Laser derusting method based on machine learning

Abstract

The invention discloses a laser rust removal method based on machine learning. The method comprises the following steps: firstly, collecting rust removal experiment data of metals of different materials under different laser rust removal parameters, covering data such as metal surface images, corrosion degrees and matrix damage before and after rust removal, and preprocessing; secondly, features related to the rust removal effect and matrix damage are extracted, and key features are screened through a feature selection algorithm; and selecting a proper machine learning model, taking the key features as input, taking the derusting effect evaluation indexes as output training models, and adopting cross validation optimization. In the actual rust removal process, data collected by a sensor is combined with the model to conduct online monitoring and prediction, and parameters are dynamically adjusted through an optimization algorithm according to the result. And evaluating the effect after rust removal, comparing the actual and predicted results, and updating the model when the deviation is large. According to the method, the rust removal effect can be accurately controlled, the base body is protected, the efficiency is improved, self-optimization is achieved, and the method has wide application prospects.

Description

Technical Field

[0001] The present invention relates to the field of laser rust removal, and in particular to a laser rust removal method based on machine learning. Background Art

[0002] In many fields such as industrial production, cultural relic restoration, and building maintenance, metal surface rust removal is a common and important task. Traditional rust removal methods, such as mechanical grinding and chemical rust removal, have many limitations. Mechanical grinding can easily damage the metal substrate and is difficult to handle workpieces with complex shapes and fine structures; chemical rust removal may cause environmental pollution and require subsequent waste liquid treatment processes.

[0003] As an emerging rust removal technology, laser rust removal has the advantages of non-contact, wear-free, high efficiency and environmental protection, and has gradually attracted widespread attention. However, the laser rust removal process is affected by many factors, such as laser power, scanning speed, pulse frequency, spot diameter, and the degree of rust on the metal surface, material properties, etc. How to accurately control these parameters to achieve the best rust removal effect while avoiding excessive damage to the substrate is a key issue facing current laser rust removal technology.

[0004] Machine learning technology has a strong ability to handle complex multivariate relationships and optimize decision-making. By learning a large amount of laser rust removal experimental data or actual rust removal case data, the machine learning model can establish an accurate mapping relationship between the rust removal effect and various influencing factors, thereby providing a scientific basis and intelligent guidance for the optimization of laser rust removal parameters. Summary of the invention

[0005] The purpose of the present invention is to provide a laser rust removal method based on machine learning to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a laser rust removal method based on machine learning, comprising the following steps:

[0007] Step 1, data collection and preprocessing: collect a large number of rust removal experimental data of metal surfaces of different materials under different laser rust removal parameters, including laser power, scanning speed, pulse frequency, and spot diameter. The rust removal experimental data includes image data of the metal surface before and after rust removal, rust degree data, and matrix damage data; clean and normalize the collected data to improve the quality and availability of the data;

[0008] Step 2, feature extraction and selection: Extract features related to laser rust removal effect and substrate damage from the preprocessed data, including laser parameter features, metal surface features, and surface quality features after rust removal; use feature selection algorithm to screen out key features that have a significant impact on laser rust removal effect and substrate damage, reduce data dimensions, and improve model training efficiency;

[0009] Step 3, model training: select a neural network model and use the features screened in step 2 as input, and use the rust removal effect evaluation index as output to train the model; use the cross-validation method to adjust the hyperparameters of the model and optimize the performance of the model so that the model can accurately predict the rust removal effect and substrate damage under different laser parameters;

[0010] Step 4: Online monitoring and parameter optimization: During the actual laser rust removal process, sensors are used to collect relevant data on the metal surface in real time, and the current rust removal effect and substrate damage are monitored and predicted online in combination with the neural network model pre-trained in step 3; according to the prediction results, the laser rust removal parameters are dynamically adjusted through the optimization algorithm to achieve the best rust removal effect and the minimum substrate damage;

[0011] Step 5. Effect evaluation and model update: After completing the laser rust removal operation, conduct a comprehensive evaluation of the rust removal effect. Specifically, use optical microscopes and scanning electron microscopes to detect the microscopic morphology and residual rust amount of the metal surface after rust removal. Compare and analyze the actual evaluation results with the model prediction results. If there is a large deviation between the two, add the rust removal data to the training data set, and update and optimize the machine learning model to continuously improve the accuracy and adaptability of the model.

[0012] Preferably, the specific implementation steps of data collection and preprocessing in step 1 are as follows:

[0013] Step 11, Experimental design and sample preparation: Select experimental samples for a variety of common metal materials; for each experimental sample, prepare samples with different degrees of corrosion. The degree of corrosion is controlled and adjusted by natural exposure time and artificial accelerated corrosion methods to cover various conditions from mild to severe corrosion;

[0014] Step 12, image data acquisition: using a high-resolution industrial camera or optical microscope to take images of the metal surface before and after rust removal to obtain image data;

[0015] Step 13, Corrosion degree data collection: chemical analysis is used to quantify the degree of corrosion, specifically, the mass loss per unit area of ​​the rusted sample before rust removal is measured to determine the amount of corrosion, a high-precision electronic balance is used to weigh the mass change of the sample before and after rust removal, and the mass thickness of the corrosion is calculated in combination with the sample surface area;

[0016] Step 14, matrix damage data collection: use a scanning electron microscope to observe the microscopic morphology of the metal matrix surface after rust removal, and check whether there are microcracks, fused layer, and heat-affected zone matrix damage characteristics; for existing damage, quantify it by measuring the depth and width of the damaged area and calculating the area ratio of the damaged area to the entire rust removal area; at the same time, use a hardness test method to compare the change in the hardness of the metal matrix before and after rust removal, and indirectly evaluate the degree of matrix damage. The hardness test uses a microhardness tester to perform multi-point measurements in the rust removal area and the untreated area and take the average value;

[0017] Step 15, outlier processing: For image data, check whether there are blurred, missing or obviously wrong pixel areas due to reasons such as camera failure, abnormal lighting, etc.; for corrosion degree data and matrix damage data, if the measured value obviously deviates from the normal range, it is necessary to further check the original measurement record or repeat the measurement to determine the accuracy of the data;

[0018] Step 16, normalizing image data: for the image data collected in the above steps, normalizing the pixel values ​​of the image, specifically mapping the pixel values ​​of the image to the interval [0, 1];

[0019] Step 17, normalization of numerical data: the corrosion degree data, substrate damage data and laser rust removal parameter data are normalized using the Min-Max normalization or Z-score normalization method.

[0020] Preferably, the specific steps of feature extraction and selection in step 2 are as follows:

[0021] Step 21: Feature extraction:

[0022] a. Laser parameter features: directly extract the laser rust removal parameters set in data collection as features;

[0023] b. Metal surface characteristics:

[0024] Roughness feature: Use a roughness measuring instrument to measure the metal surface before rust removal to obtain surface roughness parameters; Texture feature: Extract texture features from the image data of the metal surface before rust removal;

[0025] c. Surface quality characteristics after rust removal:

[0026] Residual rust amount characteristics: The residual rust amount on the metal surface after rust removal is determined by chemical analysis. Chemical analysis is an improvement on the mass loss measurement method. The residual rust amount is determined by weighing the sample again after rust removal and calculating the difference with the ideal mass after rust removal. Surface flatness characteristics: The flatness change of the metal surface after rust removal is measured using a profilometer or optical interferometer.

[0027] Step 22, feature selection: principal component analysis, first, calculate the covariance matrix of the feature data in step 21; then, solve the eigenvalues ​​and eigenvectors of the covariance matrix, the eigenvalues ​​represent the variance of the data in the direction of the corresponding eigenvectors; finally, sort the eigenvectors according to the eigenvalues, select the first k eigenvectors, and the eigenvectors constitute the principal component transformation matrix, that is, the k eigenvectors are the key features after screening.

[0028] Preferably, the specific implementation steps of the model training in step 3 are as follows:

[0029] Step 31, neural network model structure selection and initialization: select a multi-layer feedforward neural network, including an input layer, a hidden layer and an output layer; the number of input layer nodes is equal to the number of features after screening, specifically, after feature selection, the four key features of laser power, scanning speed, metal surface roughness, and residual rust amount are determined, then the number of input layer nodes is 4, and the number of output layer nodes is 2, corresponding to the rust removal rate and substrate damage degree in the rust removal effect evaluation index respectively; the number of hidden layers and the number of nodes need to be determined based on experience and experiments;

[0030] Step 32, forward propagation: Let the input feature vector be X = [x 1 , x 2 , x 3 , x 4 ], where x 1 is the laser power, x 2 is the scanning speed, x 3 is the metal surface roughness, x 4 is the residual corrosion amount. For the calculation from the input layer to the hidden layer, the input net of the jth node in the hidden layer is j The calculation formula is:

[0031]

[0032] where ω ij is the weight from the i-th node in the input layer to the j-th node in the hidden layer, b j is the bias of the jth node in the hidden layer;

[0033] The output h of the hidden layer node j The activation function is used for calculation. Here we choose the commonly used sigmoid activation function: For the calculation from the hidden layer to the output layer, the input net′ of the kth node in the output layer k The calculation formula is:

[0034]

[0035] where ω′ jkis the weight from the jth node in the hidden layer to the kth node in the output layer, b′ k is the bias of the nth node in the output layer;

[0036] The output y of the output layer node k The activation function is also used. For the rust removal rate, the sigmoid function is used to map the output to the [0, 1] interval, indicating the proportion of rust removal; for the matrix damage degree, the linear activation function is used, and the output is a numerical value, indicating the quantified value of the damage degree;

[0037] Step 33, loss function calculation: The mean square error is used as the loss function to measure the difference between the model prediction value and the true value. is the true value of the kth output, that is, the rust removal rate and substrate damage degree actually measured, yk is the model prediction value, and the loss function L is calculated as:

[0038] Step 34, back propagation and weight update: The gradient of the loss function to the weights and biases of each layer is calculated through the back propagation algorithm, and then the weights and biases are updated according to the gradient descent algorithm, where the weight ω′ from the hidden layer to the output layer jk The update formula is:

[0039]

[0040] Where η is the learning rate, which is a hyperparameter that controls the step size of weight updates. is the loss function for ω′ jk The gradient of is calculated step by step from the output layer to the hidden layer through the chain rule;

[0041] Step 35, cross-validation to adjust hyperparameters: divide the collected laser rust removal experimental data into k subsets, select k-1 subsets as training sets each time, and the remaining 1 subset as the validation set; select the hyperparameter combination with the smallest loss function value on the validation set as the final model hyperparameter, and then use the entire training data set to retrain the model under this hyperparameter combination to obtain the final trained neural network model.

[0042] Preferably, the specific implementation contents of the online monitoring and parameter optimization of step 4 are as follows:

[0043] A. Sensor data collection and transmission

[0044] Laser parameter sensor: Install a high-precision laser power sensor on the laser rust removal equipment to monitor the laser output power in real time; at the same time, install a scanning speed sensor to accurately measure the moving speed of the laser scanning head through encoders and other technical means;

[0045] Surface state sensor: Use infrared temperature sensor to measure the real-time temperature change of metal surface under the action of laser. At the same time, use optical reflectivity sensor to detect the intensity of reflected light after laser irradiates the metal surface. The intensity of reflected light will be different under different rust degree and metal matrix state. By monitoring the reflectivity, it can assist in judging the progress of rust removal.

[0046] B. Data preprocessing and feature extraction

[0047] Perform real-time preprocessing of collected sensor data, including noise removal;

[0048] Extract real-time features corresponding to the input features of the training model, directly obtain the current laser power value from the laser power sensor data, obtain the scanning speed value from the scanning speed sensor, extract the real-time temperature features of the metal surface from the infrared temperature sensor data, extract the reflectivity features from the optical reflectivity sensor data, etc. The above features will be used as input for subsequent model predictions;

[0049] C. Prediction based on machine learning model: The real-time feature data after preprocessing and extraction is input into the pre-trained neural network model in step 3. The model predicts the current rust removal rate and substrate damage degree based on the input feature data and the relationship between the features learned in the training phase and the rust removal effect and substrate damage;

[0050] D. Optimization algorithm and parameter adjustment

[0051] Optimization target setting: The optimization target is to improve the rust removal rate while ensuring that the degree of substrate damage is lower than the set threshold;

[0052] Selection and application of optimization algorithm: Genetic algorithm is used for parameter optimization. First, the laser power, scanning speed, metal surface roughness, and residual rust parameters are encoded into chromosome form. Then, a population is initialized, which contains multiple parameter combination individuals encoded in this way. The predicted rust removal rate and substrate damage degree corresponding to each individual are calculated. The fitness function is set according to the optimization goal. Through selection, crossover, and mutation genetic operations, new parameter combination individuals are continuously iterated to find the parameter combination with the highest fitness.

[0053] E. Parameter adjustment implementation: The optimized laser rust removal parameters are sent to the controller of the laser rust removal equipment. The controller adjusts the power output of the laser generator and the speed of the scanning motor according to the received parameter instructions to achieve dynamic adjustment of the laser rust removal parameters. After adjusting the parameters, continue to collect data through sensors and repeat the above monitoring, prediction and optimization process until a satisfactory rust removal effect is achieved and the substrate damage is within a controllable range.

[0054] Preferably, the specific contents of comparing and analyzing the actual evaluation results with the model prediction results in step 5 are as follows: the rust removal effect and substrate damage degree data obtained by evaluation are compared with the corresponding results predicted by the machine learning model in the laser rust removal process, and the deviation value is specifically calculated. For the rust removal rate deviation: ΔRR = |RR 实际 -RR 预测 |, for the matrix damage degree deviation ΔBD=|BD 实际 -BD 预测 |, set a reasonable deviation threshold, compare the calculated deviation value with the threshold, and judge that the deviation is large, and the machine learning model needs to be updated; if the deviation is within the threshold range, it is considered that the current model prediction effect is good, and the model can be temporarily not updated.

[0055] Compared with the prior art, the beneficial effect of the present invention is that the present invention can accurately establish the complex relationship between the rust removal effect and many parameters such as laser power, scanning speed, pulse frequency, spot diameter, and metal surface characteristics through the learning of a large amount of experimental data and actual case data by machine learning models. In the actual rust removal process, based on the data collected by the sensor in real time, the model can accurately predict the current rust removal effect and dynamically adjust the laser rust removal parameters using an optimization algorithm. This makes the rust removal process no longer rely on empirical parameter settings, but intelligently optimizes according to the real-time status of the metal surface, thereby significantly improving the rust removal rate and ensuring that the metal surface can reach a higher cleanliness standard. For example, when dealing with metal parts with complex rust conditions and mixed materials, the model can quickly adapt and find the best combination of parameters to effectively remove stubborn rust, while traditional methods often find it difficult to achieve such precise parameter matching, resulting in incomplete rust removal or over-treatment; Adapt to a variety of working conditions: Whether facing metal surfaces of different materials (such as carbon steel, stainless steel, aluminum alloy, etc.), or dealing with various degrees of rust (from mild surface oxidation to severe rust layer accumulation), this method can flexibly adjust parameter strategies through the adaptive learning ability of the machine learning model to achieve the best rust removal effect. This greatly expands the application scope of laser rust removal technology, enabling it to be more widely and efficiently used in many industrial fields (such as machinery manufacturing, automobile repair, ship maintenance, etc.) as well as special industries such as cultural relics restoration and architectural decoration, solving the problem of poor adaptability of traditional rust removal methods when dealing with different working conditions;

[0056] At the same time, while the present invention accurately controls the rust removal parameters, the machine learning model can also predict and monitor the damage to the substrate in real time. Through comprehensive analysis of metal surface features, laser parameters, and real-time monitoring data, the model can give early warning of parameter settings that may cause excessive damage to the substrate, and make timely adjustments to minimize the degree of substrate damage. For example, when it is detected that the metal surface temperature is too high or the laser energy density is too high, which may cause substrate melting or microcracks, the model will automatically adjust the scanning speed or reduce the laser power to avoid damage to the structural integrity and performance of the substrate. However, traditional laser rust removal methods with fixed parameter settings are difficult to achieve such precise damage prevention and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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.

[0059] See also Figure 1 The present invention provides a technical solution: a laser rust removal method based on machine learning, comprising the following steps:

[0060] Step 1, data collection and preprocessing: collect a large number of rust removal experimental data of metal surfaces of different materials under different laser rust removal parameters, including laser power, scanning speed, pulse frequency, and spot diameter. The rust removal experimental data includes image data of the metal surface before and after rust removal, rust degree data, and matrix damage data; clean and normalize the collected data to improve the quality and availability of the data;

[0061] The specific implementation steps of data collection and preprocessing are as follows:

[0062] Step 11, Experimental design and sample preparation: Select experimental samples for a variety of common metal materials; for each experimental sample, prepare samples with different degrees of corrosion. The degree of corrosion is controlled and adjusted by natural exposure time and artificial accelerated corrosion methods to cover various conditions from mild to severe corrosion;

[0063] Step 12, image data acquisition: using a high-resolution industrial camera or optical microscope to take images of the metal surface before and after rust removal to obtain image data;

[0064] Step 13, Corrosion degree data collection: chemical analysis is used to quantify the degree of corrosion, specifically, the mass loss per unit area of ​​the rusted sample before rust removal is measured to determine the amount of corrosion, a high-precision electronic balance is used to weigh the mass change of the sample before and after rust removal, and the mass thickness of the corrosion is calculated in combination with the sample surface area;

[0065] Step 14, matrix damage data collection: use a scanning electron microscope to observe the microscopic morphology of the metal matrix surface after rust removal, and check whether there are microcracks, fused layer, and heat-affected zone matrix damage characteristics; for existing damage, quantify it by measuring the depth and width of the damaged area and calculating the area ratio of the damaged area to the entire rust removal area; at the same time, use a hardness test method to compare the change in the hardness of the metal matrix before and after rust removal, and indirectly evaluate the degree of matrix damage. The hardness test uses a microhardness tester to perform multi-point measurements in the rust removal area and the untreated area and take the average value;

[0066] Step 15, outlier processing: For image data, check whether there are blurred, missing or obviously wrong pixel areas due to reasons such as camera failure, abnormal lighting, etc.; for corrosion degree data and matrix damage data, if the measured value obviously deviates from the normal range, it is necessary to further check the original measurement record or repeat the measurement to determine the accuracy of the data;

[0067] Step 16, normalizing image data: for the image data collected in the above steps, normalizing the pixel values ​​of the image, specifically mapping the pixel values ​​of the image to the interval [0, 1];

[0068] Step 17, normalization of numerical data: The rust degree data, substrate damage data and laser rust removal parameter data are normalized using the Min-Max normalization or Z-score normalization method; the normalization method depends on the distribution characteristics of the data and the requirements of the subsequent model. The normalized numerical data can improve the efficiency and accuracy of machine learning model training and avoid numerical instability problems during model training due to large differences in data magnitude.

[0069] Through the above detailed data collection and preprocessing steps, a high-quality and standardized data foundation can be provided for the laser rust removal method based on machine learning, thereby improving the reliability and effectiveness of subsequent model training and prediction.

[0070] Step 2, feature extraction and selection: Extract features related to laser rust removal effect and substrate damage from the preprocessed data, including laser parameter features, metal surface features, and surface quality features after rust removal; use feature selection algorithm to screen out key features that have a significant impact on laser rust removal effect and substrate damage, reduce data dimensions, and improve model training efficiency;

[0071] The specific steps of feature extraction and selection are as follows:

[0072] Step 21: Feature extraction:

[0073] a. Laser parameter features: directly extract the laser rust removal parameters set in data acquisition as features; including laser power (unit: watt, W), scanning speed (unit: millimeter per second, mm / s), pulse frequency (unit: kilohertz, kHz), spot diameter (unit: millimeter, mm). These parameters directly affect the interaction process between the laser and the metal surface, and are key factors in determining the rust removal effect and substrate damage. For example, higher laser power may provide stronger energy to remove rust, but may also increase the risk of substrate damage; faster scanning speeds may reduce laser energy deposition per unit area, affecting the thoroughness of rust removal, etc.

[0074] b. Metal surface characteristics:

[0075] Roughness characteristics: Use a roughness meter to measure the metal surface before rust removal to obtain surface roughness parameters, such as arithmetic mean roughness (Ra), root mean square roughness (Rq), etc. The roughness of the metal surface will affect the reflection and absorption characteristics of the laser, and thus affect the rust removal effect. A rougher surface may cause the laser to scatter and reduce the efficiency of rust removal, but in some cases it may also help distribute the laser energy on the surface and reduce matrix damage caused by local overheating; Texture characteristics: Extract texture features from the image data of the metal surface before rust removal. Specifically, the gray level co-occurrence matrix (GLCM) method is used to calculate texture parameters such as contrast, correlation, energy, and entropy. For example, metal surfaces with different rust textures (such as uniform rust, local concentrated rust, mottled rust, etc.) may respond differently to the laser. These texture features help describe the relationship between the initial state of the metal surface and the rust removal effect;

[0076] c. Surface quality characteristics after rust removal:

[0077] Residual rust amount characteristics: The residual rust amount on the metal surface after rust removal is determined by chemical analysis. Chemical analysis is an improvement on the mass loss measurement method. The residual rust amount is determined by weighing the sample again after rust removal and calculating the difference with the ideal mass after rust removal. Surface flatness characteristics: The flatness change of the metal surface after rust removal is measured using a profilometer or optical interferometer.

[0078] Step 22, feature selection: principal component analysis, first, calculate the covariance matrix of the feature data in step 21; then, solve the eigenvalues ​​and eigenvectors of the covariance matrix, the eigenvalues ​​represent the variance of the data in the direction of the corresponding eigenvectors; finally, sort the eigenvectors according to the eigenvalues, select the first k eigenvectors, and the eigenvectors constitute the principal component transformation matrix, that is, the k eigenvectors are the key features after screening, where the value of k is usually determined according to the cumulative contribution rate, and the k principal components corresponding to the cumulative contribution rate reaching a certain proportion (such as 85%-95%) are considered to be able to retain most of the information of the original data. For example, if there are 10 features originally, after PCA analysis, it is found that the cumulative contribution rate of the first 5 principal components reaches 90%, then the features corresponding to these 5 principal components are selected as key features, which realizes the dimensionality reduction of the data, while retaining the main information and reducing the interference of redundant features on the subsequent machine learning model training.

[0079] Step 3, model training: select a neural network model and use the features screened in step 2 as input, and use the rust removal effect evaluation index as output to train the model; use the cross-validation method to adjust the hyperparameters of the model and optimize the performance of the model so that the model can accurately predict the rust removal effect and substrate damage under different laser parameters;

[0080] The specific implementation steps of model training are as follows:

[0081] Step 31, neural network model structure selection and initialization: select a multi-layer feedforward neural network, including an input layer, a hidden layer and an output layer; the number of input layer nodes is equal to the number of features after screening, specifically, after feature selection, the four key features of laser power, scanning speed, metal surface roughness, and residual rust amount are determined, then the number of input layer nodes is 4, and the number of output layer nodes is 2, corresponding to the rust removal rate and substrate damage degree in the rust removal effect evaluation index respectively; the number of hidden layers and the number of nodes need to be determined based on experience and experiments;

[0082] Step 32, forward propagation: Let the input feature vector be X = [x 1 , x 2 , x 3 , x 4 ], where x 1 is the laser power, x 2 is the scanning speed, x 3 is the metal surface roughness, x 4 is the residual corrosion amount. For the calculation from the input layer to the hidden layer, the input net of the jth node in the hidden layer is j The calculation formula is:

[0083]

[0084] where ωij is the weight from the i-th node in the input layer to the j-th node in the hidden layer, b j is the bias of the jth node in the hidden layer;

[0085] The output h of the hidden layer node j The activation function is used for calculation. Here we choose the commonly used sigmoid activation function: For the calculation from the hidden layer to the output layer, the input net′ of the kth node in the output layer k The calculation formula is:

[0086]

[0087] where ω′ jk is the weight from the jth node in the hidden layer to the kth node in the output layer, b′ k is the bias of the nth node in the output layer;

[0088] The output y of the output layer node k The activation function is also used. For the rust removal rate, the sigmoid function is used to map the output to the [0, 1] interval, indicating the proportion of rust removal; for the matrix damage degree, the linear activation function is used, and the output is a numerical value, indicating the quantified value of the damage degree;

[0089] Step 33, loss function calculation: The mean square error is used as the loss function to measure the difference between the model prediction value and the true value. is the true value of the kth output, i.e., the rust removal rate and matrix damage degree actually measured, y k is the model prediction value, and the loss function L is calculated as:

[0090] Step 34, back propagation and weight update: The gradient of the loss function to the weights and biases of each layer is calculated through the back propagation algorithm, and then the weights and biases are updated according to the gradient descent algorithm, where the weight ω′ from the hidden layer to the output layer jk The update formula is:

[0091]

[0092] Where η is the learning rate, which is a hyperparameter that controls the step size of weight updates. is the loss function for ω′ jk The gradient of is calculated step by step from the output layer to the hidden layer through the chain rule;

[0093] Step 35, cross-validation to adjust hyperparameters: divide the collected laser rust removal experimental data into k subsets, select k-1 subsets as training sets each time, and the remaining 1 subset as validation set; select the hyperparameter combination with the smallest loss function value on the validation set as the final model hyperparameter, and then use the entire training data set to retrain the model under the hyperparameter combination to obtain the final trained neural network model;

[0094] Through the above steps, the neural network model is used to learn and model the data in the laser rust removal method based on machine learning, so as to achieve accurate prediction and analysis of the laser rust removal effect and substrate damage, and provide reliable model support for subsequent online monitoring and parameter optimization.

[0095] Step 4: Online monitoring and parameter optimization: During the actual laser rust removal process, sensors are used to collect relevant data on the metal surface in real time, and the current rust removal effect and substrate damage are monitored and predicted online in combination with the neural network model pre-trained in step 3; according to the prediction results, the laser rust removal parameters are dynamically adjusted through the optimization algorithm to achieve the best rust removal effect and the minimum substrate damage;

[0096] The specific implementation contents of the online monitoring and parameter optimization are as follows:

[0097] A. Sensor data collection and transmission

[0098] Laser parameter sensor: Install a high-precision laser power sensor on the laser rust removal equipment to monitor the laser output power in real time; at the same time, install a scanning speed sensor to accurately measure the moving speed of the laser scanning head through encoders and other technical means;

[0099] Surface state sensor: Use infrared temperature sensor to measure the real-time temperature change of metal surface under the action of laser. At the same time, use optical reflectivity sensor to detect the intensity of reflected light after laser irradiates the metal surface. The intensity of reflected light will be different under different rust degree and metal matrix state. By monitoring the reflectivity, it can assist in judging the progress of rust removal.

[0100] B. Data preprocessing and feature extraction

[0101] Perform real-time preprocessing of collected sensor data, including noise removal;

[0102] Extract real-time features corresponding to the input features of the training model, directly obtain the current laser power value from the laser power sensor data, obtain the scanning speed value from the scanning speed sensor, extract the real-time temperature features of the metal surface from the infrared temperature sensor data, extract the reflectivity features from the optical reflectivity sensor data, etc. The above features will be used as input for subsequent model predictions;

[0103] C. Prediction based on machine learning model: The real-time feature data after preprocessing and extraction is input into the pre-trained neural network model in step 3. The model predicts the current rust removal rate and substrate damage degree based on the input feature data and the relationship between the features learned in the training phase and the rust removal effect and substrate damage;

[0104] D. Optimization algorithm and parameter adjustment

[0105] Optimization target setting: The optimization target is to improve the rust removal rate while ensuring that the degree of substrate damage is lower than the set threshold;

[0106] Selection and application of optimization algorithm: Genetic algorithm is used for parameter optimization. First, the laser power, scanning speed, metal surface roughness, and residual rust parameters are encoded into chromosome form. Then, a population is initialized, which contains multiple parameter combination individuals encoded in this way. The predicted rust removal rate and substrate damage degree corresponding to each individual are calculated. The fitness function is set according to the optimization goal. Through selection, crossover, and mutation genetic operations, new parameter combination individuals are continuously iterated to find the parameter combination with the highest fitness.

[0107] E. Parameter adjustment implementation: The optimized laser rust removal parameters are sent to the controller of the laser rust removal equipment. The controller adjusts the power output of the laser generator and the speed of the scanning motor according to the received parameter instructions to achieve dynamic adjustment of the laser rust removal parameters. After adjusting the parameters, continue to collect data through sensors and repeat the above monitoring, prediction and optimization process until a satisfactory rust removal effect is achieved and the substrate damage is within a controllable range.

[0108] Through the above online monitoring and parameter optimization steps, the laser rust removal parameters can be intelligently adjusted in real time during the laser rust removal process based on the actual situation of the metal surface and the pre-trained model prediction results, thereby improving the quality and efficiency of rust removal, reducing substrate damage, and realizing automated and precise laser rust removal operations.

[0109] Step 5, effect evaluation and model update: After completing the laser rust removal operation, the rust removal effect is comprehensively evaluated. Specifically, the microscopic morphology and residual rust amount of the metal surface after rust removal are detected by optical microscope and scanning electron microscope equipment; Specifically, the image analysis method is used: the image of the metal surface after rust removal is collected using a high-resolution microscope or industrial camera. The shooting area should cover the entire rust removal surface or select multiple areas according to certain sampling rules. The collected images are analyzed using image processing software, and the metal matrix and the possible residual rust area are separated by an image segmentation algorithm. Then, the area ratio or pixel number ratio of the residual rust area to the total detection area is calculated. For example, if the total number of pixels in the detection area is 1 million, and the number of pixels in the residual rust area is 50,000 after image analysis, the residual rust area ratio is 5%. This method has the advantages of non-destructive and rapid detection, but the detection accuracy may be limited for some microscopic residual rust that is similar to the color of the matrix; Microscopic morphology observation: The microscopic morphology of the metal surface after rust removal is observed using a scanning electron microscope (SEM). SEM can provide highly magnified surface images, clearly showing whether there are matrix damage features such as microcracks, fused layers, and heat-affected zones on the metal surface. The degree of matrix damage is quantified by measuring the length, width, and depth of microcracks and counting their number and distribution density. For example, if three microcracks with a length of 10-20μm are found in an observation area of ​​10mm×10mm, the relevant data can be recorded for subsequent evaluation; Hardness test: Use a microhardness tester to perform hardness tests in the rust removal area and the untreated matrix area respectively. Select multiple test points in the rust removal area (such as selecting 9 points according to a 5mm×5mm grid distribution), measure their hardness values ​​and calculate the average value. At the same time, select the same number and distribution of test points in the untreated area for comparison. Calculate the difference or hardness change rate between the hardness value of the rust removal area and the hardness value of the untreated area to evaluate the hardness change of the matrix caused by the laser rust removal process, and then infer the degree of matrix damage;

[0110] Compare and analyze the actual evaluation results with the model prediction results. If there is a large deviation between the two, add the rust removal data to the training data set, update and optimize the machine learning model to continuously improve the accuracy and adaptability of the model;

[0111] The specific contents of the comparison and analysis between the actual evaluation results and the model prediction results are as follows: the rust removal effect and substrate damage degree data obtained by evaluation are compared with the corresponding results predicted by the machine learning model in the laser rust removal process, and the deviation value is specifically calculated. For the rust removal rate deviation: ΔRR = |RR 实际 -RR 预测 |, for the matrix damage degree deviation ΔBD=|BD 实际 -BD 预测|, set a reasonable deviation threshold, compare the calculated deviation value with the threshold value for judgment, and judge that the deviation is large, and the machine learning model needs to be updated; if the deviation is within the threshold range, it is considered that the current model prediction effect is good, and the model can be temporarily not updated; when the deviation is large, the relevant data of this laser rust removal operation (including laser rust removal parameters, initial metal surface features, real-time sensor data, and rust removal effect and matrix damage data obtained from this actual evaluation) are added to the training data set. Perform the same preprocessing operations as the original data, such as data cleaning and normalization, on the newly added data to ensure data consistency and availability; retrain the machine learning model (such as neural network) using the expanded training data set. Reset training parameters, such as the number of training rounds, learning rate, etc., and adjust them appropriately based on experience or experiments. For example, for the neural network model, increase the number of training rounds to 1.5 times the original number, and appropriately reduce the learning rate to improve the convergence accuracy of the model. During the training process, the same loss function (such as mean square error function) and optimization algorithm (such as stochastic gradient descent algorithm) as the initial training are used to update and optimize the model's weights and biases, so that the model can better learn the characteristics and rules in the new data, thereby improving the accuracy of the model's prediction of laser rust removal effects and substrate damage; after the retraining is completed, an independent test data set is used to evaluate the performance of the updated model. The prediction accuracy, mean square error and other evaluation indicators of the model on the test set are calculated and compared with the model performance indicators before the update to verify whether the model update is effective. If the performance of the updated model is significantly improved, it can be applied to the subsequent online monitoring and parameter optimization process of laser rust removal; if the performance improvement is not obvious or decreases, it is necessary to further analyze the reasons, such as checking the data quality, adjusting the model structure or training parameters, and re-updating and optimizing the model until satisfactory model performance is obtained.

[0112] The present invention can accurately establish the complex relationship between the rust removal effect and many parameters such as laser power, scanning speed, pulse frequency, spot diameter, and metal surface characteristics through the learning of a large amount of experimental data and actual case data by machine learning models. In the actual rust removal process, based on the data collected by the sensor in real time, the model can accurately predict the current rust removal effect and dynamically adjust the laser rust removal parameters using an optimization algorithm. This makes the rust removal process no longer rely on empirical parameter settings, but intelligently optimizes according to the real-time state of the metal surface, thereby significantly improving the rust removal rate and ensuring that the metal surface can reach a higher cleanliness standard. For example, when processing metal parts with complex rust conditions and mixed materials, the model can quickly adapt and find the best parameter combination to effectively remove stubborn rust, while traditional methods often find it difficult to achieve such precise parameter matching, resulting in incomplete rust removal or excessive treatment. Adaptable to various working conditions: Whether facing metal surfaces of different materials (such as carbon steel, stainless steel, aluminum alloy, etc.), or dealing with various degrees of rust (from mild surface oxidation to severe rust layer accumulation), this method can flexibly adjust parameter strategies through the adaptive learning ability of the machine learning model to achieve the best rust removal effect. This greatly expands the application scope of laser rust removal technology, enabling it to be more widely and efficiently used in many industrial fields (such as machinery manufacturing, automobile repair, ship maintenance, etc.) and special industries such as cultural relics restoration and architectural decoration, solving the problem of poor adaptability of traditional rust removal methods when dealing with different working conditions;

[0113] While the present invention accurately controls the rust removal parameters, the machine learning model can also predict and monitor the damage to the substrate in real time. Through comprehensive analysis of metal surface features, laser parameters and real-time monitoring data, the model can warn in advance of parameter settings that may cause excessive damage to the substrate, and make timely adjustments to minimize the degree of substrate damage. For example, when it is detected that the metal surface temperature is too high or the laser energy density is too high, which may cause substrate melting or microcracks, the model will automatically adjust the scanning speed or reduce the laser power to avoid damage to the structural integrity and performance of the substrate, while the traditional laser rust removal method with fixed parameter settings is difficult to achieve such precise damage prevention and control; Extend the life of metal parts: Since the damage to the substrate can be effectively reduced, the metal parts after rust removal by the method of the present invention can maintain better mechanical properties and surface quality, thereby significantly extending their service life. This advantage is particularly prominent in some fields that require high reliability and durability of metal parts, such as aerospace, high-end equipment manufacturing, etc. Compared with the potential substrate damage risks that may be brought about by traditional rust removal methods, the present invention provides a strong guarantee for the long-term stable operation of metal parts, reduces the risk of component failure caused by the rust removal process, and reduces the cost of maintenance and replacement;

[0114] In the whole laser rust removal process, the data collected by the sensor in real time is closely combined with the machine learning model to realize online monitoring and instant feedback of the rust removal effect and substrate damage. This real-time performance enables the operator to timely understand the progress of the rust removal operation. Once an abnormality occurs (such as the rust removal effect does not meet expectations or the substrate damage has an increasing trend), the system can respond quickly and adjust the parameters, avoiding the time delay and operational inconvenience caused by the need for manual regular inspection and manual adjustment of parameters in the traditional rust removal method. This not only improves the overall efficiency of the rust removal operation, but also reduces the errors and uncertainties caused by human factors, ensuring the stability and consistency of the rust removal quality; the intelligent control based on machine learning of the present invention enables the laser rust removal process to be highly automated. From data collection, model prediction, parameter optimization to the final rust removal effect evaluation and model update, the entire process does not require excessive manual intervention. It only needs to set the relevant parameters and start the equipment in the initial stage, and the system can automatically run and perform adaptive adjustments according to the actual situation. This greatly reduces the labor intensity and technical requirements of the operator, while improving the production efficiency and the standardization of the rust removal operation. In large-scale industrial production or batch rust removal tasks, the advantages of automation are more obvious, which can significantly shorten the production cycle and improve the economic benefits and market competitiveness of the enterprise.

[0115] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A laser rust removal method based on machine learning, characterized in that: The following steps are involved: Step 1, data collection and preprocessing: collect a large number of rust removal experimental data of metal surfaces of different materials under different laser rust removal parameters, including laser power, scanning speed, pulse frequency, and spot diameter. The rust removal experimental data includes image data of the metal surface before and after rust removal, rust degree data, and matrix damage data; clean and normalize the collected data to improve the quality and availability of the data; Step 2, feature extraction and selection: extract features related to laser rust removal effect and substrate damage from the preprocessed data, including laser parameter features, metal surface features, and surface quality features after rust removal; The feature selection algorithm is used to screen out the key features that have a significant impact on the laser rust removal effect and substrate damage, reduce data dimensions, and improve model training efficiency; Step 3, model training: select a neural network model and use the features screened in step 2 as input, and use the rust removal effect evaluation index as output to train the model; use the cross-validation method to adjust the hyperparameters of the model and optimize the performance of the model so that the model can accurately predict the rust removal effect and substrate damage under different laser parameters; Step 4: Online monitoring and parameter optimization: During the actual laser rust removal process, sensors are used to collect relevant data on the metal surface in real time, and the current rust removal effect and substrate damage are monitored and predicted online in combination with the neural network model pre-trained in step 3; according to the prediction results, the laser rust removal parameters are dynamically adjusted through the optimization algorithm to achieve the best rust removal effect and the minimum substrate damage; Step 5. Effect evaluation and model update: After completing the laser rust removal operation, conduct a comprehensive evaluation of the rust removal effect. Specifically, use optical microscopes and scanning electron microscopes to detect the microscopic morphology and residual rust amount of the metal surface after rust removal. Compare and analyze the actual evaluation results with the model prediction results. If there is a large deviation between the two, add the rust removal data to the training data set, and update and optimize the machine learning model to continuously improve the accuracy and adaptability of the model.

2. The laser rust removal method based on machine learning according to claim 1, characterized in that: The specific implementation steps of data collection and preprocessing in step 1 are as follows: Step 11, Experimental design and sample preparation: Select experimental samples for a variety of common metal materials; for each experimental sample, prepare samples with different degrees of corrosion. The degree of corrosion is controlled and adjusted by natural exposure time and artificial accelerated corrosion methods to cover various conditions from mild to severe corrosion; Step 12, image data acquisition: using a high-resolution industrial camera or optical microscope to take images of the metal surface before and after rust removal to obtain image data; Step 13, Corrosion degree data collection: chemical analysis is used to quantify the degree of corrosion, specifically, the mass loss per unit area of ​​the rusted sample before rust removal is measured to determine the amount of corrosion, a high-precision electronic balance is used to weigh the mass change of the sample before and after rust removal, and the mass thickness of the corrosion is calculated in combination with the sample surface area; Step 14, matrix damage data collection: use a scanning electron microscope to observe the microscopic morphology of the metal matrix surface after rust removal, and check whether there are microcracks, fused layer, and heat-affected zone matrix damage characteristics; for existing damage, quantify it by measuring the depth and width of the damaged area and calculating the area ratio of the damaged area to the entire rust removal area; at the same time, use a hardness test method to compare the change in the hardness of the metal matrix before and after rust removal, and indirectly evaluate the degree of matrix damage. The hardness test uses a microhardness tester to perform multi-point measurements in the rust removal area and the untreated area and take the average value; Step 15, outlier processing: For image data, check whether there are blurred, missing or obviously wrong pixel areas due to reasons such as camera failure, abnormal lighting, etc.; for corrosion degree data and matrix damage data, if the measured value obviously deviates from the normal range, it is necessary to further check the original measurement record or repeat the measurement to determine the accuracy of the data; Step 16, normalizing image data: for the image data collected in the above steps, normalizing the pixel values ​​of the image, specifically mapping the pixel values ​​of the image to the interval [0,1]; Step 17, normalization of numerical data: the corrosion degree data, substrate damage data and laser rust removal parameter data are normalized using the Min-Max normalization or Z-score normalization method.

3. The laser rust removal method based on machine learning according to claim 1, characterized in that: The specific steps of feature extraction and selection in step 2 are as follows: Step 21: Feature extraction: a. Laser parameter features: directly extract the laser rust removal parameters set in data acquisition as features; b. Metal surface characteristics: Roughness characteristics: Use a roughness measuring instrument to measure the metal surface before rust removal to obtain surface roughness parameters; Texture features: Extract texture features from the image data of the metal surface before rust removal; c. Surface quality characteristics after rust removal: Residual rust amount characteristics: The residual rust amount on the metal surface after rust removal is determined by chemical analysis. Chemical analysis is an improvement on the mass loss measurement method. The residual rust amount is determined by weighing the sample again after rust removal and calculating the difference with the ideal mass after rust removal. Surface flatness characteristics: The flatness change of the metal surface after rust removal is measured using a profilometer or optical interferometer. Step 22, feature selection: principal component analysis, first, calculate the covariance matrix of the feature data in step 21; then, solve the eigenvalues ​​and eigenvectors of the covariance matrix, the eigenvalues ​​represent the variance of the data in the direction of the corresponding eigenvectors; finally, sort the eigenvectors according to the eigenvalues, select the first k eigenvectors, and the eigenvectors constitute the principal component transformation matrix, that is, the k eigenvectors are the key features after screening.

4. The laser rust removal method based on machine learning according to claim 1, characterized in that: The specific implementation steps of the model training in step 3 are as follows: Step 31, neural network model structure selection and initialization: select a multi-layer feedforward neural network, including an input layer, a hidden layer and an output layer; the number of input layer nodes is equal to the number of features after screening, specifically, after feature selection, the four key features of laser power, scanning speed, metal surface roughness, and residual rust amount are determined, then the number of input layer nodes is 4, and the number of output layer nodes is 2, corresponding to the rust removal rate and substrate damage degree in the rust removal effect evaluation index respectively; the number of hidden layers and the number of nodes need to be determined based on experience and experiments; Step 32, forward propagation: Let the input feature vector be X = [x1, x2, x3, x4], where x1 is the laser power, x2 is the scanning speed, x3 is the metal surface roughness, and x4 is the residual rust. For the calculation from the input layer to the hidden layer, the input net of the jth node in the hidden layer is j The calculation formula is: where ω ij is the weight from the i-th node in the input layer to the j-th node in the hidden layer, b j is the bias of the jth node in the hidden layer; The output h of the hidden layer node j The activation function is used for calculation. Here we choose the commonly used sigmoid activation function: For the calculation from the hidden layer to the output layer, the input net′ of the kth node in the output layer k The calculation formula is: where ω′ jk is the weight from the jth node in the hidden layer to the kth node in the output layer, b′ k is the bias of the nth node in the output layer; The output y of the output layer node k The activation function is also used. For the rust removal rate, the sigmoid function is used to map the output to the [0,1] interval, indicating the proportion of rust removal; for the matrix damage degree, the linear activation function is used, and the output is a numerical value, indicating the quantitative value of the damage degree; Step 33, loss function calculation: The mean square error is used as the loss function to measure the difference between the model prediction value and the true value. is the true value of the kth output, i.e., the rust removal rate and substrate damage degree actually measured, y k is the model prediction value, and the loss function L is calculated as: Step 34, back propagation and weight update: The gradient of the loss function to the weights and biases of each layer is calculated through the back propagation algorithm, and then the weights and biases are updated according to the gradient descent algorithm, where the weight ω from the hidden layer to the output layer is j ′ k The update formula is: Where η is the learning rate, which is a hyperparameter that controls the step size of weight updates. is the loss function for ω j ′ k The gradient of is calculated step by step from the output layer to the hidden layer through the chain rule; Step 35, cross-validation to adjust hyperparameters: divide the collected laser rust removal experimental data into k subsets, select k-1 subsets as training sets each time, and the remaining 1 subset as the validation set; select the hyperparameter combination with the smallest loss function value on the validation set as the final model hyperparameter, and then use the entire training data set to retrain the model under this hyperparameter combination to obtain the final trained neural network model.

5. The laser rust removal method based on machine learning according to claim 1, characterized in that: The specific implementation contents of the online monitoring and parameter optimization of step 4 are as follows: A. Sensor data collection and transmission Laser parameter sensor: Install a high-precision laser power sensor on the laser rust removal equipment to monitor the laser output power in real time; at the same time, install a scanning speed sensor to accurately measure the moving speed of the laser scanning head through encoders and other technical means; Surface state sensor: Use infrared temperature sensor to measure the real-time temperature change of metal surface under the action of laser. At the same time, use optical reflectivity sensor to detect the intensity of reflected light after laser irradiates the metal surface. The intensity of reflected light will be different under different rust degree and metal matrix state. By monitoring the reflectivity, it can assist in judging the progress of rust removal. B. Data preprocessing and feature extraction Perform real-time preprocessing of collected sensor data, including noise removal; Extract real-time features corresponding to the input features of the training model, directly obtain the current laser power value from the laser power sensor data, obtain the scanning speed value from the scanning speed sensor, extract the real-time temperature features of the metal surface from the infrared temperature sensor data, extract the reflectivity features from the optical reflectivity sensor data, etc. The above features will be used as input for subsequent model predictions; C. Prediction based on machine learning model: The real-time feature data after preprocessing and extraction is input into the pre-trained neural network model in step 3. The model predicts the current rust removal rate and substrate damage degree based on the input feature data and the relationship between the features learned in the training phase and the rust removal effect and substrate damage; D. Optimization algorithm and parameter adjustment Optimization target setting: The optimization target is to improve the rust removal rate while ensuring that the degree of substrate damage is lower than the set threshold; Selection and application of optimization algorithm: Genetic algorithm is used for parameter optimization. First, the laser power, scanning speed, metal surface roughness, and residual rust parameters are encoded into chromosome form. Then, a population is initialized, which contains multiple parameter combination individuals encoded in this way. The predicted rust removal rate and substrate damage degree corresponding to each individual are calculated. The fitness function is set according to the optimization goal. Through selection, crossover, and mutation genetic operations, new parameter combination individuals are continuously iterated to find the parameter combination with the highest fitness. E. Parameter adjustment implementation: The optimized laser rust removal parameters are sent to the controller of the laser rust removal equipment. The controller adjusts the power output of the laser generator and the speed of the scanning motor according to the received parameter instructions to achieve dynamic adjustment of the laser rust removal parameters. After adjusting the parameters, continue to collect data through sensors and repeat the above monitoring, prediction and optimization process until a satisfactory rust removal effect is achieved and the substrate damage is within a controllable range.

6. The laser rust removal method based on machine learning according to claim 1, characterized in that: In step 5, the actual evaluation results are compared with the model prediction results. The specific contents are as follows: the rust removal effect and substrate damage degree data obtained by evaluation are compared with the corresponding results predicted by the machine learning model in the laser rust removal process, and the deviation value is specifically calculated. For the rust removal rate deviation: ΔRR = |RR 实际 -RR 预测 |, for the matrix damage degree deviation ΔBD=|BD 实际 -BD 预测 |, set a reasonable deviation threshold, and compare the calculated deviation value with the threshold. If the deviation is large, the machine learning model needs to be updated; if the deviation is within the threshold range, it is considered that the current model prediction effect is good and the model can be temporarily not updated.

Images