0009] Based on the coal flow characteristic data of the working face, an improved coal flow prediction algorithm is used to predict future changes in coal load. The improved coal flow prediction algorithm establishes a prediction model based on the nonlinear relationship between historical coal flow data and working face advance speed.

[0010] The health index of the scraper conveyor and the prediction results of the coal load change are subjected to multi-objective fusion analysis to formulate a speed regulation strategy that meets the high wear resistance constraint. The speed regulation strategy includes the target speed curve of the main drive motor and the target liquid filling volume curve of the hydraulic coupler.

[0011] Based on the deviation between the speed regulation strategy and real-time operation, a feedback adjustment mechanism is introduced to dynamically adjust the target speed curve and the target liquid filling volume curve, generate a closed-loop control command, and send it to the drive system of the scraper conveyor.

[0012] Furthermore, fault features are extracted from the real-time operating status data to generate a health index for the scraper conveyor, including:

[0013] Harmonic analysis is performed on the current of the drive motor to extract the amplitude of the characteristic harmonic components in the current and calculate the load imbalance of the drive motor.

[0014] The tension of the scraper chain is analyzed in both the time and frequency domains to calculate the average value, standard deviation, and main frequency fluctuation range of the tension signal, thereby obtaining the chain wear characteristic parameters.

[0015] The tension offset state of the scraper conveyor chain is calculated by comparing the relationship between the speed difference between the head and tail of the machine and a preset threshold.

[0016] Vibration acceleration signals at key locations in the scraper conveyor trough were collected, and envelope demodulation analysis was performed on the vibration signals to extract the characteristic frequency amplitudes reflecting the wear between the trough and the middle plate.

[0017] Combining the drive motor load imbalance, chain wear characteristic parameters, tension offset state, and characteristic frequency amplitude, a pre-trained health assessment model is used to calculate and output a quantified health index of the scraper conveyor. The chain wear state is one dimension in the output result of the health assessment model.

[0018] Furthermore, based on the coal flow characteristic data of the working face, an improved coal flow prediction algorithm is used to predict future changes in coal load, including:

[0019] Data on instantaneous coal quantity, coal particle size distribution, and gangue content ratio within a historical time period are obtained, along with data on the working face coal mining machine speed and hydraulic support push distance for the corresponding time period, to form a multi-source coal flow time series.

[0020] The multi-source coal flow time series is cleaned and aligned to eliminate noise and outliers, and then standardized to generate standardized multi-source coal flow time series data.

[0021] The standardized multi-source coal flow time series data is input into the improved coal flow prediction algorithm. The improved coal flow prediction algorithm uses a long short-term memory network to capture the long-term dependencies of the time series and uses an attention mechanism to dynamically weight the impact of different input features on future load.

[0022] The improved coal flow prediction algorithm outputs a predicted sequence of coal load within a specific future time window, the predicted sequence including the trend of coal load change and the time point of peak occurrence.

[0023] Furthermore, the working principle of the improved coal flow prediction algorithm is as follows:

[0024] A model encoder containing multi-layer long short-term memory network units is constructed to encode the input standardized multi-source coal flow time series data and extract deep temporal features of the time series.

[0025] An attention mechanism layer is connected to the top layer of the model encoder. The attention mechanism layer receives the feature vectors of each time step output by the model encoder and calculates the attention weights of the feature vectors of each historical time step at the future prediction time.

[0026] The calculated attention weights are weighted and summed on the corresponding historical time step feature vectors to generate a context vector containing global temporal information.

[0027] The context vector is input into a decoder composed of a multilayer sensing mechanism, and the decoder outputs the predicted coal load values ​​for each future prediction time point.

[0028] During model training, an adversarial training strategy is introduced, which generates simulated coal flow data through a generative adversarial network to enhance the generalization prediction ability of the improved coal flow prediction algorithm for abnormal working conditions.

[0029] Furthermore, a multi-objective fusion analysis is performed on the health index of the scraper conveyor and the predicted results of the coal load change to formulate a speed regulation strategy that meets the high wear resistance constraint, including:

[0030] Based on the chain wear state and the trough wear rate, a maximum allowable wear rate threshold for the scraper chain and the central trough is set, and equipment wear safety constraints are generated based on this threshold.

[0031] Based on the predicted results of the coal load change, the minimum operating speed required to meet the coal transportation efficiency requirements is calculated, and production efficiency constraints are generated.

[0032] Construct a multi-objective optimization problem with the goal of maximizing the efficiency of the drive system, minimizing the peak fluctuation of the chain tension, and simultaneously satisfying the equipment wear safety constraints and production efficiency constraints;

[0033] The multi-objective optimization problem is solved using a constrained multi-objective optimization algorithm. The solution results include a series of non-dominated solutions, each of which corresponds to a set of speed regulation parameters.

[0034] Based on the decision-making strategy, an optimal solution is selected from the set of non-dominated solutions, and the speed regulation parameters in the optimal solution are analyzed into the target speed curve of the main drive motor and the target fluid volume curve of the hydraulic coupler.

[0035] Furthermore, the step of solving the multi-objective optimization problem using a constrained multi-objective optimization algorithm includes:

[0036] Initialize a population containing multiple individuals, each representing a set of codes for the speed regulation parameters;

[0037] Decode each individual to obtain the corresponding speed regulation scheme, and calculate the wear constraint violation degree according to the equipment wear safety constraint condition, and calculate the efficiency constraint violation degree according to the production efficiency constraint condition;

[0038] For individuals that meet the constraints, their fitness value is calculated based on the drive system efficiency value and chain tension peak fluctuation value corresponding to the individual.

[0039] Selective, crossover, and mutation operations are performed on the population to generate new offspring populations. During the operation, the degree of constraint violation of individuals is evaluated to guide the population to evolve towards the feasible domain.

[0040] Through multiple generations of iteration, when the convergence condition is met, all non-dominated feasible individuals are selected from the final population to form the non-dominated solution set.

[0041] Furthermore, based on the deviation between the speed regulation strategy and real-time operation, a feedback adjustment mechanism is introduced to dynamically adjust the target speed curve and the target filling volume curve, including:

[0042] During the speed regulation strategy execution cycle, continuously monitor the actual operating speed, actual drive motor current and actual chain tension of the scraper conveyor;

[0043] The actual operating speed is compared with the expected operating speed given by the target speed curve, and the speed tracking error is calculated.

[0044] The actual drive motor current is compared with the expected current calculated based on the target speed curve and the predicted coal load, and the load matching error is calculated.

[0045] The actual chain tension is compared with the tension safety range set based on health indicators and operating speed to calculate the tension overshoot error;

[0046] The speed tracking error, load matching error, and tension overshoot error are input to a multi-input multi-output fuzzy controller, which outputs the speed correction amount for the target speed curve and the filling amount correction amount for the target filling amount curve.

[0047] The speed correction and the liquid filling correction are respectively superimposed on the current target speed curve and the target liquid filling curve to generate an updated closed-loop control command.

[0048] Furthermore, the working principle of the multi-input multi-output fuzzy controller includes:

[0049] Define input fuzzy variables, including speed tracking error, load matching error and tension overshoot error, and define multiple fuzzy subsets and corresponding membership functions for each input fuzzy variable;

[0050] Define output fuzzy variables, including speed correction and filling volume correction, and define multiple fuzzy subsets and corresponding membership functions for each output fuzzy variable;

[0051] Based on expert experience and a control rule base, fuzzy rules are established from input fuzzy variables to output fuzzy variables. The fuzzy rules describe the relationship between input state and output action in the form of "if...then...".

[0052] In each control cycle, the values ​​of speed tracking error, load matching error and tension overshoot error, which are collected and calculated in real time, are fuzzified through their respective membership functions to obtain the membership degree of the fuzzy subset corresponding to each input fuzzy variable.

[0053] Using the aforementioned fuzzy rules and fuzzy inference method, the activation intensity of the fuzzy subset corresponding to each output fuzzy variable is calculated;

[0054] The membership functions of the output fuzzy variables are synthesized and defuzzified using the centroid method to finally obtain the rotational speed correction and the filling volume correction.

[0055] Further, the step of weighting the calculated attention weights onto the corresponding historical time step feature vectors to generate a context vector containing global temporal information includes:

[0056] At the top layer of the model encoder, the feature vector output by the last long short-term memory network unit at the last historical time step is obtained as the encoding state at the current moment;

[0057] In the attention mechanism layer, the feature vectors output by all long short-term memory network units in the model encoder at each historical time step are obtained to form a set of historical feature vectors.

[0058] For each time point to be predicted in the future prediction sequence, a query vector is calculated in the attention mechanism layer. The query vector is formed by concatenating the encoding state of the current time and the position encoding of the time point to be predicted after a linear transformation.

[0059] The relevance score is calculated between the query vector and each historical feature vector in the set of historical feature vectors. The relevance score is calculated through a trainable multilayer perceptron network.

[0060] The correlation scores of each historical feature vector are input into the Softmax function for normalization to obtain the attention weight corresponding to each historical feature vector, and the sum of the attention weights of all historical feature vectors is 1.

[0061] Multiply each historical feature vector in the set of historical feature vectors by its corresponding attention weight to obtain a weighted feature vector.

[0062] All weighted feature vectors are summed to generate a fixed-dimensional context vector for the time point to be predicted. The context vector incorporates the time series information most relevant to the time point to be predicted from all historical time steps.

[0063] Furthermore, the method also includes a simulation verification step before the speed regulation strategy is executed:

[0064] Before the speed regulation strategy is sent to the drive system for execution, the speed regulation strategy is simulated and verified based on a high-fidelity digital twin model of the scraper conveyor.

[0065] The digital twin model includes dynamic models of the mechanical transmission system, drive motor, hydraulic coupling, and load;

[0066] The target speed curve and target fluid volume curve are run in the digital twin model to simulate the simulated running speed, simulated motor current, simulated chain tension, and simulated wear rate.

[0067] The simulation results are compared with the set safety thresholds and efficiency thresholds. If the simulation results exceed the safety thresholds or fail to reach the efficiency thresholds, the speed regulation strategy is automatically redefined. Only after the simulation verification is passed will the corresponding speed regulation strategy command be sent to the drive system of the physical scraper conveyor for execution.

[0068] Compared with the prior art, the beneficial effects of the present invention are:

[0069] Real-time operating status data such as drive motor current, scraper chain tension, and speed difference between the head and tail of the conveyor are collected. After extracting fault features from the data, a health index for the scraper conveyor is generated, including chain wear status, trough wear rate, and drive system efficiency. This index can intuitively present the wear level of the core wear-prone components of the scraper conveyor and the operating efficiency of the drive system. During the speed regulation strategy formulation process, it can directly match the constraints of high wear-resistant working conditions, avoid the mismatch between component wear and speed regulation caused by single operating parameter adjustment, reduce additional frictional contact between the scraper chain and the trough during speed regulation, maintain the operating state of the drive system in accordance with the equipment wear state, reduce the aggravation of component wear caused by unreasonable speed regulation, and keep the operating state of the drive system synchronized with the actual wear level of the equipment.

[0070] A predictive model is established based on the nonlinear relationship between historical coal flow data and the face advance speed. An improved coal flow prediction algorithm is used to predict future coal load changes. The predicted coal load can closely match the nonlinear working conditions of the actual face advance, reducing the deviation between the load prediction and the actual coal flow. This allows for advance adaptation of speed regulation parameter settings corresponding to changes in coal load, avoiding coal flow impact caused by speed regulation response lag and reducing component friction wear caused by impact. Integrating health indicators and load prediction results forms a multi-objective analysis basis. A feedback adjustment mechanism is introduced based on real-time operating deviations to dynamically adjust the target speed curve of the main drive motor and the target fluid volume curve of the hydraulic coupler. Closed-loop control commands can be precisely applied to the drive system, ensuring that speed regulation continuously matches the high-wear-resistant operating conditions and that adjustments to speed and fluid volume always adapt to the predicted coal load changes. Attached Figure Description

[0071] Figure 1 This is a flowchart illustrating the steps of an intelligent speed control method for a high wear-resistant scraper conveyor according to the present invention.

[0072] Figure 2 A flowchart for generating health indicators for scraper conveyors;

[0073] Figure 3 A flowchart illustrating the work done on the improved coal flow prediction algorithm. Detailed Implementation

[0074] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0075] See Figure 1 This invention provides an intelligent speed control method for a high wear-resistant scraper conveyor, the specific method including:

[0076] The system collects real-time operational status data and coal flow characteristic data from the working face through a sensor system deployed on the scraper conveyor. Real-time operational status data includes drive motor current, scraper chain tension, and the speed difference between the head and tail of the conveyor; coal flow characteristic data includes instantaneous coal quantity, coal particle size distribution, and gangue content. The system extracts fault features from the real-time operational status data to generate a scraper conveyor health index that quantitatively characterizes the equipment's condition. This index specifically includes chain wear status, trough wear rate, and drive system efficiency. Simultaneously, based on the working face coal flow characteristic data, the system uses an improved coal flow prediction algorithm to establish a prediction model to predict changes in coal load within a specific future time window. The scraper conveyor health index and the predicted coal load change results are input into a multi-objective fusion analysis module. Under the premise of meeting high wear resistance constraints, a speed regulation strategy is formulated, including the target speed curve of the main drive motor and the target fluid volume curve of the hydraulic coupler. Based on the deviation between the speed regulation strategy and real-time operating data, the system introduces a feedback adjustment mechanism to dynamically adjust the target speed curve and the target liquid filling volume curve, generate closed-loop control commands and send them to the drive system of the scraper conveyor, thereby achieving precise control of the equipment operation process.

[0077] In one embodiment of the present invention, during the fault feature extraction process, refer to... Figure 2Harmonic analysis is performed on the collected drive motor current signal to extract the amplitude of characteristic harmonic components of specific orders, and the load imbalance of the drive motor under different operating conditions is calculated accordingly. For the scraper chain tension signal, simultaneous time-domain and frequency-domain joint analysis is conducted. By calculating the average value, standard deviation, and fluctuation range of the dominant frequency component of the tension signal, characteristic parameters reflecting the degree of chain wear are accurately obtained. By continuously monitoring the speed difference between the head and tail drive shafts and comparing this difference with a preset safety threshold, the tension deviation state of the scraper conveyor chain can be calculated in real time. Furthermore, vibration acceleration sensors are installed at key stress locations in the scraper conveyor trough to collect high-frequency vibration signals and perform envelope demodulation analysis, effectively extracting the characteristic frequency amplitudes characterizing the wear of the trough and the middle plate. All of the above characteristic parameters—drive motor load imbalance, chain wear characteristic parameters, tension deviation state, and characteristic frequency amplitudes—are uniformly input into a pre-trained deep learning health assessment model. The model is trained on a large amount of historical fault data and can perform nonlinear mapping and fusion calculations on these multidimensional features, ultimately outputting a quantitative health index of the scraper conveyor. The chain wear status is an important dimension output by the model.

[0078] In practical implementation, the intelligent speed control system of the scraper conveyor continuously collects the three-phase drive motor current waveform data at a sampling frequency of no less than 1kHz using current sensors installed in the drive motor power supply circuit. Harmonic analysis is performed on the collected drive motor current, and a Fast Fourier Transform is used to convert the current signal from the time domain to the frequency domain, extracting the amplitude of characteristic harmonic components of specific orders, such as specific high-frequency components corresponding to rotor bar breakage or eccentricity faults. The drive motor load imbalance is calculated based on the amplitude of the characteristic harmonic components, using the following formula:

[0079]

[0080] in: Indicates the degree of load imbalance in the drive motor. This indicates the total number of selected characteristic harmonic orders. Indicates the harmonic order index. Indicates the first Weighting coefficients for subharmonics Indicates the first The absolute value of the amplitude of the second characteristic harmonic component. This represents the amplitude of the fundamental current. This formula quantifies the operating state of the drive motor under three-phase imbalance or sudden mechanical load changes.

[0081] In practice, tension sensors installed on the scraper chain monitor the chain force in real time, with a sampling frequency set to 100Hz. The scraper chain tension signal is subjected to joint time-domain and frequency-domain analysis. In the time domain, the arithmetic mean and standard deviation of the tension signal over a complete operating cycle are calculated. In the frequency domain, power spectral density analysis is used to identify the dominant frequency component of the tension signal, and the fluctuation range of the dominant frequency peak is statistically analyzed. The time-domain mean and standard deviation reflect the average level and dispersion of the chain force, while the frequency-domain dominant frequency fluctuation range reflects the periodic impact characteristics during the chain-sprocket meshing process. These combined parameters yield the chain wear characteristic parameters.

[0082] In practice, photoelectric encoders installed on the head and tail drive shafts measure the rotational speed and calculate the difference between the head and tail rotational speeds per unit time. This head-to-tail speed difference is compared to a preset threshold, which is the maximum allowable speed deviation calculated based on the scraper conveyor's design tension and the chain's elastic modulus. When the measured speed difference exceeds the preset threshold, it indicates a shift in the scraper chain tension. The system records the direction and magnitude of this shift for subsequent calculations of the tension shift status.

[0083] In practical implementation, triaxial vibration acceleration sensors are placed at key locations on the middle plate and bottom plate of the scraper conveyor trough to collect raw vibration signals with a frequency of not less than 5kHz. Envelope demodulation analysis is performed on the vibration acceleration signals. First, the raw signal is bandpass filtered to retain the characteristic frequency bands related to the wear of the trough and middle plate. Then, the envelope of the signal is extracted using Hilbert transform, and spectral analysis is performed on the envelope to extract the characteristic frequency amplitudes reflecting the wear of the trough and middle plate. These characteristic frequencies correspond to the periodic excitation of friction, jamming, or impact between the scraper chain and the middle plate.

[0084] In practical implementation, the calculated drive motor load imbalance, chain wear characteristic parameters, tension offset state, and characteristic frequency amplitude are combined into a multi-dimensional feature vector, which is then input into a pre-trained health assessment model. The health assessment model employs a deep neural network structure, trained under supervised learning using extensive historical operating data and corresponding wear state labels. The model contains multiple hidden layers and nonlinear activation functions. The health assessment model performs nonlinear mapping and feature fusion on the input multi-dimensional feature vector, outputting a quantified scraper conveyor health index vector. This vector contains independent components for chain wear state, trough wear rate, and drive system efficiency. The chain wear state component directly reflects the percentage of the scraper chain's current wear relative to its initial state.

[0085] In one embodiment of the present invention, to achieve accurate prediction of coal load, multi-source heterogeneous data such as instantaneous coal quantity, coal particle size distribution, gangue content ratio, and corresponding working face coal mining machine operating speed and hydraulic support pushing step distance are acquired within a historical time period. These data are then aligned by timestamps to construct a multi-source coal flow time series. This time series undergoes data cleaning to remove outliers and noise interference, and standardization preprocessing is performed to generate standardized multi-source coal flow time series data that meets the model input requirements. This standardized data is then input into an improved coal flow prediction algorithm, as described in [reference missing]. Figure 3 The core architecture of this algorithm includes a multi-layered long short-term memory (LSTM) network encoder used to encode the input sequence, thereby capturing the deep temporal dependencies of coal flow changes. An attention mechanism layer is connected at the top of the encoder. This layer receives the feature vectors from each time step output by the encoder and calculates the attention weights for the feature vectors from each historical time step for the future prediction time. In calculating the attention weights, the output of the last LSM network unit in the encoder at the last historical time step is first obtained and used as the current encoding state. Simultaneously, the feature vectors output by all units of the encoder at each historical time step are collected to form a historical feature vector set. For each time point to be predicted in the future prediction sequence, a query vector is generated by linearly transforming and concatenating the current encoding state with the position encoding of that time point. A trainable multi-layered perceptron network is used to calculate the relevance score between this query vector and each vector in the historical feature vector set. The scores are then normalized to attention weights using a Softmax function, ensuring that the sum of all weights is 1. These weights are then used to perform a weighted summation of the historical feature vectors to generate a context vector that incorporates global relevant information. The vector is then fed into a decoder composed of a multilayer perceptron, which outputs predicted coal load values ​​for each future time point, forming a prediction sequence that includes the trend and peak time. During the offline training phase of the model, a generative adversarial network is introduced for adversarial training. By generating data samples simulating abnormal operating conditions, the generalization robustness of the prediction algorithm in real-world complex environments is enhanced.

[0086] In practical implementation, historical data used for coal load prediction comes from a sensor network deployed at the working face. The system acquires instantaneous coal quantity data sampled at the second level over the past 24 hours, histogram data of coal block particle size distribution obtained based on image recognition, data on the proportion of gangue content detected by an online elemental analyzer, and data on the coal mining machine's walking speed and hydraulic support's step distance at the corresponding timestamps. These multi-source heterogeneous data are aligned using a unified timestamp, missing values ​​are filled, and outliers significantly exceeding the physical range are removed, forming a multi-source coal flow time series of fixed length. The multi-source coal flow time series is then standardized using the Z-score method to convert the data of each feature dimension into a standard normal distribution with a mean of zero and a variance of one, generating standardized multi-source coal flow time series data for model input.

[0087] In practical implementation, standardized multi-source coal flow time series data are input into an improved coal flow prediction algorithm. The main structure of the improved coal flow prediction algorithm is a model encoder containing three layers of long short-term memory (LSM) network units. Each LSM network unit in the model encoder processes the input sequence step by step according to time steps, capturing long-term dependencies in the coal flow data through gating mechanisms of input gates, forget gates, and output gates, and finally outputting a hidden state vector containing deep temporal features. A multi-head attention mechanism layer is connected at the top of the model encoder, receiving the feature vector set of all historical time steps output by the model encoder.

[0088] In practical implementation, for each predicted time point in the future prediction sequence, the improved coal flow prediction algorithm performs a specific attention calculation process to generate a context vector. At the top layer of the model encoder, the feature vector output by the last long short-term memory network unit at the last historical time step is obtained and marked as the current encoding state. Simultaneously, feature vectors output by all long short-term memory network units in the model encoder at each historical time step are collected to form a historical feature vector set. For a given predicted time point, the current encoding state and the position encoding of the predicted time point are concatenated and projected using a linear transformation matrix to generate a query vector. A trainable small multilayer perceptron network is used to calculate the relevance score between the query vector and each historical feature vector in the historical feature vector set. The formula for calculating the relevance score is:

[0089]

[0090] in: Represents the query vector. This represents a specific historical feature vector within the set of historical feature vectors. This represents the vector concatenation operation. The parameter is Multilayer perceptron networks, This represents the calculated relevance score. All relevance scores are input into the Softmax function for normalization, yielding the attention weight for each historical feature vector, with the sum of all weights strictly equal to 1. Each vector in the historical feature vector set is multiplied by its corresponding attention weight and then summed to generate a fixed-dimensional context vector for the time point to be predicted. This context vector incorporates the most relevant global temporal information from the historical sequence to that future moment.

[0091] In practice, the generated context vector is fed into a multilayer perceptron decoder consisting of two fully connected layers. The first layer of the decoder uses the ReLU activation function to introduce nonlinearity, while the second layer is a linear output layer. The multilayer perceptron decoder maps the context vector to the predicted coal load at a specific future time point, outputting a predicted coal load sequence at second intervals for the next ten minutes. The peak points in the sequence correspond to the predicted high load times.

[0092] In some embodiments, the improved coal flow prediction algorithm incorporates a generative adversarial network (GAN) for adversarial training during the offline training phase. The GAN comprises a generator and a discriminator. The generator receives a random noise vector and historical real coal flow data fragments as input to generate realistic simulated coal flow time series data. The discriminator attempts to distinguish whether the input data comes from a real historical dataset or simulated data generated by the generator. The GAN engages in adversarial gameplay with the improved coal flow prediction algorithm, forcing it to learn more robust feature representations, thereby improving the model's generalization prediction ability when facing unseen abnormal operating conditions.

[0093] In one embodiment of the present invention, during the multi-objective fusion analysis stage, based on the chain wear status and trough wear rate output by the health assessment model, a maximum allowable wear rate threshold for both is set within the operating cycle, serving as a safety constraint for equipment wear to ensure the equipment's service life. On the other hand, based on the predicted coal load changes and combined with the conveying task requirements, the minimum operating speed of the main drive motor required to meet the predetermined production efficiency is calculated, generating a production efficiency constraint. Based on this, a multi-objective optimization problem is constructed, whose objective function aims to maximize the efficiency of the drive system and minimize the peak fluctuation of the chain tension, while simultaneously and strictly satisfying the aforementioned safety constraint for equipment wear and the production efficiency constraint. To solve this complex constrained multi-objective optimization problem, an improved evolutionary algorithm is used. This algorithm first initializes a population containing several individuals, each represented by an encoding of a set of potential speed regulation parameter combinations. After decoding each individual to obtain a specific speed regulation scheme, it evaluates whether the scheme violates the safety constraint for equipment wear and the production efficiency constraint, and calculates the corresponding constraint violation degree. For feasible individuals that meet the constraints, their fitness values ​​are further calculated. This calculation comprehensively considers the efficiency of the drive system and the magnitude of chain tension peak fluctuations for each individual's corresponding solution. Subsequently, the algorithm performs selection, crossover, and mutation operations on the population to generate a new generation of offspring. During the evolutionary process, constraint violation information guides the population to move towards feasible solution space regions. After multiple rounds of iterative evolution, when the algorithm reaches the preset convergence condition, all non-dominated feasible individuals are selected from the final population, forming a Pareto optimal solution set. Finally, based on a preset decision preference strategy, an optimal solution is selected from this non-dominated solution set, decoded, and parsed into directly executable target speed curves for the main drive motor and target fluid volume curves for the hydraulic coupler.

[0094] In practical implementation, the multi-objective fusion analysis module, based on the chain wear status value and trough wear rate value output by the health assessment model, sets the maximum allowable wear rate threshold for the scraper chain and the central trough within a single shift operating cycle, thereby generating equipment wear safety constraints that limit excessive equipment wear. Table 1 shows typical constraint thresholds set based on the health level.

[0095] Table 1: Equipment Wear Safety Constraint Thresholds

[0096] In practical implementation, based on the predicted sequence of coal load changes and combined with the shift production and conveying volume targets issued by the mine dispatch center, the minimum average operating speed of the scraper conveyor required to meet production efficiency requirements is calculated, generating production efficiency constraints. A multi-objective optimization problem is constructed, with the objective function defined as maximizing the efficiency of the drive system while minimizing the peak fluctuation of the chain tension, and the solution must simultaneously satisfy both the equipment wear safety constraint and the production efficiency constraint.

[0097] In practical implementation, a constrained multi-objective evolutionary algorithm with an elitist strategy is used to solve the above optimization problem. The algorithm first initializes a population of 100 individuals. Each individual in the population is encoded using real numbers, with gene positions representing the main drive motor speed setpoint, the hydraulic coupler filling volume setpoint, and the speed regulation time node, respectively. For each individual, a decoding operation is performed to reconstruct the specific speed regulation parameter combination, which is then substituted into the equipment wear model to calculate the wear rate and determine whether it exceeds the threshold in Table 1 to calculate the wear constraint violation degree. Simultaneously, it is substituted into the production efficiency model to calculate the average operating speed and determine whether it is lower than the minimum requirement to calculate the efficiency constraint violation degree. For feasible individuals that satisfy all constraints, their fitness value is calculated to evaluate their quality. The calculation formula is:

[0098]

[0099] in: This represents the fitness value of an individual. and This represents the preset positive weighting coefficient. This represents the normalized value of the drive system efficiency corresponding to the speed regulation scheme after decoding. This represents the peak chain tension fluctuation amplitude simulated under this scheme. The algorithm generates offspring populations through binary tournament selection, simulated binary crossover, and Gaussian mutation genetic operations. When handling constraints, a feasibility-first criterion is adopted to guide the search direction towards the feasible region. After 200 generations of iterative optimization, convergence is determined when the proportion of non-dominated solutions in the population tends to stabilize. All non-dominated feasible individuals are selected from the final population to form a Pareto optimal solution set. In specific implementation, the final optimal solution is selected from the Pareto optimal solution set according to a preset decision strategy. This strategy tends to select a compromise solution with higher drive system efficiency and smaller chain tension fluctuations. The selected optimal solution individual is decoded, and the speed regulation parameters represented by its gene loci are parsed out, mapping and generating a target speed curve of the main drive motor and a target fluid filling volume curve of the hydraulic coupler that vary with time.

[0100] In one embodiment of the present invention, during the execution of the speed regulation strategy, the system continuously monitors the actual operating speed of the scraper conveyor, the real-time drive motor current, and the actual scraper chain tension in each control cycle. The measured operating speed is compared with the desired speed given by the target speed curve, and the deviation between the two is calculated as the speed tracking error. The measured motor current is compared with the desired current predicted based on the current target speed and coal load to obtain the load matching error. Simultaneously, the measured chain tension is compared with the tension safety upper and lower limits set according to the health index to determine whether there is a tension overshoot error. These three types of error signals are input in real time to a multi-input multi-output fuzzy controller. The fuzzy controller internally defines three input fuzzy variables: speed tracking error, load matching error, and tension overshoot error. For each input variable, multiple fuzzy subsets are defined, and corresponding triangular or trapezoidal membership functions are designed. Simultaneously, two output fuzzy variables are defined: speed correction and filling volume correction, which are also assigned fuzzy subsets and membership functions. Based on the operational experience of domain experts, a comprehensive fuzzy rule base was established. The rules adopt the logical form of "if an input variable is in a certain state, then an output variable should be adjusted in a certain way." In each control cycle, the system first converts the collected sharp numerical errors into fuzzy quantities using a membership function, determining their membership degrees on each fuzzy subset. Then, using a fuzzy inference mechanism, the corresponding output rules are activated according to the rule base, and the activation strength of the output fuzzy subset is calculated. Finally, the centroid method is used to defuzzify the output fuzzy set, transforming it into precise speed correction and fluid filling correction values. These two correction values ​​are respectively superimposed onto the currently executing target speed curve and target fluid filling curve, thereby generating updated closed-loop control commands to achieve dynamic fine-tuning of the drive system.

[0101] In practical implementation, the intelligent speed control system for the scraper conveyor monitors the actual operating speed of the scraper conveyor in real time within each control cycle through an encoder installed in the drive system, monitors the actual drive motor current through a current transformer, and monitors the actual scraper chain tension through a tension sensor. The actual operating speed is compared with the expected operating speed at the current moment based on the target speed curve generated by multi-objective optimization to calculate the speed tracking error; the actual drive motor current is compared with the expected current calculated based on the current target speed and the predicted coal load sequence to calculate the load matching error; and the actual scraper chain tension is compared with the upper and lower limits of the tension safety range set according to the health index. If the tension exceeds the safety boundary, the tension exceedance error is calculated.

[0102] In practical implementation, speed tracking error, load matching error, and tension overrun error are input as clear numerical values ​​to the multi-input multi-output (MIMO) fuzzy controller. The MIMO fuzzy controller defines three input fuzzy variables: speed tracking error, load matching error, and tension overrun error. For speed tracking error, five fuzzy subsets are defined: "negative large," "negative small," "zero," "positive small," and "positive large," and triangular membership functions are configured. For load matching error, three fuzzy subsets are defined: "low load," "normal," and "overload." For tension overrun error, three fuzzy subsets are defined: "safe," "warning," and "dangerous." Simultaneously, the MIMO fuzzy controller defines two output fuzzy variables: the main drive motor speed correction and the hydraulic coupler filling amount correction, and defines three fuzzy subsets for each output variable: "negative adjustment," "zero," and "positive adjustment."

[0103] In practice, a fuzzy rule base is established based on domain expert knowledge. The rules adopt the conditional statement form of "if...then...", which specifies the corrective actions to be taken under different input states. Some typical fuzzy rules are shown in Table 2.

[0104] Table 2: Fuzzy Control Rule Table

[0105] In each control cycle, the multi-input multi-output fuzzy controller fuzzifies the collected sharp numerical errors using their respective membership functions to determine the membership degree of each input fuzzy variable to different fuzzy subsets. Inference is performed using a fuzzy rule base to calculate the activation intensity of each output fuzzy subset, and the output membership functions are synthesized. The centroid method is then used to defuzzify the synthesized output fuzzy set, converting it into precise correction values. The calculation formula is as follows:

[0106]

[0107] in: This represents the value in the output universe of discourse (i.e., the value of the speed correction or the filling volume correction). This indicates that the output fuzzy variable is at the point The composite membership value at the location, This represents the precise correction amount after defuzzification, calculated using the center of gravity method. The calculated speed correction amount and fluid filling amount correction amount are respectively superimposed onto the current values ​​of the target speed curve and the target fluid filling amount curve, generating an updated closed-loop control command and sending it to the drive system.

[0108] In one embodiment of the present invention, to ensure the safety of the speed regulation strategy, a simulation verification step is added before the generated control commands are sent to the physical drive system. This step is based on offline simulation using a constructed high-fidelity digital twin model of the scraper conveyor. This digital twin model highly replicates the characteristics of the physical system, including accurate dynamic models of the mechanical transmission system, main drive motor, hydraulic coupler, and load. In the simulation environment, the intended target speed curve and target filling volume curve are loaded, and the digital twin model is run to simulate the simulated operating speed, simulated motor current, simulated chain tension, and simulated component wear rate of the equipment under the given strategy. The system automatically compares these simulation results with preset equipment safety thresholds and production efficiency thresholds. If the simulation results show that any indicator exceeds the safety threshold or fails to meet the minimum efficiency requirement, the system will automatically reject the current speed regulation strategy and trigger a redefined process. Only after the simulation verification confirms that the speed regulation strategy meets the requirements in all indicators will the corresponding control commands be formally sent to the physical scraper conveyor drive system on site for execution.

[0109] In practical implementation, before sending the target speed curve of the main drive motor and the target liquid filling volume curve of the hydraulic coupler generated by the multi-objective fusion analysis module to the physical drive system, the system calls the high-fidelity scraper conveyor digital twin model to perform simulation verification steps. The scraper conveyor digital twin model is a virtual entity built based on multibody dynamics and electromechanical system modeling theory. It contains a complete mechanical transmission system model, an electromagnetic torque model of the main drive motor, a fluid transmission model of the hydraulic coupler, and a dynamic model of the coal load. The model parameters are calibrated through the factory test data and actual operation data of the physical scraper conveyor, and have dynamic response characteristics that are highly consistent with the physical system.

[0110] In practical implementation, the simulation verification step loads the target speed curve of the main drive motor and the target liquid filling volume curve of the hydraulic coupler to be verified in the simulation environment. The digital twin model of the scraper conveyor is then driven to run a complete scheduling cycle, simulating the simulated operating speed curve, simulated drive motor current curve, simulated scraper chain tension curve, and simulated trough wear rate curve under this speed regulation strategy. The system presets safety thresholds and production efficiency thresholds for equipment operation. The safety thresholds include the maximum allowable chain tension, the maximum allowable drive motor current, and the maximum allowable trough wear rate. The production efficiency threshold is the minimum average transport speed required to meet shift production requirements.

[0111] In practice, the simulated operation results are automatically compared with preset thresholds. If the simulated speed curve shows an average speed lower than the production efficiency threshold, or the peak value of the simulated drive motor current curve exceeds the maximum allowable current of the motor, or the maximum value of the simulated scraper chain tension curve exceeds the maximum allowable chain tension, or the simulated trough wear rate exceeds the maximum allowable wear rate, then the current speed regulation strategy is deemed to have failed simulation verification. Once simulation verification fails, the system automatically triggers the speed regulation strategy redefinition process, feeding back the out-of-limit information in the simulation results to the multi-objective fusion analysis module to start a new round of optimization calculations. Only when the simulated operating speed, simulated drive motor current, simulated scraper chain tension, and simulated wear rate are all within the preset safety threshold range, and the average transport speed is not lower than the production efficiency threshold, is the speed regulation strategy considered to have passed verification, and the corresponding closed-loop control command is permitted to be issued to the drive system of the physical scraper conveyor for execution.

[0112] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for intelligent speed control of a high wear-resistant scraper conveyor, characterized in that, The method includes: The system collects real-time operating status data of the scraper conveyor and coal flow characteristic data of the working face. The real-time operating status data includes drive motor current, scraper chain tension, and speed difference between the head and tail of the conveyor. The coal flow characteristic data of the working face includes instantaneous coal quantity, coal particle size distribution, and gangue content ratio. Fault features are extracted from the real-time operating status data to generate a scraper conveyor health index, which includes chain wear status, trough wear rate, and drive system efficiency. Based on the coal flow characteristic data of the working face, an improved coal flow prediction algorithm is used to predict future changes in coal load. The improved coal flow prediction algorithm establishes a prediction model based on the nonlinear relationship between historical coal flow data and working face advance speed. The health index of the scraper conveyor and the prediction results of the coal load change are subjected to multi-objective fusion analysis to formulate a speed regulation strategy that meets the high wear resistance constraint. The speed regulation strategy includes the target speed curve of the main drive motor and the target liquid filling volume curve of the hydraulic coupler. Based on the deviation between the speed regulation strategy and real-time operation, a feedback adjustment mechanism is introduced to dynamically adjust the target speed curve and the target liquid filling volume curve, generate a closed-loop control command, and send it to the drive system of the scraper conveyor.

2. The intelligent speed control method for a high wear-resistant scraper conveyor according to claim 1, characterized in that, Fault features are extracted from the real-time operating status data to generate a health index for the scraper conveyor, including: Harmonic analysis is performed on the current of the drive motor to extract the amplitude of the characteristic harmonic components in the current and calculate the load imbalance of the drive motor. The tension of the scraper chain is analyzed in both the time and frequency domains to calculate the average value, standard deviation, and main frequency fluctuation range of the tension signal, thereby obtaining the chain wear characteristic parameters. The tension offset state of the scraper conveyor chain is calculated by comparing the relationship between the speed difference between the head and tail of the machine and a preset threshold. Vibration acceleration signals at key locations in the scraper conveyor trough were collected, and envelope demodulation analysis was performed on the vibration signals to extract the characteristic frequency amplitudes reflecting the wear between the trough and the middle plate. Combining the drive motor load imbalance, chain wear characteristic parameters, tension offset state, and characteristic frequency amplitude, a pre-trained health assessment model is used to calculate and output a quantified health index of the scraper conveyor. The chain wear state is one dimension in the output result of the health assessment model.

3. The intelligent speed control method for a high wear-resistant scraper conveyor according to claim 1, characterized in that, Based on the coal flow characteristic data of the working face, an improved coal flow prediction algorithm is used to predict future changes in coal load, including: Data on instantaneous coal quantity, coal particle size distribution, and gangue content ratio within a historical time period are obtained, along with data on the working face coal mining machine speed and hydraulic support push distance for the corresponding time period, to form a multi-source coal flow time series. The multi-source coal flow time series is cleaned and aligned to eliminate noise and outliers, and then standardized to generate standardized multi-source coal flow time series data. The standardized multi-source coal flow time series data is input into the improved coal flow prediction algorithm. The improved coal flow prediction algorithm uses a long short-term memory network to capture the long-term dependencies of the time series and uses an attention mechanism to dynamically weight the impact of different input features on future load. The improved coal flow prediction algorithm outputs a predicted sequence of coal load within a specific future time window, the predicted sequence including the trend of coal load change and the time point of peak occurrence.

4. The intelligent speed control method for a high wear-resistant scraper conveyor according to claim 3, characterized in that, The working principle of the improved coal flow prediction algorithm is as follows: A model encoder containing multi-layer long short-term memory network units is constructed to encode the input standardized multi-source coal flow time series data and extract deep temporal features of the time series. An attention mechanism layer is connected to the top layer of the model encoder. The attention mechanism layer receives the feature vectors of each time step output by the model encoder and calculates the attention weights of the feature vectors of each historical time step at the future prediction time. The calculated attention weights are weighted and summed on the corresponding historical time step feature vectors to generate a context vector containing global temporal information. The context vector is input into a decoder composed of a multilayer sensing mechanism, and the decoder outputs the predicted coal load values ​​for each future prediction time point. During model training, an adversarial training strategy is introduced, which generates simulated coal flow data through a generative adversarial network to enhance the generalization prediction ability of the improved coal flow prediction algorithm for abnormal working conditions.

5. The intelligent speed control method for a high wear-resistant scraper conveyor according to claim 1, characterized in that, A multi-objective fusion analysis is performed on the health index of the scraper conveyor and the predicted results of the coal load change to formulate a speed regulation strategy that meets the high wear resistance constraint, including: Based on the chain wear state and the trough wear rate, a maximum allowable wear rate threshold for the scraper chain and the central trough is set, and equipment wear safety constraints are generated based on this threshold. Based on the predicted results of the coal load change, the minimum operating speed required to meet the coal transportation efficiency requirements is calculated, and production efficiency constraints are generated. Construct a multi-objective optimization problem with the goal of maximizing the efficiency of the drive system, minimizing the peak fluctuation of the chain tension, and simultaneously satisfying the equipment wear safety constraints and production efficiency constraints; The multi-objective optimization problem is solved using a constrained multi-objective optimization algorithm. The solution results include a series of non-dominated solutions, each of which corresponds to a set of speed regulation parameters. Based on the decision-making strategy, an optimal solution is selected from the set of non-dominated solutions, and the speed regulation parameters in the optimal solution are analyzed into the target speed curve of the main drive motor and the target fluid volume curve of the hydraulic coupler.

6. The intelligent speed control method for a high wear-resistant scraper conveyor according to claim 5, characterized in that, The method of solving the multi-objective optimization problem using a constrained multi-objective optimization algorithm includes: Initialize a population containing multiple individuals, each representing a set of codes for the speed regulation parameters; Decode each individual to obtain the corresponding speed regulation scheme, and calculate the wear constraint violation degree according to the equipment wear safety constraint condition, and calculate the efficiency constraint violation degree according to the production efficiency constraint condition; For individuals that meet the constraints, their fitness value is calculated based on the drive system efficiency value and chain tension peak fluctuation value corresponding to the individual. Selective, crossover, and mutation operations are performed on the population to generate new offspring populations. During the operation, the degree of constraint violation of individuals is evaluated to guide the population to evolve towards the feasible domain. Through multiple generations of iteration, when the convergence condition is met, all non-dominated feasible individuals are selected from the final population to form the non-dominated solution set.

7. The intelligent speed control method for a high wear-resistant scraper conveyor according to claim 1, characterized in that, Based on the deviation between the speed regulation strategy and real-time operation, a feedback adjustment mechanism is introduced to dynamically adjust the target speed curve and the target filling volume curve, including: During the speed regulation strategy execution cycle, continuously monitor the actual operating speed, actual drive motor current and actual chain tension of the scraper conveyor; The actual operating speed is compared with the expected operating speed given by the target speed curve, and the speed tracking error is calculated. The actual drive motor current is compared with the expected current calculated based on the target speed curve and the predicted coal load, and the load matching error is calculated. The actual chain tension is compared with the tension safety range set based on health indicators and operating speed to calculate the tension overshoot error; The speed tracking error, load matching error, and tension overshoot error are input to a multi-input multi-output fuzzy controller, which outputs the speed correction amount for the target speed curve and the filling amount correction amount for the target filling amount curve. The speed correction and the liquid filling correction are respectively superimposed on the current target speed curve and the target liquid filling curve to generate an updated closed-loop control command.

8. The intelligent speed control method for a high wear-resistant scraper conveyor according to claim 7, characterized in that, The working principle of the multi-input multi-output fuzzy controller includes: Define input fuzzy variables, including speed tracking error, load matching error and tension overshoot error, and define multiple fuzzy subsets and corresponding membership functions for each input fuzzy variable; Define output fuzzy variables, including speed correction and filling volume correction, and define multiple fuzzy subsets and corresponding membership functions for each output fuzzy variable; Based on expert experience and a control rule base, fuzzy rules are established from input fuzzy variables to output fuzzy variables. The fuzzy rules describe the relationship between input state and output action in the form of "if...then...". In each control cycle, the values ​​of speed tracking error, load matching error and tension overshoot error, which are collected and calculated in real time, are fuzzified through their respective membership functions to obtain the membership degree of the fuzzy subset corresponding to each input fuzzy variable. Using the aforementioned fuzzy rules and fuzzy inference method, the activation intensity of the fuzzy subset corresponding to each output fuzzy variable is calculated; The membership functions of the output fuzzy variables are synthesized and defuzzified using the centroid method to finally obtain the rotational speed correction and the filling volume correction.

9. The intelligent speed control method for a high wear-resistant scraper conveyor according to claim 4, characterized in that, The step of weighting and summing the calculated attention weights onto the corresponding historical time step feature vectors to generate a context vector containing global temporal information includes: At the top layer of the model encoder, the feature vector output by the last long short-term memory network unit at the last historical time step is obtained as the encoding state at the current moment; In the attention mechanism layer, the feature vectors output by all long short-term memory network units in the model encoder at each historical time step are obtained to form a set of historical feature vectors. For each time point to be predicted in the future prediction sequence, a query vector is calculated in the attention mechanism layer. The query vector is formed by concatenating the encoding state of the current time and the position encoding of the time point to be predicted after a linear transformation. The relevance score is calculated between the query vector and each historical feature vector in the set of historical feature vectors. The relevance score is calculated through a trainable multilayer perceptron network. The correlation scores of each historical feature vector are input into the Softmax function for normalization to obtain the attention weight corresponding to each historical feature vector, and the sum of the attention weights of all historical feature vectors is 1. Multiply each historical feature vector in the set of historical feature vectors by its corresponding attention weight to obtain a weighted feature vector. All weighted feature vectors are summed to generate a fixed-dimensional context vector for the time point to be predicted. The context vector incorporates the time series information most relevant to the time point to be predicted from all historical time steps.

10. The intelligent speed control method for a high wear-resistant scraper conveyor according to claim 1, characterized in that, The method also includes a simulation verification step before the speed regulation strategy is executed: Before the speed regulation strategy is sent to the drive system for execution, the speed regulation strategy is simulated and verified based on a high-fidelity digital twin model of the scraper conveyor. The digital twin model includes dynamic models of the mechanical transmission system, drive motor, hydraulic coupling, and load; The target speed curve and target fluid volume curve are run in the digital twin model to simulate the simulated running speed, simulated motor current, simulated chain tension, and simulated wear rate. The simulation results are compared with the set safety thresholds and efficiency thresholds. If the simulation results exceed the safety thresholds or fail to reach the efficiency thresholds, the speed regulation strategy is automatically redefined. Only after the simulation verification is passed will the corresponding speed regulation strategy command be sent to the drive system of the physical scraper conveyor for execution.

Images