PSO-BPNN-based laser paint removal process parameter optimization method and medium

Abstract

The invention relates to a PSO-BPNN-based laser paint removal process parameter optimization method and a medium, and the method comprises the steps: collecting the physical property parameters of a substrate and a paint layer of a to-be-processed workpiece, and constructing a multi-physical field simulation model of a paint removal process; under different set laser paint removal process parameter combinations, multi-physical field numerical simulation of interaction of laser, a paint layer and a base material is carried out, and corresponding surface topography parameter simulation results are obtained; according to the method, a data set is constructed, an optimal laser paint removal process parameter combination is predicted by adopting a PSO-BPNN prediction optimization algorithm, in the PSO-BPNN prediction optimization algorithm, a BPNN algorithm is utilized to construct a material-process-performance mapping relation, particle swarm optimization is performed by utilizing the PSO algorithm, and the optimal laser paint removal process parameter combination is obtained. Each particle in the particle swarm represents a laser paint removal process parameter combination. Compared with the prior art, the laser paint removing device has the advantages of improving the laser paint removing precision and the like.

Description

Technical Field

[0001] This invention relates to the fields of laser cleaning technology and artificial intelligence optimization control technology, and in particular to a method and medium for optimizing laser paint removal process parameters based on PSO-BPNN. Background Technology

[0002] Laser paint removal, as a highly efficient, clean, and non-contact surface treatment technology, has garnered widespread attention in fields such as aerospace, rail transportation, precision manufacturing, and cultural relic restoration. Compared to traditional methods like chemical peeling, sandblasting, and mechanical grinding, laser paint removal offers advantages such as high processing precision, minimal environmental pollution, and strong adaptability, making it particularly suitable for applications requiring high surface integrity and processing control. However, the laser paint removal process involves a complex coupling of multiple factors, including laser beam parameters, material thermophysical properties, and coating structure. Different combinations of process parameters can lead to drastically different paint removal results, including incomplete removal, thermal damage, ablation, or substrate deformation. Therefore, achieving precise control and parameter optimization of the laser paint removal process remains a key challenge for the technology's intelligent and engineering applications.

[0003] Currently, the selection of laser paint removal process parameters largely relies on the operator's experience or a large amount of experimental data, determining a reasonable parameter combination through repeated trial and error. This approach is not only inefficient and costly, but also difficult to adapt to the personalized paint removal needs of different materials and structural types, and cannot fully reveal the nonlinear mapping relationship between laser process parameters and paint removal quality. While traditional theoretical analysis methods can reveal the basic laws of laser-material interaction to some extent, they struggle to account for the combined effects of complex boundary conditions, multi-scale structures, and multiple physical processes in actual engineering. Therefore, it is necessary to introduce artificial intelligence methods with strong nonlinear fitting capabilities and global optimization capabilities to model and optimize the laser paint removal process. Summary of the Invention

[0004] The purpose of this invention is to provide a method and medium for optimizing laser paint removal process parameters based on PSO-BPNN to improve the accuracy of laser paint removal.

[0005] The objective of this invention can be achieved through the following technical solutions:

[0006] A method for optimizing laser paint removal process parameters based on PSO-BPNN includes the following steps:

[0007] Collect the physical properties of the substrate and paint layer of the workpiece to be processed, and construct a multiphysics simulation model of the paint removal process;

[0008] Based on the multiphysics simulation model of the paint removal process, multiphysics numerical simulation of the interaction between the laser and the paint layer and the substrate is carried out under different combinations of laser paint removal process parameters to obtain the simulation results of the corresponding surface morphology parameters.

[0009] Based on simulation results of different material properties, laser paint removal process parameter combinations, and surface morphology parameters, a dataset is constructed. The optimal combination of laser paint removal process parameters is predicted using the PSO-BPNN prediction optimization algorithm. In the PSO-BPNN prediction optimization algorithm, the BPNN algorithm is used to construct the mapping relationship between material, process, and performance, and the PSO algorithm is used for particle swarm optimization to obtain the optimal combination of laser paint removal process parameters. Each particle in the particle swarm represents a combination of laser paint removal process parameters.

[0010] Furthermore, the step of constructing a multiphysics simulation model of the paint removal process includes:

[0011] Based on the physical properties of the substrate and paint layer of the workpiece to be processed, an initial multiphysics simulation model of the paint removal process is established in the simulation software.

[0012] The initial multiphysics simulation model is meshed, the laser action area is locally densified, and boundary conditions are set during the simulation process to obtain the final multiphysics simulation model of the paint removal process.

[0013] Furthermore, the boundary conditions include the initial temperature field, laser heat source loading conditions, thermal convection and thermal radiation boundaries, and thermal conduction interface conditions.

[0014] Furthermore, the step of obtaining the corresponding surface morphology parameter simulation results includes:

[0015] Based on the existing laser design, multiple sets of laser paint removal process parameter combinations are constructed as simulation input variables;

[0016] The laser paint removal process parameters are input one by one into the multiphysics simulation model of the paint removal process for simulation. The simulation results are obtained to obtain the three-dimensional morphological changes and paint peeling characteristics of the surface of the workpiece after laser treatment, and the simulation results of surface morphology parameters closely related to paint removal are extracted from them.

[0017] Furthermore, it also includes preprocessing and classifying the database, clustering the data using cluster analysis, and using principal component analysis to reduce dimensionality and quantify the sensitivity of key parameters.

[0018] Furthermore, the laser paint removal process parameters in the laser paint removal process parameter combination include power P, frequency f, scanning speed v1, scanning spacing s1, and number of scans n.

[0019] Furthermore, the step of predicting the optimal combination of laser paint removal process parameters includes:

[0020] 1) Initialize the particle swarm and define the multi-objective optimization function, and construct the mapping relationship between material-process-performance based on the BPNN algorithm to calculate the objective function value of each individual particle;

[0021] 2) Calculate the fitness value of individual particles and redetermine the extreme values ​​of individual particles and the global extreme value;

[0022] 3) Using the objective function value as the basis for updating the particle's individual direction, update the particle's velocity vector and position vector, where the update expressions are as follows:

[0023]

[0024] In the formula, Let the updated velocity of the i-th particle in the d-th dimension be the velocity at iteration k+1. Let be the velocity of the i-th particle in the d-th dimension at the current k iterations; w be the inertia weight, used to balance the capabilities of global and local search; r1 and r2 be random numbers used to introduce randomness; and l1 and l2 be individual and social learning factors. The optimal position of particle i in the d-th dimension at the current k iterations is used as the historical best position. Let i be the current position of particle i in the d-th dimension. This represents the globally optimal position of the entire group in the d-th dimension. Let i be the updated position of particle i in the d-th dimension, which will be used as the position in the (k+1)th iteration.

[0025] 4) Calculate the fitness values ​​of the new generation of individual particles and redetermine the individual particle extreme values ​​and the global extreme values;

[0026] 5) Determine if the iteration conditions are met. If yes, stop the iteration, output the optimal individual particle, and obtain the optimal combination of laser paint removal process parameters. If not, return to step 3) to perform iterative calculations until the iteration conditions are met.

[0027] Furthermore, the expression for the multi-objective optimization function is:

[0028] f(x) = max(0, A / A) t -1)+max(0,Ra-Ra max )+max(0,1-R r )

[0029] In the formula, f(x) is a multi-objective optimization function, and A / A t Ra is the surface area ratio, and Ra is the roughness.max For maximum roughness, R r For paint removal rate, A t A and B represent the surface areas before and after laser treatment, respectively.

[0030] Furthermore, the training steps using the BPNN algorithm include:

[0031] (1) Initialization: Randomly initialize weights and biases;

[0032] (2) Forward propagation: The laser paint removal process parameters are input through the input layer, and the current layer output is obtained through the transfer function, which is then used as the input for the next layer, until the output layer obtains the predicted surface morphology parameters. The expression for the transfer function is:

[0033]

[0034] In the formula, F sigmoid (z) is the Sigmoid activation function, used to map the input z to the (0,1) interval, where z is the weighted input of the neuron, i.e., the combination of laser paint removal process parameters, H. j Let n be the output of the j-th hidden layer neuron, n be the number of input layer neurons, p be the number of hidden layer neurons, and w be the output of the j-th hidden layer neuron. ij Let x be the connection weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer. i Let a be the input value of the i-th neuron in the input layer, m be the number of neurons in the output layer, and a be the input value of the i-th neuron in the input layer. j y is the bias of the j-th neuron in the hidden layer. k w is the output of the k-th neuron in the output layer. jk Let b be the connection weight from the j-th neuron in the hidden layer to the k-th neuron in the output layer. k This is the bias of the k-th neuron in the output layer;

[0035] (3) Calculation error: The error between the predicted surface topography parameters and the corresponding simulated surface topography parameters is calculated using a loss function, where the expression for the loss function is:

[0036]

[0037] In the formula, L is the loss function. This represents the true value of the k-th output neuron, i.e., the simulation result of the surface morphology parameters;

[0038] (4) Backpropagation: Calculate the output layer error using the loss function, and then backpropagate the hidden layer error layer by layer, and then calculate the gradient of each layer.

[0039] (5) Update weights and biases: Update the weights and biases based on the backpropagation results. The hidden layer bias update expression is as follows:

[0040]

[0041] In the formula, a j (T+1) represents the bias of the j-th neuron in the hidden layer at the (T+1)-th iteration, i.e., the updated hidden layer bias, a j (T) represents the bias of the j-th neuron in the hidden layer at the T-th iteration, and μ is the learning rate, controlling the magnitude of parameter updates. For the loss function L with respect to bias a j The gradient;

[0042] The expression for updating the weights from the input layer to the hidden layer is:

[0043]

[0044] In the formula, w ij (T+1) represents the weights from the input layer to the hidden layer at the (T+1)th iteration, i.e., the updated weights from the input layer to the hidden layer, w. ij (T) represents the weights from the input layer to the hidden layer at the T-th iteration. For the loss function L with respect to the weights w ij The gradient;

[0045] The output layer bias update expression is:

[0046]

[0047] In the formula, b k (T+1) represents the bias of the k-th neuron in the output layer at the (T+1)-th iteration, i.e., the updated output layer bias, b k (T) represents the bias of the k-th neuron in the output layer during the T-th iteration. For the loss function L with respect to bias b k The gradient;

[0048] The expression for updating the weights from the hidden layer to the output layer is:

[0049]

[0050] In the formula, w jk (T+1) represents the weights from the hidden layer to the output layer at the (T+1)th iteration, i.e., the updated weights from the hidden layer to the output layer, w. jk (T) represents the weights from the hidden layer to the output layer at the T-th iteration. For the loss function L with respect to the weights w jk The gradient;

[0051] Iteration: Return to step (2) and repeat the iteration until the iteration condition is met to obtain the trained BPNN algorithm.

[0052] The present invention also provides a computer-readable storage medium including one or more programs executable by one or more processors of an electronic device, the one or more programs including instructions for performing the laser paint removal process parameter optimization method based on PSO-BPNN as described above.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] (1) This invention achieves efficient correlation modeling between laser process parameters and paint removal effect by combining multiphysics numerical simulation and machine learning. By using particle swarm optimization algorithm to globally optimize the prediction results of BPNN algorithm, the optimal combination of laser paint removal process parameters adapted to different workpieces and paint layer characteristics is successfully obtained, which can significantly improve the accuracy of laser paint removal.

[0055] (2) By predicting and analyzing the surface morphology parameters, this invention can effectively avoid problems such as incomplete paint removal or substrate ablation that occur in traditional laser paint removal, and achieve high-quality, low-damage paint removal treatment.

[0056] (3) The present invention uses the PSO algorithm to perform global search optimization of laser process parameters, which avoids the problems of traditional trial and error methods that rely on experience and are inefficient, and significantly reduces experimental costs and time consumption.

[0057] (4) The optimal combination of laser paint removal process parameters can be dynamically adjusted according to different workpiece materials, coating types and processing requirements, making it suitable for precise laser cleaning in complex structures and various application scenarios.

[0058] (5) This invention not only significantly improves the quality and efficiency of paint removal, but also effectively reduces thermal damage to the substrate, achieving the goal of high-precision, low-damage laser paint removal. It is particularly suitable for high-end manufacturing fields such as aircraft and rail transportation where surface treatment quality requirements are extremely high. It can achieve efficient and safe removal of surface coatings and has good engineering application value and promotion prospects. Attached Figure Description

[0059] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0060] Figure 2 This is a schematic diagram of the physical model structure for laser paint removal according to the present invention;

[0061] Figure 3 This is a schematic diagram of the BPNN algorithm prediction method of the present invention;

[0062] Figure 4 This is a flowchart of the PSO optimization process of the present invention. Detailed Implementation

[0063] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0064] In recent years, artificial intelligence has been increasingly widely used in manufacturing process modeling and parameter optimization. Among these, backpropagation neural networks (BPNNs), due to their excellent generalization and function approximation capabilities, can be used to establish mapping models between laser process parameters and surface morphology parameters. Particle swarm optimization (PSO) algorithms, with their simple structure and fast convergence speed, can efficiently search for optimal solutions in a multi-parameter space to achieve optimal paint removal quality indicators. Combining BPNNs and PSOs to construct a prediction and optimization system for precise laser paint removal can reduce reliance on experiments while enabling intelligent control of the processing, improving the consistency of paint removal quality and the level of substrate protection. This is of great significance for promoting the high-quality development of laser cleaning technology.

[0065] Therefore, this embodiment utilizes artificial intelligence technology to provide a method for optimizing laser paint removal process parameters based on PSO-BPNN, such as... Figure 1 As shown, the method includes the following steps:

[0066] Step 1: Collect the physical property parameters of the base layer and paint layer of the workpiece to be processed, establish a multiphysics simulation model of the laser paint removal process based on the actual workpiece structure, and complete the mesh generation and boundary condition setting.

[0067] Based on the typical paint layer structure of the workpiece to be treated (such as the outer surface of an aircraft), determine its lamination material composition, including the surface paint layer and the bottom substrate material.

[0068] By consulting material databases or conducting experimental tests, the thermal properties of each layer of material can be obtained, including key physical property data such as thermal conductivity, density, specific heat capacity, melting point, and latent heat of vaporization.

[0069] A three-dimensional multiphysics simulation model of the paint removal process was established using simulation software, with reasonable definition of the geometric structure and setting of material properties. The established multiphysics simulation model was meshed, with appropriate mesh size and density to balance computational accuracy and efficiency, and localized refinement was applied to the laser-affected area.

[0070] Set the boundary conditions involved in the simulation process, including laser heat source loading conditions (such as Gaussian spot, moving speed, pulse parameters, etc.), thermal convection and thermal radiation boundaries, and thermal conduction interface conditions. At the same time, apply an appropriate initial temperature field to ensure that the model has good convergence and rationality, and obtain the final multiphysics simulation model of the paint removal process. Complete the initialization and solution preparation of the multiphysics simulation model to lay the foundation for subsequent parameter simulation and optimization.

[0071] The multiphysics simulation model constructed in this embodiment is as follows: Figure 2 As shown, Figure 2 Midpoint-1: Pressure constraint point; Faces 1-8: Convective boundary, no-slip wall; Face 9: Adiabatic boundary, no-slip wall; Face 10: Air-material interface; Face 11: Outlet boundary, pressure outlet; Domain I: Air layer; Domain II (material coating): Heat flow (including heat source, thermal radiation, and convective heat transfer); Domain III (material matrix). The boundary conditions are following the colon.

[0072] Step 2: Under different combinations of laser paint removal process parameters, conduct multiphysics numerical simulations of the interaction between the laser and the coating and substrate to obtain the simulation results of the corresponding surface morphology parameters.

[0073] Based on existing laser models, typical parameter ranges such as laser power, pulse frequency, and scanning speed are set, and multiple representative combinations of laser process parameters are constructed as simulation input variables.

[0074] The constructed laser paint removal process parameters were input one by one into the established multiphysics simulation model, and the simulation was run to obtain the three-dimensional morphological changes and paint peeling characteristics of the workpiece surface after laser treatment.

[0075] Extract surface morphology parameters closely related to paint removal quality, specifically including: pits and bumps (h). b , pit depth h d Paint removal rate R r Initial coating thickness d0 before cleaning, and radius of transverse pit r hor Longitudinal pit radius r ver Roughness Ra, average transverse radius R hor Longitudinal average radius R ver and surface area ratio A / A i These, etc., are used as output data for the training and validation of subsequent neural network models.

[0076] Step 3: Establish an intelligent database containing material properties, laser process parameters, and surface morphology parameters, and use a backpropagation neural network (BPNN) to construct a mapping relationship model between materials, processes, and properties.

[0077] like Figure 3As shown; specifically:

[0078] Based on the aforementioned multiphysics simulation results, a training sample database was constructed. The samples cover a variety of typical materials (such as different matrix materials: aluminum alloy, carbon fiber composite materials; and epoxy resin paint layers of different thicknesses), multiple sets of laser paint removal process parameter combinations and corresponding surface morphology parameters.

[0079] The constructed database is preprocessed and classified using big data processing and cloud computing technologies. Cluster analysis, principal component analysis and other methods are introduced to analyze the similarity and sensitivity of key parameters among different material samples.

[0080] The backpropagation neural network algorithm was selected, and the model performance was optimized by adjusting the learning rate, weight initialization method, and number of hidden layer nodes. It learns from the database and establishes a mapping relationship between materials, processes, and properties; among other things, it combines... Figure 3 The training steps of the backpropagation neural network algorithm shown include the following:

[0081] (1) Initialization: Randomly initialize weights and biases;

[0082] (2) Forward Propagation: The laser paint removal process parameter combination (P, f, v1, s1, n) is input through the input layer, and the current layer output is obtained through the transfer function, which is then used as the input for the next layer, until the output layer obtains the predicted surface morphology parameters, denoted as... Where P is power, f is frequency, v1 is scanning speed, s1 is scanning interval, and n is the number of scans. The transfer function expression for the predicted pit protrusion, pit depth, transverse pit radius, longitudinal pit radius, roughness, transverse average radius, longitudinal average radius, and surface area ratio is as follows:

[0083]

[0084] In the formula, F sigmoid (z) is the Sigmoid activation function, used to map the input z to the (0,1) interval, where z is the weighted input of the neuron, i.e., the combination of laser paint removal process parameters, H. j Let n be the output of the j-th hidden layer neuron, n be the number of input layer neurons, p be the number of hidden layer neurons, and w be the output of the j-th hidden layer neuron. ij Let x be the connection weight from the i-th neuron in the input layer to the j-th neuron in the hidden layer. i Let a be the input value of the i-th neuron in the input layer, m be the number of neurons in the output layer, and a be the input value of the i-th neuron in the input layer. j y is the bias (threshold) of the j-th neuron in the hidden layer. k w is the output (predicted value) of the k-th neuron in the output layer. jkLet b be the connection weight from the j-th neuron in the hidden layer to the k-th neuron in the output layer. k This is the bias (threshold) for the k-th neuron in the output layer;

[0085] (3) Calculation error: The error between the predicted surface topography parameters and the corresponding simulated surface topography parameters is calculated using a loss function. The expression for the loss function (mean squared error) is as follows:

[0086]

[0087] In the formula, L is the loss function. The true value (target value) of the k-th output neuron is the simulation result of the surface morphology parameters, and the coefficients are... This is to simplify the calculation when taking the derivative (the derivative of the squared term is multiplied by 2 during gradient descent, which cancels out this coefficient);

[0088] (4) Backpropagation: Calculate the output layer error using the loss function, and then backpropagate the hidden layer error layer by layer, and then calculate the gradient of each layer.

[0089] (5) Update weights and biases: Update the weights and biases based on the backpropagation results. The hidden layer bias update expression is as follows:

[0090]

[0091] In the formula, a j (T+1) represents the bias of the j-th neuron in the hidden layer at the (T+1)-th iteration, i.e., the updated hidden layer bias, a j (T) represents the bias of the j-th neuron in the hidden layer at the T-th iteration, and μ is the learning rate (step size), which controls the magnitude of the parameter update. For the loss function L with respect to bias a j The gradient (partial derivative);

[0092] The expression for updating the weights from the input layer to the hidden layer is:

[0093]

[0094] In the formula, w ij (T+1) represents the weights from the input layer to the hidden layer at the (T+1)th iteration, i.e., the updated weights from the input layer to the hidden layer, w. ij (T) represents the weights from the input layer to the hidden layer at the T-th iteration. For the loss function L with respect to the weights w ij The gradient (partial derivative);

[0095] The output layer bias update expression is:

[0096]

[0097] In the formula, b k (T+1) represents the bias of the k-th neuron in the output layer at the (T+1)-th iteration, i.e., the updated output layer bias, b k (T) represents the bias of the k-th neuron in the output layer during the T-th iteration. For the loss function L with respect to bias b k The gradient (partial derivative);

[0098] The expression for updating the weights from the hidden layer to the output layer is:

[0099]

[0100] In the formula, w jk (T+1) represents the weights from the hidden layer to the output layer at the (T+1)th iteration, i.e., the updated weights from the hidden layer to the output layer, w. jk (T) represents the weights from the hidden layer to the output layer at the T-th iteration. For the loss function L with respect to the weights w jk The gradient (partial derivative);

[0101] Iteration: Return to step (2) and repeat the iteration until the iteration condition is met to obtain the trained BPNN algorithm.

[0102] Finally, cross-validation was used to train the model. By adjusting algorithm parameters and increasing training data, the prediction accuracy and generalization ability of the model were improved. At the same time, the performance of the model was evaluated and verified to ensure that it has good prediction accuracy, stability and robustness under various materials and different laser parameters, providing reliable performance prediction support for subsequent algorithm optimization.

[0103] Step 4: Construct an objective function based on the paint removal quality evaluation index, embed the BPNN prediction model into the particle swarm optimization (PSO) algorithm, perform global optimization of the laser process parameters, and obtain the laser paint removal process parameter combination corresponding to the optimal paint removal quality.

[0104] This step primarily utilizes the trained model to predict surface morphology parameters after laser treatment under different material and process parameters. Based on the prediction results, a paint removal quality evaluation index system is constructed using the particle swarm optimization algorithm, and a multi-objective optimization function is established accordingly. Common optimization objectives include: maximum paint removal rate, minimum surface roughness, minimum unit area ratio (or unit area energy consumption), and minimum residual paint film thickness. Under the set parameters such as particle number, iteration number, inertia weight, and learning factor, the PSO algorithm is run to iteratively search for the optimal solution, obtaining a set of laser paint removal process parameter combinations corresponding to the optimal paint removal effect, thus achieving global optimization configuration of process parameters under complex working conditions.

[0105] Specifically, in combination Figure 4 This step specifically includes the following:

[0106] 1) Initialize the particle swarm and define the multi-objective optimization function, and construct the mapping relationship between material-process-performance based on the BPNN algorithm to calculate the objective function value of each individual particle;

[0107] 2) Calculate the fitness value of individual particles and redetermine the extreme values ​​of individual particles and the global extreme value;

[0108] 3) The objective function value is used as the sole basis for updating the direction, to "evaluate" the quality of the particle's current position, and thus update the particle swarm position. The update expressions for the velocity vector and position vector of each individual particle are as follows:

[0109] v i (v i1 ,v i2 ,…,v id )

[0110] x i (x i1 ,x i2 ,…,x id )

[0111]

[0112] In the formula, v i Let v be the velocity vector of the i-th particle. id x is the velocity component of the particle in the d-th dimension. i Let x be the position vector of the i-th particle. id It is the position component of the particle in the d-th dimension. Let the updated velocity of the i-th particle in the d-th dimension be the velocity at iteration k+1. Let be the velocity of the i-th particle in the d-th dimension at the current k iterations; w be the inertia weight, used to balance the capabilities of global and local search; r1 and r2 be random numbers used to introduce randomness; and l1 and l2 be individual and social learning factors. The optimal position of particle i in the d-th dimension at the current k iterations is used as the historical best position. Let r be the current position of the particle in the d-th dimension. This represents the globally optimal position of the entire group in the d-th dimension. Let i be the updated position of particle i in the d-th dimension, which will be used as the position in the (k+1)th iteration.

[0113] 4) Calculate the fitness values ​​of the new generation of individual particles and redetermine the individual particle extreme values ​​and the global extreme values;

[0114] 5) Determine if the iteration conditions are met. If yes, stop the iteration, output the optimal individual particle, and obtain the optimal combination of laser paint removal process parameters. If not, return to step 3) to perform iterative calculations until the iteration conditions are met.

[0115] Step 5: Output the optimized laser process parameters and verify and apply them in the actual laser paint removal process to achieve efficient, accurate and low-damage coating removal on the target workpiece.

[0116] The process involves verification and application to achieve efficient, precise, and low-damage coating removal from the target workpiece; specifically:

[0117] Using laser processing equipment, parameters such as laser power, frequency, and scanning speed are set according to the optimization results to perform paint removal on actual workpieces.

[0118] The paint-removed area is characterized by a three-dimensional topography measuring instrument, optical microscope or scanning electron microscope, and microscopic detection and quantitative analysis are performed on the paint-removed area to obtain key evaluation indicators such as the geometric characteristics of the pits, the paint residue rate, and the range of the heat-affected zone (HAZ).

[0119] The paint removal effect under optimized parameters was compared and analyzed with the processing results under traditional empirical parameters. The advantages of the process were evaluated in terms of paint removal efficiency, substrate damage rate and surface quality. The results showed that the optimized parameters can effectively improve the paint removal quality and stability, meet the comprehensive performance requirements of laser paint removal process in high-end manufacturing fields in terms of precision, safety and process consistency, and have good practical application and promotion value.

[0120] In summary, the laser paint removal process parameter optimization method based on PSO-BPNN proposed in this invention achieves efficient correlation modeling between laser process parameters and paint removal effect through a combination of multiphysics numerical simulation and machine learning. By using particle swarm optimization algorithm to globally optimize the neural network prediction model, the optimal combination of laser paint removal process parameters adapted to different workpiece and paint layer characteristics was successfully obtained. Practical application verification shows that this method not only significantly improves paint removal quality and efficiency but also effectively reduces thermal damage to the substrate, achieving the goal of high-precision, low-damage laser paint removal, and has good engineering application value and promising prospects for widespread application.

[0121] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0122] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0123] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0124] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0125] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0126] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0127] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for optimizing process parameters of laser paint removal based on PSO-BPNN, characterized in that, The method comprises the following steps: Collecting physical parameters of the substrate and paint layer of the workpiece to be processed, and constructing a multi-physical field simulation model of the paint removal process; Based on the multi-physical field simulation model of the paint removal process, under the set different laser paint removal process parameter combinations, the multi-physical field numerical simulation of the interaction between laser and paint layer and substrate is carried out, and the corresponding surface topography parameter simulation results are obtained; Based on different material physical parameters, laser paint removal process parameter combinations and surface topography parameter simulation results, a data set is constructed, and the PSO-BPNN prediction optimization algorithm is used to predict the optimal laser paint removal process parameter combination, wherein the BPNN algorithm is used to construct the mapping relationship between material-technology-performance, and the PSO algorithm is used for particle swarm optimization to obtain the optimal laser paint removal process parameter combination, and each particle in the particle swarm represents a laser paint removal process parameter combination.

2. The method of claim 1, wherein the method is a PSO-BPNN-based laser paint removal process parameter optimization method. The step of constructing the multi-physical field simulation model of the paint removal process comprises: Based on the physical parameters of the substrate and paint layer of the workpiece to be processed, an initial multi-physical field simulation model of the paint removal process is established in the simulation software; The initial multi-physical field simulation model is meshed, the laser action area is locally encrypted, and the boundary conditions in the simulation process are set to obtain the final multi-physical field simulation model of the paint removal process.

3. The method of claim 2, wherein the method is based on PSO-BPNN. The boundary conditions include initial temperature field, laser heat source loading condition, heat convection and heat radiation boundary, and heat conduction interface condition.

4. The method of claim 1, wherein the method is a PSO-BPNN-based laser paint removal process parameter optimization method. The step of obtaining the corresponding surface topography parameter simulation results comprises: According to the existing laser, a plurality of laser paint removal process parameter combinations are constructed as simulation input variables; The laser paint removal process parameter combinations are input into the multi-physical field simulation model of the paint removal process one by one for simulation, the three-dimensional topography change and paint layer peeling characteristics of the workpiece surface after laser treatment are obtained, and the surface topography parameter simulation results closely related to paint removal are extracted.

5. The method of claim 1, wherein the method is based on PSO-BPNN for laser paint removal process parameter optimization. It also includes the steps of preprocessing and classifying the database, using cluster analysis method for clustering grouping, using principal component analysis method for dimension reduction and quantifying the sensitivity of key parameters.

6. The method of claim 1, wherein the method is a PSO-BPNN-based laser paint removal process parameter optimization method. The laser paint removal process parameters in the laser paint removal process parameter combination include power P, frequency f, scanning speed v1, scanning spacing s1 and scanning times n.

7. The method of claim 1, wherein the method is a PSO-BPNN-based laser paint removal process parameter optimization method. The step of predicting the optimal laser paint removal process parameter combination comprises: 1) Initialize the particle swarm and define the multi-objective optimization function, and calculate the objective function value of each particle individual based on the BPNN algorithm to construct the mapping relationship between material-technology-performance; 2) Calculate the fitness value of the particle individual and re-determine the particle individual extreme value and global extreme value; 3) The target function value is used as the basis for updating the direction of the particle individual, and the velocity vector and position vector of the particle individual are updated, and the update expressions are respectively: wherein, is the updated velocity of the ith particle individual on the dth dimension, as the velocity at iteration k+1, is the velocity of the ith particle individual on the dth dimension at the current iteration k, w is the inertia weight used to balance the ability of global and local search, r1, r2 are random numbers used to introduce randomness, and l1, l2 are individual, social learning factors, is the individual best position of the particle individual i on the dth dimension at the current iteration k, as the historical best position, is the current position of the particle individual i on the dth dimension, is the global best position of the whole swarm on the dth dimension, is the updated position of the particle individual i on the dth dimension, as the position at iteration k+1. 4) Calculate the fitness value of the new generation of particle individuals and re-determine the particle individual extreme value and global extreme value; 5) Determine whether the iteration condition is met, if yes, stop iteration, output the optimal particle individual, and obtain the optimal laser paint removal process parameter combination, if not, return to step 3) for iterative calculation until the iteration condition is met.

8. The PSO-BPNN-based laser paint removal process parameter optimization method according to claim 7, characterized in that, The expression of the multi-objective optimization function is: f(x) = max(0, A / A t -1) + max(0, Ra-Ra max ) + max(0, 1-R r ) In the formula, f(x) is a multi-objective optimization function, A / A t is a surface area ratio, Ra is a roughness, Ra max is a maximum roughness, R r is a paint removal rate, A t , and A are surface areas before and after laser treatment, respectively.

9. The PSO-BPNN based laser paint removal process parameter optimization method according to claim 1, wherein, The training step using the BPNN algorithm comprises: (1) Initialization: randomly initializing the weights and biases; (2) Forward propagation: inputting the laser paint removal process parameter combination through the input layer, obtaining the output of the current layer through a transfer function, and taking the output as the input of the next layer until the predicted surface topography parameter is obtained at the output layer, wherein the expression of the transfer function is: where F sigmoid (z) is a sigmoid activation function used to map the input z to the interval (0, 1), z is the weighted input of a neuron, i.e., the laser paint removal process parameter combination, H j is the output of the jth hidden layer neuron, n is the number of input layer neurons, p is the number of hidden layer neurons, w ij is the connection weight from the ith neuron of the input layer to the jth neuron of the hidden layer, x i is the input value of the ith neuron of the input layer, m is the number of output layer neurons, a j is the bias of the jth neuron of the hidden layer, y k is the output of the kth neuron of the output layer, w jk is the connection weight from the jth neuron of the hidden layer to the kth neuron of the output layer, b k is the bias of the kth neuron of the output layer; (3) Error calculation: calculating the error between the predicted surface topography parameter and the corresponding surface topography parameter simulation result using a loss function, wherein the expression of the loss function is: In the formula, L is a loss function, is the true value of the kth output neuron, i.e., the surface topography parameter simulation result; (4) Backward propagation: calculating the output layer error using the loss function, and recursively calculating the hidden layer error in reverse, and then calculating the gradient of each layer; (5) Updating the weights and biases: updating the weights and biases according to the backward propagation result, wherein the expression for updating the hidden layer bias is: wherein a j (T+1) is the bias of the jth neuron of the hidden layer at the (T+1)th iteration, i.e., the updated hidden layer bias, a j (T) is the bias of the jth neuron of the hidden layer at the Tth iteration, μ is the learning rate, controlling the magnitude of the parameter update, is the gradient of the loss function L with respect to the bias a j . The expression for updating the weights from the input layer to the hidden layer is: where w ij (T+1) is the weight from the input layer to the hidden layer at the T+1th iteration, i.e., the updated weight from the input layer to the hidden layer, w ij (T) is the weight from the input layer to the hidden layer at the Tth iteration, is the gradient of the loss function L with respect to the weight w ij . The expression for updating the output layer bias is: where b k (T+1) is the bias of the kth neuron of the output layer at the T+1th iteration, i.e., the updated output layer bias, b k (T) is the bias of the kth neuron of the output layer at the Tth iteration, is the gradient of the loss function L with respect to the bias b k ; The expression for updating the weights from the hidden layer to the output layer is: where w jk (T+1) is the weight from the hidden layer to the output layer at the T+1th iteration, i.e., the updated weight from the hidden layer to the output layer, w jk (T) is the weight from the hidden layer to the output layer at the Tth iteration, is the gradient of the loss function L with respect to the weight w jk . Iteration: returning to step (2) for repeated iteration until the iteration condition is reached, and obtaining the trained BPNN algorithm.

10. A computer-readable storage medium, characterized in that, One or more programs for execution by one or more processors of an electronic device, the one or more programs including instructions for performing the PSO-BPNN-based laser paint removal process parameter optimization method according to any one of claims 1-9.

Images