Hydraulic support laser cladding process parameter optimization method based on big data

Abstract

The invention belongs to the technical field of hydraulic support stand column surface laser cladding, and particularly relates to a hydraulic support laser cladding process parameter optimization method based on big data. Comprising the steps that S1, historical data of hydraulic support laser cladding from different coal mines under different working conditions are collected, and the historical data comprise technological parameters and corresponding coating performance; s2, cleaning, duplicate removal and standardization are carried out on the collected historical data; s3, finding out the correlation between the process parameters and the coating performance, extracting key process features and performance features, and constructing feature vectors; s4, on the basis of the extracted feature vectors, using a neural network to construct a prediction model between the process parameters and the coating performance; s5, optimizing laser cladding process parameters according to the constructed prediction model; and S6, the optimized laser cladding process parameters are applied to actual production. According to the invention, the automation and intelligence of the cladding process are realized, the individual requirements of different manufacturers are met, and the bonding performance of the cladding layer is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of laser cladding on the surface of hydraulic support columns, and specifically is a method for optimizing process parameters of laser cladding on hydraulic support based on big data. Background Art

[0002] Traditional process parameter setting mainly relies on simple statistical analysis and designers’ experience and judgment, and cannot fully utilize big data analysis technology to explore deep patterns and features hidden in the data. This leads to a lack of scientific basis for the adjustment process of process parameter setting, which is prone to deviations and cannot meet the needs. In the existing technology, the limited data analysis tools and methods make it impossible to extract key factors from the complex data, and the optimization accuracy of process parameters is limited.

[0003] The existing parameter adjustment process usually lacks effective simulation verification methods, and the adjustment of process parameters mainly relies on the experience of designers and a small amount of actual trial feedback. This not only increases the cost of trial and error, but may also cause the process parameters to fail to achieve the expected results in actual production, affecting the hardness and corrosion resistance of the cladding layer. In the existing technology, the shortcomings of simulation models and verification methods limit the reliability and feasibility of parameter optimization solutions, resulting in poor application results in actual production.

[0004] In order to overcome these problems in the existing technology, it is necessary to propose a method for optimizing the remanufacturing process parameters of the laser cladding technology on the surface of mining hydraulic support columns based on big data to improve the efficiency and accuracy of the parameter adjustment process. Summary of the invention

[0005] In order to solve the problem that the processing parameters are selected based on experience and the principle of a large safety factor, which leads to the inability to reasonably match the process parameters and the surface quality of the cladding layer does not reach an ideal level, the present invention provides a method for optimizing the remanufacturing process parameters of the laser cladding technology on the surface of a mining hydraulic support column for specific equipment and hydraulic cylinder raw materials.

[0006] The present invention adopts the following technical scheme: a method for optimizing the remanufacturing process parameters of a mine hydraulic support column surface laser cladding technology, comprising: S1: Collect historical data of laser cladding of hydraulic supports under different working conditions from different coal mines, including process parameters and corresponding coating performance; S2: Clean, remove duplicates and standardize the collected historical data; S3: Find the correlation between process parameters and coating performance, extract key process characteristics and performance characteristics, and construct feature vectors; S4: Based on the extracted feature vectors, a prediction model between process parameters and coating performance is constructed using a neural network; S5: Optimize laser cladding process parameters based on the constructed prediction model; S6: Apply the optimized laser cladding process parameters to actual production, perform performance tests on the coating after cladding, and compare and verify the test results with the predicted results; if there is a deviation, feed the deviation information back to the neural network system, retrain the model and optimize the parameters to continuously improve the accuracy of process parameter optimization.

[0007] In some embodiments, in step S1: Historical data of laser cladding of hydraulic supports. The process parameters include cladding laser power P, scanning speed V, and coating properties include hardness H, wear resistance W, and corrosion resistance C of the coating after cladding.

[0008] In some embodiments, step S2 includes: S21: organize the historical data into transactions, each transaction represents a laser cladding experiment or production process, and a transaction is {laser power, scanning speed, powder feeding rate, overlap rate, hardness, wear resistance, corrosion resistance}, where wear resistance is divided into three levels: poor, medium, and good, coded as 0, 1, and 2 respectively. The coating wear mark is greater than or equal to the substrate wear mark, which is poor, the coating wear mark is less than the substrate wear mark by less than 200 microns, which is medium, and the coating wear mark is less than the substrate wear mark by more than 200 microns, which is good; corrosion resistance is divided into three levels: poor, medium, and good, coded as 0, 1, and 2 respectively. The coating corrosion potential is less than the substrate corrosion potential by more than 30mV, which is poor, the coating corrosion potential is between 30mV and the substrate corrosion potential, which is medium, and the coating corrosion potential is greater than the substrate corrosion potential by more than 30mV, which is good; S22: Each transaction is marked with Ax, where A represents the subject of “process parameter optimization study” and x represents the serial number of the transaction, and Min-Max normalization is used.

[0009] In some embodiments, step S3 includes: S31: Calculate the support of process parameters; The calculation formula of support is: , where Support (X) represents the support of item set X, which includes laser power P, scanning speed V, powder feeding rate R, and overlap rate O; S32: Scan the process parameter support data set obtained in step S31, filter out single items whose support is greater than or equal to the minimum support, and generate frequent 1-item sets; S33: Combine frequent 1-item sets to generate candidate 2-item sets, scan the data set again, and calculate the support of each candidate 2-item set.

[0010] S34: Repeat steps S33 and S34, continuously combine frequent item sets to generate higher-level candidate item sets, and calculate support for screening until no more frequent item sets can be generated; S35: For each frequent item set, generate association rules. Association rules are usually expressed as X→Y, that is, X will occur Y to a certain extent; S36: Calculate the confidence of each association rule, and determine the minimum confidence of each association rule based on experience or multiple experiments; S37: Scan the confidence data set of each association rule obtained in step S36, interpret the association rules that meet the confidence requirements, and stop mining for the association rules that do not meet the confidence requirements; S38: The processed key process features and performance features are stitched together in a certain order to form a feature vector.

[0011] In some embodiments, in step S36, the confidence calculation formula is: , where Confidence (E→F) represents the confidence of the probability of item set F appearing when item set E has already appeared; Support (E∪F) represents the frequency of item sets E and F appearing at the same time; Support (E) represents the frequency of item set E appearing alone.

[0012] In some embodiments, in step S4, the neural network adopts a multilayer perceptron neural network, which includes an input layer, a hidden layer and an output layer, wherein the input is a feature variable, the output is a target variable, the number of neural units in the input layer is 4, and the number of neural units in the output layer is 3.

[0013] In some embodiments, step S4 includes: S41: Prepare data. Laser power P, scanning speed V, powder feeding rate R, and overlap rate O are feature variables. Coating hardness H, wear resistance W, and corrosion resistance C are target variables. Divide the feature vector data into two subsets according to the ratio of 80% training set and 20% test set. S42: Load the divided data into a multi-layer perceptron neural network training framework for training; Among them, the hidden layer activation function type is ReLU function; Use mean squared error loss function; The weights and biases are updated using the Adam optimizer.

[0014] In some embodiments, step S5 includes: S51: Determine the population size, crossover probability and mutation probability of the genetic algorithm.

[0015] S52: using real number coding, taking laser power P, scanning speed V, powder feeding rate R, and overlap rate O as genes in the chromosome, and the value range is determined according to the range of actual process parameters; S53: randomly generate an initial population according to the encoding method and population size, each individual represents a set of laser cladding process parameter combinations, and input each individual in the population into the neural network prediction model to obtain the corresponding coating performance prediction value, including hardness, wear resistance, and corrosion resistance; S54: Roulette wheel selection method is used to select some individuals as parents of the next generation according to their fitness; S55: Use the arithmetic crossover method to perform a crossover operation on the selected parent individuals to generate new offspring individuals; S56: Set the mutation probability, and the mutation mode is uniform mutation; S57: Add the newly generated offspring individuals to the population, replace some individuals with lower fitness, form a new population, and then evaluate the fitness of each individual in the new population again, repeat the above selection, crossover, and mutation operations until the stopping condition is met, that is, the fitness converges.

[0016] In some embodiments, in step S54, parent individual A 1 =[P 1 ,V 1， R 1, O 1 ] and A 2 =[P 2 ,V 2 R 1, O 1 ], where p represents the laser power and v represents the scanning speed, and the progeny B is generated by arithmetic crossover 1 and B 2 The formula is as follows: where α is the crossover coefficient.

[0017] In some embodiments, the fitness function is: Fitness=w 1 ×Hardness+w 2 × Wear resistance + w 3 ×Corrosion resistance, where w 1 、w 2、 w 3 is the weight coefficient.

[0018] Compared with the prior art, the present invention realizes the automatic adjustment and optimization of process parameters, improves the performance of the cladding layer, and significantly improves the optimization efficiency and accuracy of laser cladding process parameters through multi-channel data collection, construction and enhanced processing of comprehensive data models, deep learning and nonlinear relationship extraction, intelligent optimization algorithm and simulation verification, realizes the automation and intelligence of the cladding process, meets the personalized needs of different manufacturers, and improves the bonding performance of the cladding layer. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0021] like Figure 1 As shown, a method for optimizing process parameters of remanufacturing technology for the surface of a mining hydraulic support column by laser cladding technology includes: S1: Collect historical data of laser cladding of hydraulic supports from different coal mines and under different working conditions. The historical data include process parameters and corresponding coating performance. Historical data of laser cladding of hydraulic support, process parameters include cladding laser power P, scanning speed V, powder feeding rate R, overlap rate O and coating properties including hardness H, wear resistance W and corrosion resistance C of the coating after cladding.

[0022] Specifically, we obtained the original data related to the laser cladding process parameters from the public information of the enterprise, collected different sets of process parameter data used by the enterprise, and used the YDJG600-3000 high-speed laser cladding machine with a power range of 0-8 kW, a spot radius of 2 mm, and coaxial powder feeding. The process parameter table is shown in Table 1: Process parameter table 1 S2: Clean, deduplicate and standardize the collected historical data.

[0023] S21: organize the historical data into transactions, each transaction represents a laser cladding experiment or production process, and a transaction is {laser power, scanning speed, powder feeding rate, overlap rate, hardness, wear resistance, corrosion resistance}, where wear resistance is divided into three levels: poor, medium, and good, coded as 0, 1, and 2 respectively. The coating wear mark is greater than or equal to the substrate wear mark, which is poor, the coating wear mark is less than the substrate wear mark by less than 200 microns, which is medium, and the coating wear mark is less than the substrate wear mark by more than 200 microns, which is good; corrosion resistance is divided into three levels: poor, medium, and good, coded as 0, 1, and 2 respectively. The coating corrosion potential is less than the substrate corrosion potential by more than 30mV, which is poor, the coating corrosion potential is between 30mV and the substrate corrosion potential, which is medium, and the coating corrosion potential is greater than the substrate corrosion potential by more than 30mV, which is good.

[0024] S22: Each transaction is marked with Ax, where A represents the subject of "process parameter optimization research" and x represents the serial number of the transaction (1, 2, 3...). Min-Max standardization is used, and the formula is: , where x is the original data, x min is the minimum value of the feature in the data set, x max is the maximum value of this feature in the data set, x norm is the normalized data. Table 2 is obtained.

[0025] S3: Find the correlation between process parameters and coating performance, extract key process characteristics and performance characteristics, and construct feature vectors.

[0026] Step S3 includes: S31: Calculate the support of process parameters; The calculation formula of support is: , where Support (X) represents the support of the item set X, which includes laser power P, scanning speed V, powder feeding rate R, and overlap rate O. For example, the support of laser power P is , P represents the laser power. And based on experience or multiple tests, the minimum support of each process parameter is determined.

[0027] , Similarly, Confidence (P→C) is calculated using the same method.

[0028] S32: Scan the process parameter support data set obtained in step S31, filter out single items whose support is greater than or equal to the minimum support, and generate frequent 1-item sets; S33: Combine frequent 1-item sets to generate candidate 2-item sets, scan the data set again, and calculate the support of each candidate 2-item set.

[0029] S34: Repeat steps S32 and S33, continuously combine frequent item sets to generate higher-level candidate item sets, and calculate support for screening until no more frequent item sets can be generated.

[0030] S35: For each frequent item set, generate association rules. Association rules are usually expressed as X→Y, that is, X will occur Y to a certain extent.

[0031] S36: Calculate the confidence of each association rule. The confidence calculation formula is: , where Confidence (E→F) represents the confidence of the probability of item set F appearing when item set E has already appeared; Support (E∪F) represents the frequency of item sets E and F appearing at the same time; Support (E) represents the frequency of item set E appearing alone. And based on experience or multiple experiments, determine the minimum confidence of each association rule.

[0032] S37: Scan the confidence data set of each association rule obtained in step S37, interpret the association rules that meet the confidence requirements, and stop mining for the association rules that do not meet the confidence requirements.

[0033] S38: The processed key process features and performance features are stitched together in a certain order to form a feature vector.

[0034] S4: Based on the extracted feature vectors, a prediction model between process parameters and coating performance is constructed using a neural network.

[0035] S41: Prepare data. Laser power P, scanning speed V, powder feeding rate R, and overlap rate O are feature variables. Coating hardness H, wear resistance W, and corrosion resistance C are target variables. Divide the feature vector data into two subsets according to the ratio of 80% training set and 20% test set. S42: Load the divided data into the multi-layer perceptron (MLP) neural network training framework; The multi-layer perceptron (MLP) structure includes an input layer, a hidden layer, and an output layer. The input is a feature variable and the output is a target variable, so the number of neural units in the input layer is 4 and the number of neural units in the output layer is 3.

[0036] Among them, the hidden layer activation function type is ReLU function, and the expression of ReLU function is .

[0037] The multi-layer perceptron (MLP) uses the mean square error (MSE) loss function, which is given by , where n is the number of samples and y is the true value, is the predicted value.

[0038] Use Adam optimizer to update weights and biases. The weight update formula is , where w ij represents the weight of the connection between the jth neuron in the i-th layer of the neural network and the neurons in the next layer, α is the learning rate, is the loss function L with respect to weight w ij gradient.

[0039] S5: Optimize the laser cladding process parameters based on the constructed prediction model.

[0040] S51: Determine the population size, crossover probability and mutation probability of the genetic algorithm.

[0041] S52: using real number coding, taking laser power P, scanning speed V, powder feeding rate R, and overlap rate O as genes in the chromosome, and the value range is determined according to the range of actual process parameters; S53: randomly generate an initial population according to the encoding method and population size, each individual represents a set of laser cladding process parameter combinations, and input each individual in the population into the neural network prediction model to obtain the corresponding coating performance prediction value, including hardness, wear resistance, and corrosion resistance; S54: Roulette wheel selection method is used to select some individuals as parents of the next generation according to their fitness; Input each individual (process parameter) in the population into the neural network prediction model to obtain the corresponding coating performance prediction value (hardness, wear resistance, corrosion resistance). The fitness function is: Fitness=w 1 ×Hardness+w 2 × Wear resistance + w 3 ×Corrosion resistance, where w 1 、w 2、 w 3 is the weight coefficient.

[0042] S55: Use the arithmetic crossover method to perform a crossover operation on the selected parent individuals to generate new offspring individuals; parent individual A 1 =[P 1 ,V 1， R 1, O 1 ] and A 2 =[P 2 ,V 2 R 1, O 1 ], where p represents the laser power and v represents the scanning speed, and the progeny B is generated by arithmetic crossover 1 and B 2 The formula is as follows: where α is the crossover coefficient.

[0043] S56: Set the mutation probability to 0.05 and the mutation mode to uniform mutation.

[0044] S57: Add the newly generated offspring individuals to the population, replace some individuals with lower fitness, form a new population, and then evaluate the fitness of each individual in the new population again, repeat the above selection, crossover, and mutation operations until the stopping condition is met, that is, the fitness converges.

[0045] S6: Apply the optimized laser cladding process parameters to actual production, perform performance tests on the coating after cladding, and compare and verify the test results with the predicted results; if there is a deviation, feed the deviation information back to the neural network system, retrain the model and optimize the parameters to continuously improve the accuracy of process parameter optimization.

[0046] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing process parameters of laser cladding technology for the surface of hydraulic support columns for mining, characterized in that: include: S1: Collect historical data of laser cladding of hydraulic supports under different working conditions from different coal mines, including process parameters and corresponding coating performance; S2: Clean, remove duplicates and standardize the collected historical data; S3: Find the correlation between process parameters and coating performance, extract key process characteristics and performance characteristics, and construct feature vectors; S4: Based on the extracted feature vectors, a prediction model between process parameters and coating performance is constructed using a neural network; S5: Optimize laser cladding process parameters based on the constructed prediction model; S6: Apply the optimized laser cladding process parameters to actual production, perform performance tests on the cladding coating, and compare and verify the test results with the predicted results; If there is a deviation, the deviation information will be fed back to the neural network system, the model will be retrained and the parameters will be optimized to continuously improve the accuracy of process parameter optimization.

2. The method for optimizing process parameters of the remanufacturing technology of the surface laser cladding technology of the mining hydraulic support column according to claim 1 is characterized in that: In step S1: Historical data of laser cladding of hydraulic supports. The process parameters include cladding laser power P, scanning speed V, and coating properties include hardness H, wear resistance W, and corrosion resistance C of the coating after cladding.

3. The method for optimizing process parameters of the remanufacturing technology of the surface laser cladding technology of the mining hydraulic support column according to claim 1 is characterized in that: The step S2 comprises: S21: organize the historical data into transactions, each transaction represents a laser cladding experiment or production process, and a transaction is {laser power, scanning speed, powder feeding rate, overlap rate, hardness, wear resistance, corrosion resistance}, where wear resistance is divided into three levels: poor, medium, and good, coded as 0, 1, and 2 respectively. The coating wear mark is greater than or equal to the substrate wear mark, which is poor, the coating wear mark is less than the substrate wear mark by less than 200 microns, which is medium, and the coating wear mark is less than the substrate wear mark by more than 200 microns, which is good; corrosion resistance is divided into three levels: poor, medium, and good, coded as 0, 1, and 2 respectively. The coating corrosion potential is less than the substrate corrosion potential by more than 30mV, which is poor, the coating corrosion potential is between 30mV and the substrate corrosion potential, which is medium, and the coating corrosion potential is greater than the substrate corrosion potential by more than 30mV, which is good; S22: Each transaction is marked with Ax, where A represents the subject of "process parameter optimization study" and x represents the serial number of the transaction, and Min-Max normalization is used.

4. The method for optimizing process parameters of the remanufacturing process of the surface laser cladding technology of the mining hydraulic support column according to claim 3 is characterized in that: The step S3 comprises: S31: Calculate the support of process parameters; The calculation formula of support is: , where Support (X) represents the support of item set X, which includes laser power P, scanning speed V, powder feeding rate R, and overlap rate O; S32: Scan the process parameter support data set obtained in step S31, filter out single items whose support is greater than or equal to the minimum support, and generate frequent 1-item sets; S33: Combine frequent 1-item sets to generate candidate 2-item sets, scan the data set again, and calculate the support of each candidate 2-item set. S34: Repeat steps S33 and S34, continuously combine frequent item sets to generate higher-level candidate item sets, and calculate support for screening until no more frequent item sets can be generated; S35: For each frequent item set, generate association rules. Association rules are usually expressed as X→Y, that is, X will occur Y to a certain extent; S36: Calculate the confidence of each association rule, and determine the minimum confidence of each association rule based on experience or multiple experiments; S37: Scan the confidence data set of each association rule obtained in step S36, interpret the association rules that meet the confidence requirements, and stop mining for the association rules that do not meet the confidence requirements; S38: The processed key process features and performance features are stitched together in a certain order to form a feature vector.

5. The method for optimizing process parameters of the remanufacturing technology of the surface laser cladding technology of the mining hydraulic support column according to claim 4 is characterized in that: In step S36, the confidence calculation formula is: , where Confidence (E→F) represents the confidence of the probability of item set F appearing when item set E has already appeared; Support (E∪F) represents the frequency of item sets E and F appearing at the same time; Support (E) indicates the frequency with which item set E appears alone.

6. The method for optimizing process parameters of the remanufacturing technology of the surface laser cladding technology of the mining hydraulic support column according to claim 1 is characterized in that: In step S4, the neural network adopts a multilayer perceptron neural network, which includes an input layer, a hidden layer and an output layer, wherein a feature variable is input and a target variable is output, the number of neural units in the input layer is 4, and the number of neural units in the output layer is 3.

7. The method for optimizing process parameters of the remanufacturing technology of the surface laser cladding technology of the mining hydraulic support column according to claim 6 is characterized in that: Step S4 includes: S41: Prepare data. Laser power P, scanning speed V, powder feeding rate R, and overlap rate O are feature variables. Coating hardness H, wear resistance W, and corrosion resistance C are target variables. Divide the feature vector data into two subsets according to the ratio of 80% training set and 20% test set. S42: Load the divided data into a multi-layer perceptron neural network training framework for training; Among them, the hidden layer activation function type is ReLU function; Use mean squared error loss function; The weights and biases are updated using the Adam optimizer.

8. The method for optimizing process parameters of the remanufacturing technology of the surface laser cladding technology of the mining hydraulic support column according to claim 4 is characterized in that: The step S5 comprises: S51: Determine the population size, crossover probability and mutation probability of the genetic algorithm. S52: using real number coding, taking laser power P, scanning speed V, powder feeding rate R, and overlap rate O as genes in the chromosome, and the value range is determined according to the range of actual process parameters; S53: randomly generate an initial population according to the encoding method and population size, each individual represents a set of laser cladding process parameter combinations, and input each individual in the population into the neural network prediction model to obtain the corresponding coating performance prediction value, including hardness, wear resistance, and corrosion resistance; S54: Roulette wheel selection method is used to select some individuals as parents of the next generation according to their fitness; S55: Use the arithmetic crossover method to perform a crossover operation on the selected parent individuals to generate new offspring individuals; S56: Set the mutation probability, and the mutation mode is uniform mutation; S57: Add the newly generated offspring individuals to the population, replace some individuals with lower fitness, form a new population, and then evaluate the fitness of each individual in the new population again, repeat the above selection, crossover, and mutation operations until the stopping condition is met, that is, the fitness converges.

9. The method for optimizing process parameters of the remanufacturing technology of the surface laser cladding technology of the mining hydraulic support column according to claim 8 is characterized in that: In step S55, the parent individual A1=[P1,V 1， R 1, O1] and A2=[P2,V2 R 1, O1], where p represents the laser power and v represents the scanning speed. The formula for generating the offspring B1 and B2 by arithmetic crossover is as follows: where α is the crossover coefficient.

10. The method for optimizing process parameters of the remanufacturing technology of the surface laser cladding technology of the mining hydraulic support column according to claim 8 is characterized in that: Fitness function is: Fitness = w1×hardness+w2×wear resistance+w3×corrosion resistance, where w1, w 2、 w3 is the weight coefficient.

Images