Method for preparing metal alloy powder for on-line cladding of yaw large disc of fan generator set

Abstract

The invention relates to the technical field of power equipment control processing, and provides a preparation method of metal alloy powder for online cladding of a yaw large disc of a fan generator set, which realizes intelligent monitoring of a preparation process by acquiring multi-dimensional preparation data and utilizing a random forest model to construct a prediction model. The method specifically comprises the steps of data acquisition, model training, real-time data comparison, error range matching, model optimization and the like so as to improve the accuracy of preparation data, and when a preparation error exceeds a preset range, a genetic algorithm is adopted to optimize the model and adjust preparation parameters, so that the quality of a cladding layer is improved. According to the invention, the problem of insufficient data accuracy detection in the prior art is effectively solved, the intelligence of the preparation process is realized, and the cladding quality is obviously improved.

Description

Technical Field

[0001] The invention relates to the technical field of power equipment control processing, and in particular to a method for preparing metal alloy powder for online cladding of a yaw disc of a wind turbine generator set. Background Art

[0002] Online cladding of the yaw disc of a wind turbine generator set is a repair technology. It uses specific metal alloy powder as welding material, and the alloy powder is clad on the repair surface of the yaw disc through a laser cladding process. The online cladding technology of the yaw disc of a wind turbine generator set can achieve metallurgical bonding, completely replacing the traditional mechanical bonding method. It not only improves the hardness and wear resistance of the repaired part, but also meets the strict requirements for precision and roughness. When preparing metal alloy powder, it is necessary to accurately control the composition and proportion of the alloy powder according to factors such as the base material of the yaw disc, the requirements of the repair surface, and the welding process parameters to ensure the quality of the cladding layer.

[0003] The specific steps of data processing for the preparation of metal alloy powder for online cladding of the yaw disc of the wind turbine generator set include collecting the material type, composition ratio, physical and chemical properties of the base material, welding material and cladding layer. Then, according to the size, shape and missing condition of the repaired surface, as well as welding process parameters such as laser power and welding speed, data analysis and processing are carried out to determine the preparation plan of the alloy powder, including the ratio of alloy elements, particle size distribution of powder, etc. Finally, the quality of the prepared metal alloy powder is tested.

[0004] The preparation of metal alloy powder for online cladding of the yaw plate of a wind turbine generator set has the following technical pain points. During the preparation process, a large amount of data needs to be collected and processed, such as the physical and chemical properties of the base material, welding material and cladding layer, as well as welding process parameters. The precision and accuracy of the data directly affect the preparation plan of the alloy powder and the final cladding quality. The existing technology is unable to detect the data accuracy of each process in the preparation process of metal alloy powder for online cladding of the yaw plate of a wind turbine generator set, which affects the cladding quality. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a method for preparing metal alloy powder for online cladding of a yaw disk of a wind turbine generator set, thereby solving the problem that the prior art is unable to detect the data accuracy of each process in the process of preparing metal alloy powder for online cladding of a yaw disk of a wind turbine generator set, thus affecting the cladding quality.

[0006] The present invention provides a method for preparing metal alloy powder for online cladding of a yaw disk of a wind turbine generator set, comprising:

[0007] Step S101, obtaining preparation data, the preparation data including base material data, welding material data, cladding layer requirement data, repair surface data, welding process parameters, environmental parameters, quality inspection data and process monitoring data;

[0008] Step S102, dividing the preparation data into a training set and a test set, selecting a random forest as a model, using the training set to train the random forest model, using the cross-validation method to evaluate the model performance under different parameter combinations on the test set data, and selecting the optimal parameter combination to obtain a metal alloy powder preparation prediction model for online cladding of the yaw disc of a wind turbine generator set;

[0009] Step S103, receiving a preparation task, the preparation task including preparation equipment information and preparation task parameters, sending the task to the device end for execution, collecting preparation data after the device end executes the preparation task, substituting the preparation task into a metal alloy powder preparation prediction model for online cladding of a yaw disk of a wind turbine generator set, outputting an expected value of the preparation data of the metal alloy powder preparation prediction model for online cladding of a yaw disk of a wind turbine generator set, comparing the expected value of the preparation data of the metal alloy powder preparation prediction model for online cladding of a yaw disk of a wind turbine generator set with the preparation data after the device end executes the preparation task, and obtaining a preparation data comparison result;

[0010] Step S104, matching the preparation task in a preset error range knowledge base to obtain an error range corresponding to the preparation task, and comparing the preparation data comparison result with the error range corresponding to the preparation task to obtain a preparation error range comparison result;

[0011] Step S105, obtaining preparation history data and preparation history tasks. If the preparation error range comparison result exceeds the error range corresponding to the preparation task, the preparation prediction model of the metal alloy powder for online cladding of the yaw disk of the wind turbine generator set is optimized using a genetic algorithm according to the preparation data comparison result, the error of the error range corresponding to the preparation task, and the preparation history data, to obtain an optimized preparation prediction model for online cladding of the yaw disk of the wind turbine generator set. The preparation task is processed using the optimized preparation prediction model for online cladding of the yaw disk of the wind turbine generator set to obtain corrected preparation task parameters, and the corrected preparation task parameters are sent to the device end.

[0012] Furthermore, the method for preparing metal alloy powder for online cladding of a yaw disk of a wind turbine generator set according to the present invention, the step S101 comprises:

[0013] The matrix material data includes the type, composition, hardness and strength of the yaw plate matrix material;

[0014] Welding material data include chemical composition, particle size distribution, and melting point of metal alloy powder;

[0015] The required data of the cladding layer include the required thickness, hardness, wear resistance and corrosion resistance of the cladding layer;

[0016] The repair surface data includes the state of the surface to be repaired of the yaw disk, and the state of the surface to be repaired of the yaw disk includes the degree of damage, surface roughness and cleanliness;

[0017] Welding process parameters include welding current, voltage, welding speed and powder feeding rate;

[0018] Environmental parameters include temperature, humidity, and wind speed during the preparation process;

[0019] Quality inspection data includes hardness test and metallographic analysis;

[0020] The process monitoring data includes data collected in real time during the preparation process, and the data collected in real time during the preparation process includes fluctuation data of welding current and voltage.

[0021] Furthermore, in the method for preparing metal alloy powder for online cladding of a yaw disk of a wind turbine generator set according to the present invention, step S102 comprises:

[0022] Preprocessing the obtained preparation data, and dividing the preprocessed preparation data into a training set and a test set;

[0023] The random forest model is trained using the training set data. During the training process, the base material data and welding material data are used as input features, and the corrected preparation task parameters are used as output targets.

[0024] In the cross-validation process, different random forest parameter combinations are used, and the performance indicators of different random forest parameter combinations on the test set are calculated, and the parameter combination with the best performance indicator is selected as the parameter of the final model.

[0025] Furthermore, the method for preparing metal alloy powder for online cladding of the yaw disk of a wind turbine generator set according to the present invention, the step S103 comprises:

[0026] Receive preparation tasks. Preparation equipment information includes equipment model and equipment status. Preparation task parameters include base material type, welding material type and cladding layer requirements.

[0027] Send the preparation task to the designated equipment end, which includes metal powder making equipment, powder processing equipment, laser cladding equipment, powder feeding system, grinder, grinder, cladding robot, welding equipment and sensor, and start the equipment to execute the preparation task;

[0028] During the equipment's preparation task, actual preparation data is collected through sensors and data interfaces. The actual preparation data includes the thickness, hardness, chemical composition and surface quality of the cladding layer.

[0029] Establish a data comparison item for the expected value of the preparation data of the metal alloy powder preparation prediction model for online cladding of the yaw plate of the wind turbine generator set and the preparation data after the equipment side executes the preparation task;

[0030] For each comparison item, the difference between the predicted value and the actual value is calculated to obtain the preparation data comparison result.

[0031] Furthermore, the method for preparing metal alloy powder by online cladding of the yaw disk of a wind turbine generator set according to the present invention, the step S104 comprises:

[0032] Loading a preset error range knowledge base, the preset error range knowledge base includes acceptable error ranges corresponding to different types of preparation tasks;

[0033] Perform feature extraction on the preparation task to obtain the required features of the base material, welding material type and cladding layer, match the corresponding error range in the error range knowledge base, and extract the error range value from the matched error range knowledge base entry;

[0034] Compare the preparation data comparison results with the error range corresponding to the preparation task. For each comparison item, compare the actual difference value with the corresponding error range to determine whether the actual difference falls within the acceptable range.

[0035] Furthermore, the method for preparing metal alloy powder by online cladding of the yaw disk of a wind turbine generator set according to the present invention, the step S105 comprises:

[0036] Acquire preparation history data, which includes preparation task parameters, actual preparation data, and corresponding preparation error range comparison results;

[0037] Receive and set the optimization target, and define the optimization target as minimizing the difference between the predicted value and the actual value based on the preparation error range comparison result;

[0038] Determine the basic parameters of the genetic algorithm, which include population size, number of iterations, crossover probability, and mutation probability;

[0039] The parameters of the prediction model for the preparation of metal alloy powder for online cladding of the yaw disk of a wind turbine generator set are encoded into genotypes that can be processed by a genetic algorithm;

[0040] Initialize the population and randomly generate a set of initial model parameters as the population of the genetic algorithm;

[0041] Each individual represents a model parameter combination;

[0042] Evaluate fitness, use the optimization objective function to evaluate the fitness of each individual, that is, calculate the corresponding model prediction error;

[0043] Selection operation, which selects a part of individuals as parents according to their fitness to generate the next generation;

[0044] Crossover operation: perform crossover operation on the selected parent individuals to generate new offspring individuals;

[0045] Mutation operation: perform mutation operation on offspring individuals to introduce new genotypes;

[0046] The mutation operation can randomly change the gene positions in individuals, replace some or all of the parent individuals with newly generated offspring individuals, form a new population, and repeat the fitness evaluation, selection, crossover and mutation operations until the predetermined number of iterations or optimization goal is reached;

[0047] The best individual is selected, and the individual with the highest fitness is selected in the last generation of the population, that is, the optimized model parameter combination, and the optimized model parameter combination is used to update the metal alloy powder preparation prediction model for online cladding of the yaw disk of the wind turbine generator set.

[0048] Furthermore, the method for preparing metal alloy powder by online cladding of the yaw disk of a wind turbine generator set according to the present invention, the step S105 comprises:

[0049] Preprocessing the received preparation task data, and inputting the preprocessed preparation task data into the optimized metal alloy powder preparation prediction model for online cladding of the yaw plate of the wind turbine generator set;

[0050] The optimized prediction model for the preparation of metal alloy powder for online cladding of the yaw disc of the wind turbine generator set predicts the parameters of the preparation process based on the input data. The parameters predicted by the data include the thickness, hardness and chemical composition of the cladding layer.

[0051] Based on the prediction results of the optimized prediction model for the online cladding of the yaw disk of the wind turbine generator set, the preparation task parameters are corrected and sent to the equipment end.

[0052] Beneficial effects of the present invention;

[0053] The present invention can improve the data accuracy of the metal alloy powder preparation process of the online cladding of the yaw plate of the wind turbine generator set by introducing machine learning algorithms (such as random forests) and real-time data monitoring mechanisms. By comprehensively collecting, preprocessing and training the preparation data, the present invention can predict the expected data value of the preparation process and compare it with the actual preparation data, so as to timely discover and correct data deviations and improve the quality and performance of the cladding layer.

[0054] The present invention realizes intelligent monitoring and management of the preparation process. By collecting data from the device end in real time and comparing it with the expected value of the prediction model, the system can automatically determine the deviation of the preparation process and trigger the corresponding optimization mechanism. This intelligent monitoring method significantly reduces the dependence on manual intervention and improves production efficiency and automation level.

[0055] The optimization methods such as genetic algorithms adopted by the present invention can continuously optimize the prediction model and control parameters according to real-time data feedback, ensuring that the preparation process always remains in the optimal state. This adaptive ability enables the present invention to adapt to the changing needs of different preparation tasks, improving the flexibility and robustness of the system. Due to the introduction of intelligent monitoring and optimization mechanisms, the present invention can significantly reduce data deviations and error rates in the preparation process, reduce scrap rates and rework rates, thereby improving production efficiency and economic benefits. At the same time, the optimized preparation task parameters can guide the equipment to perform preparation tasks more efficiently, further reducing production costs.

[0056] By ensuring the data accuracy of the preparation process and the high quality of the cladding layer, the present invention helps to extend the service life of equipment such as the yaw disc of a wind turbine generator set. The high-quality cladding layer can provide better wear resistance, corrosion resistance and mechanical properties, reduce wear and failure of the equipment during operation, and reduce maintenance costs.

[0057] In summary, the present invention significantly improves the preparation quality of metal alloy powder for online cladding of the yaw disc of a wind turbine generator set by introducing an intelligent monitoring and optimization mechanism, and has significant beneficial effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on the drawings without paying any creative labor.

[0059] Figure 1 A schematic flow chart of a method for preparing metal alloy powder by online cladding of a yaw disc of a wind turbine generator set provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. The technical solutions provided by the embodiments of the present invention are described in detail below in conjunction with the drawings.

[0061] In order to better understand the purpose of the present invention, the present invention is described in further detail below.

[0062] The present invention provides a method for preparing metal alloy powder for online cladding of a yaw disk of a wind turbine generator set, comprising:

[0063] Step S101, obtaining preparation data, the preparation data including base material data, welding material data, cladding layer requirement data, repair surface data, welding process parameters, environmental parameters, quality inspection data and process monitoring data;

[0064] In step S101, obtaining preparation data is the basic link of the technical solution of the present invention, which provides input for subsequent model training, data comparison and optimization. These data cover multiple aspects of information during the online cladding process of the yaw plate of the wind turbine generator set, including:

[0065] Base material data: the type of base material of the yaw plate, such as steel, alloy, etc. The chemical composition of the base material, which will affect its compatibility with the welding material and the cladding effect. The physical properties of the base material have a direct impact on the performance requirements of the cladding layer.

[0066] Welding material data: The chemical composition of metal alloy powder determines the performance of the cladding layer. The particle size and distribution of alloy powder affect the uniformity and density of the cladding layer. The melting point of alloy powder is an important basis for selecting welding process parameters.

[0067] Cladding layer requirement data: The thickness of the cladding layer needs to be achieved to meet the repair or enhancement requirements. The cladding layer requirement data includes hardness, wear resistance and corrosion resistance, and the performance requirements of the cladding layer so that the repaired yaw disc has sufficient mechanical properties and durability.

[0068] Repair surface data:

[0069] The damage degree includes the damage of the surface of the yaw plate to be repaired, such as cracks, wear, etc.

[0070] Surface roughness includes the roughness of the surface to be repaired, which affects the adhesion and quality of the cladding layer.

[0071] Cleanliness includes the cleanliness of the surface to be repaired, and it must be ensured that there is no oil, impurities, etc. to ensure the cladding effect.

[0072] Welding process parameters:

[0073] Welding current and voltage control the energy input of the welding process and affect the formation and performance of the cladding layer.

[0074] The welding speed determines the deposition rate and cooling rate of the cladding layer, affecting the microstructure and properties.

[0075] The powder feeding rate includes controlling the feeding speed of alloy powder to ensure the uniformity and thickness of the cladding layer.

[0076] Environmental parameters:

[0077] Temperature, humidity and wind speed comprise the environmental conditions during the preparation process and may affect the welding results and the properties of the alloy powder.

[0078] Quality inspection data:

[0079] Hardness testing includes hardness testing of the cladding layer to verify its mechanical properties.

[0080] Metallographic analysis involves observing the microstructure of the cladding layer and evaluating its quality and properties.

[0081] Process monitoring data:

[0082] The fluctuation data of welding current and voltage include data collected in real time during the preparation process, which is used to monitor the stability of the welding process. By comprehensively collecting and analyzing the preparation data, the present invention can more accurately predict and optimize the preparation process of metal alloy powder for online cladding of the yaw disc of the wind turbine generator set, thereby improving the quality and performance of the cladding layer.

[0083] Step S102, dividing the preparation data into a training set and a test set, selecting a random forest as a model, using the training set to train the random forest model, using the cross-validation method to evaluate the model performance under different parameter combinations on the test set data, and selecting the optimal parameter combination to obtain a metal alloy powder preparation prediction model for online cladding of the yaw disc of a wind turbine generator set;

[0084] The prepared data obtained in step S101 is divided into a training set and a test set. The training set is used to train the random forest model so that it can learn the rules and patterns in the data; the test set is used to evaluate the performance of the model to ensure that the model can also show good prediction ability on unseen data.

[0085] Random forest is an integrated learning method that improves the accuracy and stability of the model by building multiple decision trees and combining their prediction results. Random forest performs well in processing high-dimensional data, avoiding overfitting, and handling nonlinear relationships, so it is selected as the prediction model of the present invention.

[0086] The random forest model is trained using the training set. During the training process, the model learns the relationship between the features in the data (such as base material data, welding material data, etc.) and the target variable (such as the quality of the cladding layer), thereby building a model that can predict the quality of the cladding layer.

[0087] In order to evaluate the model performance under different parameter combinations, the cross-validation method is used. Cross-validation is a commonly used model evaluation technique that divides the data set into multiple subsets, takes turns using one of the subsets as the test set and the remaining subsets as the training set, and repeats the training and testing process multiple times. By calculating the average performance of multiple tests, a more reliable model performance evaluation result can be obtained.

[0088] During cross-validation, different combinations of parameters (such as the number of trees, maximum depth, minimum number of samples required for a split, etc.) are tried and the impact of these parameters on the model performance is observed.

[0089] According to the results of cross-validation, the optimal parameter combination is selected. The optimal parameter combination refers to the parameter setting that can achieve the best model performance in a given parameter space. After selecting the optimal parameter combination, the model is retrained using the entire training set to obtain the final prediction model for the preparation of metal alloy powder for online cladding of the yaw disc of the wind turbine generator set.

[0090] Through step S102, the present invention successfully constructs a prediction model based on random forest, which can accurately predict the key parameters and cladding layer quality in the process of preparing metal alloy powder for online cladding of the yaw plate of the wind turbine generator set. This provides a solid foundation for subsequent real-time data monitoring and comparison, model optimization and parameter adjustment.

[0091] Step S103, receiving a preparation task, the preparation task including preparation equipment information and preparation task parameters, sending the task to the device end for execution, collecting preparation data after the device end executes the preparation task, substituting the preparation task into a metal alloy powder preparation prediction model for online cladding of a yaw disk of a wind turbine generator set, outputting an expected value of the preparation data of the metal alloy powder preparation prediction model for online cladding of a yaw disk of a wind turbine generator set, comparing the expected value of the preparation data of the metal alloy powder preparation prediction model for online cladding of a yaw disk of a wind turbine generator set with the preparation data after the device end executes the preparation task, and obtaining a preparation data comparison result;

[0092] The system receives the preparation task, which includes the preparation equipment information and preparation task parameters. The preparation equipment information, such as equipment model and equipment status, is used to ensure that the task can be executed by the correct equipment; the preparation task parameters include key information such as base material type, welding material type, and cladding layer requirements, which are the basis for executing the preparation task.

[0093] The system sends the received preparation tasks to the designated equipment end (such as metal powder making equipment, laser cladding equipment, etc.) for execution. During the execution process, the equipment will perform corresponding operations according to the task parameters, such as preparing metal alloy powder and performing laser cladding.

[0094] When the equipment is performing the preparation task, the system collects preparation data in real time through sensors, data interfaces, etc. These data include key indicators such as the thickness, hardness, chemical composition, and surface quality of the cladding layer, which are used to evaluate the completion of the preparation task.

[0095] Substitute the relevant parameters of the preparation task (such as base material type, welding material type, cladding layer requirements, etc.) into the previously constructed metal alloy powder preparation prediction model for online cladding of the yaw plate of the wind turbine generator set. Based on these input parameters, the model predicts the expected data values ​​of the preparation process, such as the expected cladding layer thickness, hardness, etc.

[0096] The model outputs the predicted expected values ​​of the preparation data, which represent the quality standards that the preparation task should achieve under ideal conditions without external interference and errors.

[0097] Compare the expected value of the preparation data output by the model with the preparation data collected after the equipment actually executes the preparation task. Through the comparison, the difference between the actual preparation result and the expected target can be evaluated, and then the completion of the preparation task can be judged.

[0098] After the comparison process is completed, the system generates preparation data comparison results. These results include the difference between the predicted value and the actual value of each key indicator (such as cladding layer thickness, hardness, etc.), and whether the difference is within an acceptable range.

[0099] Through the execution of step S103, the technical solution of the present invention can realize real-time monitoring and evaluation of the metal alloy powder preparation process of the online cladding of the yaw plate of the wind turbine generator set. Once it is found that there is a significant deviation between the actual preparation result and the expected target, the system can take timely measures to adjust and optimize, thereby ensuring the quality and stability of the preparation process.

[0100] Step S104, matching the preparation task in a preset error range knowledge base to obtain an error range corresponding to the preparation task, and comparing the preparation data comparison result with the error range corresponding to the preparation task to obtain a preparation error range comparison result;

[0101] Step S104 is an important step in ensuring the quality of cladding in the technical solution of the present invention. It evaluates the completion of the preparation task by comparing the actual preparation data with the preset error range. The specific process is as follows:

[0102] The knowledge base pre-stores the acceptable error ranges corresponding to different types of preparation tasks. These error ranges are set based on past experience, industry standards or experimental data and are used to measure the quality standards for the completion of preparation tasks.

[0103] For each received preparation task, it is matched in the error range knowledge base according to its characteristics (such as base material, welding material type, cladding layer requirements, etc.). By extracting the relevant features of the preparation task and comparing them with the entries in the knowledge base, the corresponding error range value is found.

[0104] The preparation data comparison result obtained in step S103 (i.e., the difference between the predicted value and the actual value) is compared with the error range matched in step S104. This step involves comparing the actual difference value of each comparison item (such as cladding layer thickness, hardness, etc.) with the corresponding error range.

[0105] By comparison, it is determined whether the actual difference falls within the preset acceptable error range. If the actual difference does not exceed the error range in all comparison items, it is considered that the quality of the preparation task is completed in accordance with the requirements; otherwise, it is considered that there is a quality problem and further analysis and processing is required.

[0106] Through the comparison and analysis of step S104, the technical solution of the present invention can effectively monitor the quality of the metal alloy powder preparation process of the online cladding of the yaw disk of the wind turbine generator set. Once it is found that the actual preparation data has a significant deviation from the expected and exceeds the acceptable range, the system will trigger the subsequent optimization process (such as step S105) to ensure the quality and stability of the preparation process.

[0107] Step S105, obtaining preparation history data and preparation history tasks. If the preparation error range comparison result exceeds the error range corresponding to the preparation task, the preparation prediction model of the metal alloy powder for online cladding of the yaw disk of the wind turbine generator set is optimized using a genetic algorithm according to the preparation data comparison result, the error of the error range corresponding to the preparation task, and the preparation history data, to obtain an optimized preparation prediction model for online cladding of the yaw disk of the wind turbine generator set. The preparation task is processed using the optimized preparation prediction model for online cladding of the yaw disk of the wind turbine generator set to obtain corrected preparation task parameters, and the corrected preparation task parameters are sent to the device end.

[0108] The system first obtains the past preparation history data and related historical task information. These data and information include the parameter settings of previous preparations, actual execution results, error records, etc., which provide valuable experience and reference for subsequent model optimization.

[0109] According to the preparation error range comparison result obtained in step S104, the system determines whether the error of the current preparation task exceeds the preset acceptable range. If the error is within the acceptable range, no further optimization is required; if it is beyond the range, the optimization process is entered.

[0110] When the preparation error exceeds the acceptable range, the system uses genetic algorithms to optimize the prediction model of metal alloy powder preparation for online cladding of the yaw plate of the wind turbine generator set. Genetic algorithms are an optimization search algorithm that simulates the principles of natural selection and genetics. It continuously iterates and evolves a set of better model parameter combinations through operations such as selection, crossover and mutation.

[0111] During the optimization process, the system combines the current preparation data comparison results, the error range corresponding to the preparation task, and the preparation history data to provide the genetic algorithm with a fitness function and constraints, guiding the algorithm to evolve towards a more accurate prediction direction.

[0112] After multiple rounds of iterative optimization, the genetic algorithm outputs a set of better model parameter combinations. The system retrains the prediction model based on these parameters to obtain an optimized prediction model for the preparation of metal alloy powder for online cladding of the yaw disc of the wind turbine generator set.

[0113] The optimized prediction model is applied to the current preparation task processing. The model outputs more accurate expected values ​​of preparation data based on task parameters and historical data.

[0114] Based on the output of the optimization model and the current preparation error, the system generates revised preparation task parameters. These parameters are designed to reduce or eliminate preparation errors and improve preparation quality. Finally, the system sends the revised preparation task parameters to the device to instruct the device to perform preparation operations according to the new parameter settings.

[0115] Through the execution of step S105, the technical solution of the present invention can realize dynamic optimization and real-time adjustment of the metal alloy powder preparation process for online cladding of the yaw disk of the wind turbine generator set, thereby effectively improving the preparation quality and efficiency.

[0116] Specifically, the method for preparing metal alloy powder for online cladding of a yaw disk of a wind turbine generator set according to the present invention, the step S101 comprises:

[0117] The matrix material data includes the type, composition, hardness and strength of the yaw plate matrix material;

[0118] Welding material data include chemical composition, particle size distribution, and melting point of metal alloy powder;

[0119] The required data of the cladding layer include the required thickness, hardness, wear resistance and corrosion resistance of the cladding layer;

[0120] The repair surface data includes the state of the surface to be repaired of the yaw disk, and the state of the surface to be repaired of the yaw disk includes the degree of damage, surface roughness and cleanliness;

[0121] Welding process parameters include welding current, voltage, welding speed and powder feeding rate;

[0122] Environmental parameters include temperature, humidity, and wind speed during the preparation process;

[0123] Quality inspection data includes hardness test and metallographic analysis;

[0124] The process monitoring data includes data collected in real time during the preparation process, and the data collected in real time during the preparation process includes fluctuation data of welding current and voltage.

[0125] Specifically, the method for preparing metal alloy powder for online cladding of a yaw disk of a wind turbine generator set according to the present invention, the step S102 comprises:

[0126] Preprocessing the obtained preparation data, and dividing the preprocessed preparation data into a training set and a test set;

[0127] The random forest model is trained using the training set data. During the training process, the base material data and welding material data are used as input features, and the corrected preparation task parameters are used as output targets.

[0128] In the cross-validation process, different random forest parameter combinations are used, and the performance indicators of different random forest parameter combinations on the test set are calculated, and the parameter combination with the best performance indicator is selected as the parameter of the final model.

[0129] First, the prepared data obtained from the device or other data sources is preprocessed. This includes steps such as data cleaning (such as processing missing values, outliers, etc.), data transformation (such as normalization, standardization, etc.), and feature selection (selecting features that have an important impact on the prediction target) to ensure the quality and consistency of the data.

[0130] The preprocessed data is divided into a training set and a test set. Usually, the training set occupies a larger proportion (such as 70%-80%) and is used to train the random forest model; the test set is used to evaluate the performance of the model.

[0131] The random forest model is trained using the training set data. In this process, the base material data (such as material type, composition ratio, etc.) and welding material data (such as welding material type, diameter, composition, etc.) are used as input features, and the corrected preparation task parameters (such as cladding layer thickness, hardness, chemical composition, etc.) are used as output targets. The random forest model will build a model that can predict the target parameters by learning the relationship between these input features and the target parameters.

[0132] During the training process, in order to evaluate the generalization ability of the model on different data subsets, the cross-validation method is used. Specifically, the training set is further divided into multiple subsets (such as k subsets in k-fold cross-validation), and one subset is selected as the validation set each time, and the remaining subsets are used as the training set for model training. This process is repeated k times, and a different subset is selected as the validation set each time.

[0133] In each cross-validation, try different random forest parameter combinations (such as the number of trees, maximum depth, minimum number of samples, etc.), and record the performance indicators of each parameter combination on the validation set (such as accuracy, recall, F1 score, etc.). Finally, select the best parameter combination as the parameters of the final model based on the performance indicators. This usually means selecting those parameter combinations with the best average performance in all cross-validation rounds.

[0134] Through the above steps, a random forest prediction model can be constructed for the metal alloy powder preparation process of online cladding of the yaw disk of the wind turbine generator set in the large-span abandoned mine cable structure photovoltaic power station system.

[0135] Specifically, the method for preparing metal alloy powder for online cladding of a yaw disk of a wind turbine generator set according to the present invention, the step S103 comprises:

[0136] Receive preparation tasks. Preparation equipment information includes equipment model and equipment status. Preparation task parameters include base material type, welding material type and cladding layer requirements.

[0137] Send the preparation task to the designated equipment end, which includes metal powder making equipment, powder processing equipment, laser cladding equipment, powder feeding system, grinder, grinder, cladding robot, welding equipment and sensor, and start the equipment to execute the preparation task;

[0138] During the equipment's preparation task, actual preparation data is collected through sensors and data interfaces. The actual preparation data includes the thickness, hardness, chemical composition and surface quality of the cladding layer.

[0139] Establish a data comparison item for the expected value of the preparation data of the metal alloy powder preparation prediction model for online cladding of the yaw plate of the wind turbine generator set and the preparation data after the equipment side executes the preparation task;

[0140] For each comparison item, the difference between the predicted value and the actual value is calculated to obtain the preparation data comparison result.

[0141] The system receives a preparation task, which contains detailed preparation equipment information and preparation task parameters. The preparation equipment information usually includes the equipment model and equipment status to ensure that the task can be executed by the correct equipment. The preparation task parameters cover key information such as base material type, welding material type and cladding layer requirements, which are the basis for executing the preparation task.

[0142] The system sends the received preparation tasks to the designated equipment. These equipment may include metal powder making equipment, powder processing equipment, laser cladding equipment, powder feeding system, grinder, grinding machine, cladding robot, welding equipment and sensors, etc., which together constitute the hardware foundation of the preparation process. The system starts these devices and enables them to start executing the preparation tasks according to the task requirements.

[0143] When the equipment is performing the preparation task, the system collects the actual preparation data in real time through sensors and data interfaces. These data cover key indicators such as the thickness, hardness, chemical composition and surface quality of the cladding layer, which directly reflect the actual execution effect of the preparation task.

[0144] The system establishes data comparison items for the expected value of the preparation data of the metal alloy powder preparation prediction model for the online cladding of the yaw plate of the wind turbine generator set and the actual preparation data collected after the equipment side executes the preparation task. These comparison items correspond one to one, which is convenient for subsequent difference calculation.

[0145] For each comparison item, the system calculates the difference between the predicted value and the actual value. This usually involves mathematical operations, such as finding absolute values, calculating percentage errors, etc., to quantify the deviation between the predicted value and the actual value. After the calculation is completed, the system generates preparation data comparison results, which list the difference values ​​of each comparison item in detail, providing a basis for subsequent error analysis and model optimization.

[0146] Through the above steps, step S103 realizes the execution of the preparation task, the collection of actual preparation data, and the comparison between the predicted value and the actual value.

[0147] Specifically, the method for preparing metal alloy powder for online cladding of a yaw disk of a wind turbine generator set according to the present invention, the step S104 comprises:

[0148] Loading a preset error range knowledge base, the preset error range knowledge base includes acceptable error ranges corresponding to different types of preparation tasks;

[0149] Perform feature extraction on the preparation task to obtain the required features of the base material, welding material type and cladding layer, match the corresponding error range in the error range knowledge base, and extract the error range value from the matched error range knowledge base entry;

[0150] Compare the preparation data comparison results with the error range corresponding to the preparation task. For each comparison item, compare the actual difference value with the corresponding error range to determine whether the actual difference falls within the acceptable range.

[0151] Load the preset error range knowledge base. The system first loads the preset error range knowledge base. This knowledge base contains the acceptable error ranges set for different types of preparation tasks. These error ranges are determined based on historical data, industry standards, and actual application requirements, and are used to evaluate whether the actual execution effect of the current preparation task meets the standards.

[0152] Next, the system extracts features from the received preparation tasks. These features usually include key information such as base material, welding material type, and cladding layer requirements. Then, the system matches these features in the error range knowledge base and finds the error range knowledge base entry that best matches the current preparation task.

[0153] The corresponding error range values ​​are extracted from the matching error range knowledge base entries. These error range values ​​will be used as the benchmark for evaluating the performance of the current preparation task.

[0154] The preparation data comparison result obtained in step S103 is compared with the error range matched in step S104. For each comparison item (such as cladding layer thickness, hardness, chemical composition, surface quality, etc.), the system compares the actual difference value (i.e., the difference between the predicted value and the actual value) with the corresponding error range.

[0155] Determine whether the actual difference value falls within the acceptable error range. If the actual difference value is within the error range, it is considered that the execution effect of the current preparation task meets the standard; if the actual difference value exceeds the error range, it is considered that the execution effect does not meet the standard, and further analysis of the reasons and corresponding measures need to be taken to improve it.

[0156] Through the above steps, step S104 realizes the comparison and evaluation of the error range of the preparation task execution effect. This process not only helps to ensure that the execution quality of the preparation task meets the preset standards, but also provides an important basis for subsequent improvement and optimization.

[0157] Specifically, the method for preparing metal alloy powder for online cladding of a yaw disk of a wind turbine generator set according to the present invention, the step S105 comprises:

[0158] Acquire preparation history data, which includes preparation task parameters, actual preparation data, and corresponding preparation error range comparison results;

[0159] Receive and set the optimization target, and define the optimization target as minimizing the difference between the predicted value and the actual value based on the preparation error range comparison result;

[0160] Determine the basic parameters of the genetic algorithm, which include population size, number of iterations, crossover probability, and mutation probability;

[0161] The parameters of the prediction model for the preparation of metal alloy powder for online cladding of the yaw disk of a wind turbine generator set are encoded into genotypes that can be processed by a genetic algorithm;

[0162] Initialize the population and randomly generate a set of initial model parameters as the population of the genetic algorithm;

[0163] Each individual represents a model parameter combination;

[0164] Evaluate fitness, use the optimization objective function to evaluate the fitness of each individual, that is, calculate the corresponding model prediction error;

[0165] Selection operation, which selects a part of individuals as parents according to their fitness to generate the next generation;

[0166] Crossover operation: perform crossover operation on the selected parent individuals to generate new offspring individuals;

[0167] Mutation operation: perform mutation operation on offspring individuals to introduce new genotypes;

[0168] The mutation operation can randomly change the gene positions in individuals, replace some or all of the parent individuals with newly generated offspring individuals, form a new population, and repeat the fitness evaluation, selection, crossover and mutation operations until the predetermined number of iterations or optimization goal is reached;

[0169] The best individual is selected, and the individual with the highest fitness is selected in the last generation of the population, that is, the optimized model parameter combination, and the optimized model parameter combination is used to update the metal alloy powder preparation prediction model for online cladding of the yaw disk of the wind turbine generator set.

[0170] Obtain preparation history data: Obtain preparation history data from the database, which includes preparation task parameters (such as base material type, welding material type, cladding layer requirements, etc.), actual preparation data (such as actual thickness, hardness, chemical composition, surface quality, etc. of the cladding layer) and corresponding preparation error range comparison results.

[0171] Receive and set optimization goals: Based on the comparison results of the preparation error range, the optimization goal is to minimize the difference between the predicted value and the actual value. This means that we hope to make the model's prediction results closer to the actual preparation data by optimizing the model parameters.

[0172] Determine the basic parameters of the genetic algorithm: Set the basic parameters of the genetic algorithm, including population size (i.e. the number of model parameter combinations processed simultaneously), number of iterations (i.e. the number of rounds the algorithm runs), crossover probability (i.e. the probability that two parent individuals cross to generate offspring individuals), and mutation probability (i.e. the probability that individual genes mutate).

[0173] Model parameter encoding: Encode the parameters of the prediction model for the online cladding of the metal alloy powder of the wind turbine generator set yaw disc into genotypes that can be processed by the genetic algorithm. This usually involves discretizing continuous parameters or using binary encoding.

[0174] Initialize the population: Randomly generate a set of initial model parameters as the population of the genetic algorithm. Each individual represents a specific combination of model parameters.

[0175] Evaluate fitness:

[0176] The fitness of each individual is evaluated using an optimization objective function (i.e., the difference between the predicted value and the actual value). This usually involves calculating the model prediction error and ranking or assigning a score to the individuals based on the size of the error.

[0177] Selection operation: select a part of individuals as parents according to their fitness to generate the next generation. This usually adopts strategies such as roulette selection and tournament selection to ensure that excellent individuals have more opportunities to be selected.

[0178] Crossover operation: Perform a crossover operation on the selected parent individuals to generate new offspring individuals. This usually involves exchanging some gene segments of the parent individuals to combine individuals with new characteristics.

[0179] Mutation operation: mutation operation is performed on offspring individuals to introduce new genotypes. This usually involves randomly changing one or some gene positions in individuals to increase the diversity of the population.

[0180] Iterative optimization: Replace some or all of the parent individuals with the newly generated offspring individuals to form a new population. Then repeat the fitness evaluation, selection, crossover and mutation operations until the predetermined number of iterations or optimization target (such as the prediction error is less than a certain threshold) is reached.

[0181] Select the best individual: Select the individual with the highest fitness in the last generation of the population, that is, the optimized model parameter combination. These parameter combinations represent the best solution after optimization by the genetic algorithm.

[0182] Update prediction model: Use the optimized model parameter combination to update the prediction model for the metal alloy powder preparation of the online cladding of the yaw plate of the wind turbine generator set. In this way, the model can more accurately predict the actual preparation results and provide stronger support for subsequent preparation tasks.

[0183] Through the above steps, step S105 implements a model optimization process based on a genetic algorithm, aiming to improve the accuracy of the prediction model for the preparation of metal alloy powder for online cladding of the yaw disk of a wind turbine generator set.

[0184] Specifically, the method for preparing metal alloy powder for online cladding of a yaw disk of a wind turbine generator set according to the present invention, the step S105 comprises:

[0185] Preprocessing the received preparation task data, and inputting the preprocessed preparation task data into the optimized metal alloy powder preparation prediction model for online cladding of the yaw plate of the wind turbine generator set;

[0186] The optimized prediction model for the preparation of metal alloy powder for online cladding of the yaw disc of the wind turbine generator set predicts the parameters of the preparation process based on the input data. The parameters predicted by the data include the thickness, hardness and chemical composition of the cladding layer.

[0187] Based on the prediction results of the optimized prediction model for the online cladding of the yaw disk of the wind turbine generator set, the preparation task parameters are corrected and sent to the equipment end.

[0188] The prediction model is applied and the task parameters are modified to preprocess the received preparation task data. The preprocessing steps may include data cleaning (such as removing missing values ​​and outlier processing), data transformation (such as normalization and standardization), and feature selection to ensure the quality and consistency of the input data, thereby improving the accuracy and reliability of the prediction model.

[0189] The preprocessed preparation task data is input into the optimized prediction model for the preparation of metal alloy powder for online cladding of the yaw plate of the wind turbine generator set. The model predicts the parameters of the preparation process based on the input data, which usually include key indicators such as the thickness, hardness and chemical composition of the cladding layer.

[0190] Carefully analyze the output of the prediction model. These results not only provide predicted values ​​of the preparation parameters, but may also contain information such as the uncertainty or confidence interval of the prediction, which helps to further evaluate the reliability of the prediction results.

[0191] Based on the prediction results of the prediction model, necessary corrections are made to the original preparation task parameters, which may involve adjusting the thickness of the cladding layer, selecting the appropriate type of welding material, optimizing the cladding process parameters, etc., to ensure that the preparation process can meet the expected quality requirements and performance indicators.

[0192] The modified preparation task parameters are sent to the equipment side, which adjusts the preparation process and procedures based on these parameters to ensure that the preparation task can be executed according to the optimized plan.

[0193] During the execution of the preparation task, the deviation between the actual preparation data and the predicted data is continuously monitored. If the deviation exceeds a preset threshold, the process of steps S103 to S105 may need to be re-executed to further optimize and adjust the prediction model.

[0194] By implementing step S105, the present invention not only improves the execution efficiency and accuracy of the preparation task, but also adapts to the changing requirements of different preparation tasks by continuously optimizing the prediction model, thereby realizing the intelligence and automation of the preparation process.

[0195] The technical solution of the present invention solves the problem that the prior art cannot detect the data accuracy of each process in the preparation process of metal alloy powder for online cladding of the yaw disk of a wind turbine generator set, thereby affecting the cladding quality through the following steps:

[0196] Acquire multi-dimensional data, including base material data, welding material data, cladding layer requirement data, etc., and perform data cleaning, conversion and feature selection. Use the random forest model to build a prediction model, evaluate the model performance through cross-validation method, select the optimal parameter combination, and obtain the preparation prediction model. Collect data in real time when performing preparation tasks, and compare it with the expected value of the prediction model to obtain the preparation data comparison result. Match the preset error range to evaluate whether the preparation error is within the acceptable range. If the preparation error exceeds the acceptable range, the model optimization process is triggered, and the prediction model is optimized using a genetic algorithm and the model parameters are updated. Continuously monitor data changes throughout the preparation process, and fine-tune the model according to actual conditions to ensure that the model can always accurately predict and guide the preparation process.

[0197] Through these steps, the present invention realizes intelligent monitoring and optimization of the preparation process, ensuring the data accuracy and cladding quality of the preparation process.

Claims

1. A method for preparing metal alloy powder for online cladding of a yaw disc of a wind turbine generator set, characterized in that: include: Step S101, obtaining preparation data, the preparation data including base material data, welding material data, cladding layer requirement data, repair surface data, welding process parameters, environmental parameters, quality inspection data and process monitoring data; Step S102, dividing the preparation data into a training set and a test set, selecting a random forest as a model, using the training set to train the random forest model, using the cross-validation method to evaluate the model performance under different parameter combinations on the test set data, and selecting the optimal parameter combination to obtain a metal alloy powder preparation prediction model for online cladding of the yaw disc of a wind turbine generator set; Step S103, receiving a preparation task, the preparation task including preparation equipment information and preparation task parameters, sending the task to the device end for execution, collecting preparation data after the device end executes the preparation task, substituting the preparation task into a metal alloy powder preparation prediction model for online cladding of a yaw disk of a wind turbine generator set, outputting an expected value of the preparation data of the metal alloy powder preparation prediction model for online cladding of a yaw disk of a wind turbine generator set, comparing the expected value of the preparation data of the metal alloy powder preparation prediction model for online cladding of a yaw disk of a wind turbine generator set with the preparation data after the device end executes the preparation task, and obtaining a preparation data comparison result; Step S104, matching the preparation task in a preset error range knowledge base to obtain an error range corresponding to the preparation task, and comparing the preparation data comparison result with the error range corresponding to the preparation task to obtain a preparation error range comparison result; Step S105, obtaining preparation history data and preparation history tasks. If the preparation error range comparison result exceeds the error range corresponding to the preparation task, the preparation prediction model of the metal alloy powder for online cladding of the yaw disk of the wind turbine generator set is optimized using a genetic algorithm according to the preparation data comparison result, the error of the error range corresponding to the preparation task, and the preparation history data, to obtain an optimized preparation prediction model for online cladding of the yaw disk of the wind turbine generator set. The preparation task is processed using the optimized preparation prediction model for online cladding of the yaw disk of the wind turbine generator set to obtain corrected preparation task parameters, and the corrected preparation task parameters are sent to the device end.

2. The method for preparing metal alloy powder for online cladding of the yaw disk of a wind turbine generator set according to claim 1, characterized in that: The step S101 includes: The matrix material data includes the type, composition, hardness and strength of the yaw plate matrix material; Welding material data include chemical composition, particle size distribution, and melting point of metal alloy powder; The required data of the cladding layer include the required thickness, hardness, wear resistance and corrosion resistance of the cladding layer; The repair surface data includes the state of the surface to be repaired of the yaw disk, and the state of the surface to be repaired of the yaw disk includes the degree of damage, surface roughness and cleanliness; Welding process parameters include welding current, voltage, welding speed and powder feeding rate; Environmental parameters include temperature, humidity, and wind speed during the preparation process; Quality inspection data includes hardness test and metallographic analysis; The process monitoring data includes data collected in real time during the preparation process, and the data collected in real time during the preparation process includes fluctuation data of welding current and voltage.

3. The method for preparing metal alloy powder for online cladding of the yaw disk of a wind turbine generator set according to claim 1, characterized in that: The step S102 includes: Preprocessing the obtained preparation data, and dividing the preprocessed preparation data into a training set and a test set; The random forest model is trained using the training set data. During the training process, the base material data and welding material data are used as input features, and the corrected preparation task parameters are used as output targets. In the cross-validation process, different random forest parameter combinations are used, and the performance indicators of different random forest parameter combinations on the test set are calculated, and the parameter combination with the best performance indicator is selected as the parameter of the final model.

4. The method for preparing metal alloy powder for online cladding of a yaw disk of a wind turbine generator set according to claim 1, characterized in that: The step S103 includes: Receive preparation tasks. Preparation equipment information includes equipment model and equipment status. Preparation task parameters include base material type, welding material type and cladding layer requirements. Send the preparation task to the designated equipment end, which includes metal powder making equipment, powder processing equipment, laser cladding equipment, powder feeding system, grinder, grinder, cladding robot, welding equipment and sensor, and start the equipment to execute the preparation task; During the equipment's preparation task, actual preparation data is collected through sensors and data interfaces. The actual preparation data includes the thickness, hardness, chemical composition and surface quality of the cladding layer. Establish a data comparison item for the expected value of the preparation data of the metal alloy powder preparation prediction model for online cladding of the yaw plate of the wind turbine generator set and the preparation data after the equipment side executes the preparation task; For each comparison item, the difference between the predicted value and the actual value is calculated to obtain the preparation data comparison result.

5. The method for preparing metal alloy powder for online cladding of a yaw disk of a wind turbine generator set according to claim 1, characterized in that: The step S104 includes: Loading a preset error range knowledge base, the preset error range knowledge base includes acceptable error ranges corresponding to different types of preparation tasks; Perform feature extraction on the preparation task to obtain the required features of the base material, welding material type and cladding layer, match the corresponding error range in the error range knowledge base, and extract the error range value from the matched error range knowledge base entry; Compare the preparation data comparison results with the error range corresponding to the preparation task. For each comparison item, compare the actual difference value with the corresponding error range to determine whether the actual difference falls within the acceptable range.

6. The method for preparing metal alloy powder for online cladding of a yaw disk of a wind turbine generator set according to claim 1, characterized in that: The step S105 includes: Acquire preparation history data, which includes preparation task parameters, actual preparation data, and corresponding preparation error range comparison results; Receive and set the optimization target, and define the optimization target as minimizing the difference between the predicted value and the actual value based on the preparation error range comparison result; Determine the basic parameters of the genetic algorithm, which include population size, number of iterations, crossover probability, and mutation probability; The parameters of the prediction model for the preparation of metal alloy powder for online cladding of the yaw disk of a wind turbine generator set are encoded into genotypes that can be processed by a genetic algorithm; Initialize the population and randomly generate a set of initial model parameters as the population of the genetic algorithm; Each individual represents a combination of model parameters; Evaluate fitness, use the optimization objective function to evaluate the fitness of each individual, that is, calculate the corresponding model prediction error; Selection operation, which selects a part of individuals as parents according to their fitness to generate the next generation; Crossover operation: perform crossover operation on the selected parent individuals to generate new offspring individuals; Mutation operation: perform mutation operation on offspring individuals to introduce new genotypes; The mutation operation can randomly change the gene positions in individuals, replace some or all of the parent individuals with newly generated offspring individuals, form a new population, and repeat the fitness evaluation, selection, crossover and mutation operations until the predetermined number of iterations or optimization goal is reached; The best individual is selected, and the individual with the highest fitness is selected in the last generation of the population, that is, the optimized model parameter combination, and the optimized model parameter combination is used to update the metal alloy powder preparation prediction model for online cladding of the yaw disk of the wind turbine generator set.

7. The method for preparing metal alloy powder for online cladding of a yaw disk of a wind turbine generator set according to claim 6, characterized in that: The step S105 includes: Preprocessing the received preparation task data, and inputting the preprocessed preparation task data into the optimized metal alloy powder preparation prediction model for online cladding of the yaw plate of the wind turbine generator set; The optimized prediction model for the preparation of metal alloy powder for online cladding of the yaw disc of the wind turbine generator set predicts the parameters of the preparation process based on the input data. The parameters predicted by the data include the thickness, hardness and chemical composition of the cladding layer. Based on the prediction results of the optimized prediction model for the online cladding of the yaw disk of the wind turbine generator set, the preparation task parameters are corrected and sent to the equipment end.

Images