Reverse prediction method and device for niobium alloy laser welding process and medium

Abstract

The embodiment of the invention discloses a reverse prediction method and device for a niobium alloy laser welding process and a medium, and relates to the technical field of laser welding. The welding process performance demand information comprises any one or more of a high-temperature oxidation resistance index, a welding depth index, a welding width-depth ratio index, a heat affected zone width index, a tensile strength index and a shear strength index; and according to the welding process performance demand information, welding process parameters corresponding to niobium alloy laser welding are determined by adopting a pre-constructed laser welding process parameter reverse prediction model, and the laser welding process parameter reverse prediction model is a back propagation neural network model based on grey wolf optimization and differential evolution algorithm optimization. According to the method, a large number of repeated tests and cost waste caused by a traditional trial and error method are avoided, and the problem that performance and parameter matching is difficult in the niobium alloy laser welding process is solved.

Description

Technical Field

[0001] The present invention relates to the field of laser welding technology, and in particular to a reverse prediction method, equipment and medium for niobium alloy laser welding process. Background Art

[0002] With the booming commercial aerospace industry, the demand for materials that are both lightweight and have excellent strength is growing. In this context, niobium alloys have emerged as an ideal material choice in the pursuit of high performance and lightweight spacecraft with their excellent high specific strength, high temperature resistance and corrosion resistance. In order to fully tap the potential of niobium alloys, precise welding technology has become a crucial link in connecting these high-performance materials. Among many welding processes, laser welding technology has occupied a preferred position in the connection of niobium alloys due to its high precision, high energy density and flexibility. However, the laser welding process is extremely complex, covering multiple interrelated process parameters such as laser power, welding speed, focus positioning and shielding gas. The synergistic effect of these parameters plays a decisive role in the quality of welding. Improper parameter settings may cause welding defects such as cracks, pores, oxidation, etc. These defects will affect the quality of the weld and the performance of the joint; in addition, niobium alloys have a high melting point and poor molten pool fluidity, which requires extremely high welding accuracy and equipment stability, and is difficult to control.

[0003] In actual application scenarios, the laser welding process often pays more attention to performance requirements. Traditional predictions all use forward predictions. When faced with specific performance requirements, technicians still need to have rich experience to accurately obtain input parameters. In other words, when selecting traditional laser welding process parameters, most of them rely on experience and are inefficient, and cannot meet the continuous stability requirements of welding quality. Summary of the invention

[0004] One or more embodiments of the present specification provide a reverse prediction method, device and medium for niobium alloy laser welding process, which are used to solve the following technical problems: When selecting traditional laser welding process parameters, most of them rely on experience and are inefficient, and cannot meet the requirements of continuous stability of welding quality.

[0005] One or more embodiments of this specification adopt the following technical solutions:

[0006] One or more embodiments of the present specification provide a reverse prediction method for a niobium alloy laser welding process, the method comprising: obtaining welding process performance requirement information corresponding to niobium alloy laser welding, wherein the welding process performance requirement information comprises any one or more of a high temperature oxidation resistance index, a welding depth index, a welding width-to-depth ratio index, a heat affected zone width index, a tensile strength index, and a shear strength index; determining the welding process parameters corresponding to the niobium alloy laser welding according to the welding process performance requirement information through a pre-constructed laser welding process parameter reverse prediction model, wherein the laser welding process parameter reverse prediction model is a back propagation neural network model optimized based on grey wolf optimization and differential evolution algorithms.

[0007] Furthermore, before determining the welding process parameters corresponding to the niobium alloy laser welding according to the welding process performance requirement information through the pre-constructed laser welding process parameter inverse prediction model, the method also includes: acquiring an initial niobium alloy laser welding data set, wherein the initial niobium alloy laser welding data set includes niobium alloy welding process parameters of multiple thicknesses and corresponding welding performance indicators; performing data feature analysis on the initial niobium alloy laser welding data set to determine at least one main welding performance indicator and at least one key welding process parameter to construct a laser welding process data set; constructing a inverse mapping model between welding performance indicators and welding process parameters based on a back propagation neural network method optimized by grey wolf optimization and differential evolution algorithms, wherein the input of the inverse mapping model is the at least one main welding performance indicator, and the output is the at least one key welding process parameter; training the inverse mapping model through the laser welding process data set to determine the laser welding process parameter inverse prediction model.

[0008] Furthermore, data feature analysis is performed on the initial niobium alloy laser welding data set to determine at least one main welding performance index and at least one key welding process parameter, specifically including: obtaining actual application environment parameters of the niobium alloy laser welded formed part to determine at least one main welding performance index corresponding to the niobium alloy laser welded formed part based on the actual application environment parameters; and using the fuzzy-decision laboratory analysis method to identify at least one key welding process parameter in the welding process according to the at least one main welding performance index.

[0009] Furthermore, the fuzzy-decision laboratory analysis method is used to identify at least one key welding process parameter in the welding process according to the at least one main welding performance indicator, specifically including: using fuzzy triangular number form to score the influence between multiple welding process parameters, obtaining multiple expert scoring data, and constructing a corresponding fuzzy influence matrix based on each expert scoring data; summarizing the multiple fuzzy influence matrices by weighted averaging to obtain a comprehensive fuzzy matrix, defuzzifying the comprehensive fuzzy matrix by using the centroid method, and normalizing it to determine a normalized matrix; calculating the total influence matrix between the multiple welding process parameters according to the normalized matrix, so as to calculate the influence degree between each welding process parameter and the influence degree between each welding process parameter based on the total influence matrix; sorting the multiple welding process parameters according to the influence degree between each welding process parameter and the influence degree between each welding process parameter to determine at least one key welding process parameter.

[0010] Further, according to the normalized matrix, the total influence matrix between the plurality of welding process parameters is calculated, specifically comprising: according to the normalized matrix, according to the formula , calculate the total influence matrix between the multiple welding process parameters, where T is the total influence matrix and I is the unit matrix.

[0011] Furthermore, the plurality of welding process parameters are sorted by the influence between each welding process parameter and the influence between each welding process parameter to determine at least one key welding process parameter, specifically comprising: obtaining the importance coefficient of each welding process parameter according to the sum of the influence between each welding process parameter and the influence between each welding process parameter; and sorting the plurality of welding process parameters in order from large to small according to the importance coefficient of each welding process parameter to determine at least one key welding process parameter among the plurality of welding process parameters.

[0012] Furthermore, based on the back propagation neural network method optimized by gray wolf optimization and differential evolution algorithm, an inverse mapping model between welding performance indicators and welding process parameters is constructed, which specifically includes: setting the parameters of gray wolf optimization algorithm and differential evolution algorithm, and representing each individual as the weight and bias of the neural network; performing fitness calculation through the error function of BP neural network to determine the fitness value of each individual, wherein the fitness value of each individual is used to represent the error of each weight and bias combination on the training set; performing a global search through the encirclement and hunting behavior of the gray wolf algorithm, using three gray wolves to guide the search, and updating the position of the gray wolves; when performing mutation operation, three individuals are randomly selected, and the mutation of individuals is realized by calculating the difference vector between them, and a crossover operation is used to generate new individuals, and the dominant individuals are selected according to the fitness values ​​of the new and old individuals; the BP neural network is trained using the optimized weights and biases to construct an inverse prediction model between performance indicators and input parameters.

[0013] Further, after the inverse mapping model is trained through the laser welding process data set to determine the inverse prediction model of the laser welding process parameters, the method further includes: calculating at least one model evaluation index according to the true value and the predicted value corresponding to the individual with the smallest fitness value, so as to evaluate the inverse prediction model of the laser welding process parameters based on the at least one model evaluation index, wherein the model evaluation index includes any one of the model determination coefficient, the model mean absolute error and the model mean deviation error; wherein, , R² is the coefficient of determination of the model, and N is the number of samples; It is the true value corresponding to the individual with the smallest fitness value; is the predicted value corresponding to the individual with the smallest fitness value, , MAE is the mean absolute error of the model, , MBE is the model mean bias error.

[0014] One or more embodiments of this specification provide a reverse prediction device for a niobium alloy laser welding process, including:

[0015] at least one processor; and,

[0016] a memory communicatively connected to the at least one processor; wherein,

[0017] The memory stores instructions that can be executed by the at least one processor. The instructions are executed by the at least one processor so that the at least one processor can perform the above-mentioned inverse prediction method for the niobium alloy laser welding process.

[0018] One or more embodiments of the present specification provide a non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions are configured to: execute the above-mentioned reverse prediction method for the niobium alloy laser welding process.

[0019] At least one of the above technical solutions adopted in the embodiments of this specification can achieve the following beneficial effects: through the above technical solution, by directly predicting welding parameters according to specific performance requirements such as high-temperature oxidation resistance, welding depth, welding width-to-depth ratio, heat-affected zone width, tensile strength and shear strength, it can be ensured that the welding results accurately meet the design requirements, avoiding a large number of repeated experiments and cost waste caused by the traditional trial and error method, and improving production efficiency; a back propagation (BP) neural network model optimized based on the gray wolf optimization (GWO) and differential evolution (DE) algorithms is constructed. The model can reversely predict the optimal welding process parameters according to the required welding performance, optimize the welding parameters based on a large amount of data, ensure the stability of the welding process, reduce welding defects, and by accurately controlling the welding parameters, it can significantly improve the performance consistency between welded parts to meet high-quality production requirements; the model can automatically predict the optimal welding parameters according to performance requirements, reduce manual intervention, improve the intelligence level of the production process, and solve the problem of difficult performance and parameter matching in the laser welding process of niobium alloys. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art description. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. In the drawings:

[0021] Figure 1 A schematic flow chart of a reverse prediction method for a niobium alloy laser welding process provided in an embodiment of this specification;

[0022] Figure 2 A schematic diagram of a model construction process of a laser welding process parameter reverse prediction model provided in an embodiment of this specification;

[0023] Figure 3 A schematic diagram for visualizing the influence of a process parameter provided in an embodiment of this specification;

[0024] Figure 4 A schematic diagram of a BP neural network structure provided in the embodiments of this specification;

[0025] Figure 5 A schematic diagram of the structure of a reverse prediction device for a niobium alloy laser welding process provided in an embodiment of this specification. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0027] The embodiments of this specification provide a reverse prediction method for a niobium alloy laser welding process. It should be noted that the execution subject in the embodiments of this specification may be a server or any device with data processing capabilities. Figure 1 A schematic diagram of a reverse prediction method for a niobium alloy laser welding process provided in an embodiment of this specification is shown in FIG. Figure 1 As shown, it mainly includes the following steps:

[0028] Step S101, obtaining welding process performance requirement information corresponding to niobium alloy laser welding.

[0029] The welding process performance requirement information includes any one or more of a high temperature oxidation resistance index, a welding depth index, a welding width-to-depth ratio index, a heat affected zone width index, a tensile strength index and a shear strength index.

[0030] In one embodiment of the present specification, the performance standards required for niobium alloy laser welding are determined, and performance indicators directly related to welding quality are collected and analyzed, including but not limited to high temperature oxidation resistance index, welding depth index, welding width-to-depth ratio index, heat-affected zone width index, tensile strength index and shear strength index. Specific welding process performance requirement information can select and combine corresponding performance indicators according to specific application scenarios.

[0031] Step S102: According to the welding process performance requirement information, a pre-built laser welding process parameter reverse prediction model is used to determine the welding process parameters corresponding to the niobium alloy laser welding.

[0032] Among them, the inverse prediction model of laser welding process parameters is a back propagation neural network model based on grey wolf optimization and differential evolution algorithm optimization.

[0033] In one embodiment of the present specification, the welding process performance requirement information obtained in step S101 is input into the reverse prediction model. The model automatically calculates the welding process parameters that meet these performance requirements based on the built-in algorithm and training data, including laser power, welding speed, focus position and shielding gas, etc., and outputs one or more groups of welding process parameter combinations that meet the performance requirements for reference in actual welding operations.

[0034] According to the welding process performance requirement information, before determining the welding process parameters corresponding to the niobium alloy laser welding using a pre-constructed laser welding process parameter reverse prediction model, the method also includes: obtaining an initial niobium alloy laser welding data set, wherein the initial niobium alloy laser welding data set includes niobium alloy welding process parameters of multiple thicknesses and corresponding welding performance indicators; performing data feature analysis on the initial niobium alloy laser welding data set to determine at least one main welding performance indicator and at least one key welding process parameter to construct a laser welding process data set; constructing a reverse mapping model between welding performance indicators and welding process parameters based on a back propagation neural network method optimized by grey wolf optimization and differential evolution algorithms, wherein the input of the reverse mapping model is the at least one main welding performance indicator, and the output is the at least one key welding process parameter; and training the reverse mapping model through the laser welding process data set to determine the laser welding process parameter reverse prediction model.

[0035] In one embodiment of the present specification, by analyzing the performance after welding and predicting the optimal welding process parameters, a prediction model based on the Grey Wolf algorithm combined with differential evolution and optimized BP neural network is pre-constructed to provide efficient, accurate and reliable parameter guidance for the welding process. Figure 2 A schematic diagram of a model construction process of a laser welding process parameter reverse prediction model provided in an embodiment of this specification, such as Figure 2 As shown in the figure, the model building process includes the following steps: niobium alloy laser welding data collection, welding data feature analysis, data set division and normalization, construction and training of GWO-DE-BP neural network inverse model, model evaluation and other steps.

[0036] In one embodiment of the present specification, the laser welding of niobium alloy includes tantalum-niobium refractory alloy, niobium-hafnium alloy, etc., and its thickness is a key factor affecting the selection of laser welding parameters. According to the thickness of niobium alloy, it is divided into three categories: thin plate (thickness ≤ 1mm), medium plate (1mm < thickness ≤ 3mm) and thick plate (3mm < thickness). Therefore, during the data collection process, the welding process parameters are adjusted according to the thickness of the alloy, including laser power, welding speed, shielding gas flow rate and oscillation frequency, etc., and the performance indicators after welding are evaluated, such as welding depth, welding width-to-depth ratio, heat-affected zone width, tensile strength, high temperature oxidation resistance and shear strength, etc. A detailed and accurate niobium alloy laser welding process database is constructed based on the data collection results to provide solid process support for the establishment of the prediction model.

[0037] In order to avoid increasing the difficulty of model training due to too many parameters, it is necessary to analyze the data characteristics of the initial niobium alloy laser welding data set to clarify the core process parameters and key performance indicators that affect the welding performance. The GWO-DE-BP neural network method (Grey Wolf Optimizer - Differential Evolution - Backpropagation Neural Network, BPNN) optimized based on the Grey Wolf Optimizer (GWO) and Differential Evolution (DE) algorithm is used to construct a reverse mapping model between welding performance and parameters. The post-welding performance data such as tensile strength, high temperature oxidation resistance and welding depth are used as input, and the welding process parameters such as laser power, welding speed, oscillation frequency and focal distance are used as output to reveal the complex nonlinear relationship between performance and process parameters.

[0038] The initial niobium alloy laser welding data set is subjected to data feature analysis to determine at least one main welding performance index and at least one key welding process parameter, specifically comprising: obtaining actual application environment parameters of the niobium alloy laser welded formed part to determine at least one main welding performance index corresponding to the niobium alloy laser welded formed part based on the actual application environment parameters; and using a fuzzy-decision laboratory analysis method to identify at least one key welding process parameter in the welding process according to the at least one main welding performance index.

[0039] In one embodiment of the present specification, the main performance indicators that must be met by the welded molded part are determined according to the actual application environment of the welded molded part. In order to determine the main welding performance indicators of the niobium alloy laser welded molded part, it is first necessary to understand its actual application environment. The environmental parameters include but are not limited to working temperature, working pressure, vibration and shock, size and shape, etc. The working temperature is used to understand the temperature range in which the molded part needs to work to determine its tolerance requirements for high temperature environments. The working pressure indicates the pressure that the molded part may withstand during work to evaluate its structural strength and durability. The working environment medium includes whether there are corrosive gases, liquids or solid substances, which will affect the corrosion resistance and service life of the welded joint. Vibration and shock refer to the vibration and shock that the molded part may be subjected to during work, which will affect its fatigue resistance. Size and shape refer to the size and shape of the molded part that have a direct impact on its welding performance and manufacturing difficulty. Based on the actual application environment parameters obtained, at least one of the following main welding performance indicators can be determined: tensile strength, shear strength, fatigue strength, corrosion resistance, etc. Tensile strength is used to evaluate the bearing capacity of the welded joint under tensile force to ensure its structural stability. Shear strength is used to measure the performance of welded joints under shear force to ensure that they will not fail when subjected to shear force. Fatigue strength is used to evaluate the ability of welded joints to continue working under cyclic loads and their fatigue resistance. Corrosion resistance is used to evaluate the performance of welded joints in corrosive environments based on the working environment medium to ensure their service life. Fuzzy Decision Laboratory Analysis (FDLA) is then used to identify the key process parameters that play a decisive role in the welding process based on these performance indicators.

[0040] The fuzzy-decision laboratory analysis method is used to identify at least one key welding process parameter in the welding process according to the at least one main welding performance index, specifically including: using fuzzy triangular number form to score the influence between multiple welding process parameters, obtaining multiple expert scoring data, and constructing a corresponding fuzzy influence matrix based on each expert scoring data; summarizing multiple fuzzy influence matrices by weighted average to obtain a comprehensive fuzzy matrix, defuzzifying the comprehensive fuzzy matrix by using the centroid method, and normalizing it to determine a normalized matrix; calculating the total influence matrix between multiple welding process parameters according to the normalized matrix, so as to calculate the influence degree between each welding process parameter and the influence degree between each welding process parameter based on the total influence matrix; sorting multiple welding process parameters according to the influence degree between each welding process parameter and the influence degree between each welding process parameter to determine at least one key welding process parameter.

[0041] According to the normalized matrix, a total influence matrix between the plurality of welding process parameters is calculated, specifically comprising: according to the normalized matrix, according to the formula , calculate the total influence matrix between multiple welding process parameters, where T is the total influence matrix and I is the unit matrix.

[0042] According to the influence degree between each welding process parameter and the influence degree between each welding process parameter, a plurality of welding process parameters are sorted to determine at least one key welding process parameter, specifically comprising: obtaining the importance coefficient of each welding process parameter according to the sum of the influence degree between each welding process parameter and the influence degree between each welding process parameter; and sorting the plurality of welding process parameters in order from large to small according to the importance coefficient of each welding process parameter to determine at least one key welding process parameter among the plurality of welding process parameters.

[0043] In one embodiment of the present specification, based on the laser welding process parameters of niobium alloy, such as laser power, welding speed, oscillation frequency, etc., experts are invited to score the influence of process parameters in the form of fuzzy triangular numbers, and each expert constructs a fuzzy influence matrix.

[0044] (1)

[0045] Among them, i and j represent different process parameters, represents the influence of process parameter i on parameter j, and k represents the fuzzy influence matrix corresponding to the kth expert. It should be noted that the scoring criteria are set as follows: no influence (0, 0, 0.25); weak influence (0.25, 0.5, 0.75); medium influence (0.5, 0.75, 1.0); strong influence (0.75, 1.0, 1.25); extremely strong influence (1.0, 1.25, 1.5).

[0046] The scores of all experts are combined by weighted average to obtain the comprehensive fuzzy matrix , and then use the centroid method to defuzzify it to obtain an influence matrix with clear values ,in, To affect the elements in the matrix, normalize them at the same time to obtain the normalized matrix , where each element of the normalized matrix The calculation formula is:

[0047] (2)

[0048] According to the normalized matrix Calculate the full impact matrix , the calculation formula is as follows:

[0049] (3)

[0050] in, is the unit matrix, and T is the total influence matrix among various process parameters.

[0051] Calculate the influence degree between each process parameter based on the full influence matrix and impact , where the calculation formula is as follows:

[0052] (4)

[0053] Among them, the impact is the total influence of process parameter i on other process parameters; is the degree to which process parameter i is affected by other process parameters, and are all matrix elements in the total impact matrix. + Sort the process parameters. + The larger the value, the higher the importance of process parameter i, and the ranking of the influence of process parameters can be obtained accordingly; + and - The horizontal and vertical axes are used to visualize the mutual influence relationship between process parameters.

[0054] Based on the back propagation neural network method optimized by gray wolf optimization and differential evolution algorithm, an inverse mapping model between welding performance indicators and welding process parameters is constructed, which specifically includes: setting the parameters of gray wolf optimization algorithm and differential evolution algorithm, and representing each individual as the weight and bias of the neural network; performing fitness calculation through the error function of BP neural network to determine the fitness value of each individual, wherein the fitness value of each individual is used to represent the error of each weight and bias combination on the training set; performing global search through the encirclement and hunting behavior of the gray wolf algorithm, using three gray wolves to guide the search, and updating the position of the gray wolves; when performing mutation operation, three individuals are randomly selected, and the individual mutation is realized by calculating the difference vector between them, and the crossover operation is used to generate new individuals, and the dominant individuals are selected according to the fitness values ​​of the new and old individuals; the BP neural network is trained using the optimized weights and biases to construct an inverse prediction model between performance indicators and input parameters.

[0055] In one embodiment of the present specification, the laser welding process data set is precisely divided according to a specific ratio (such as 8:2) to form independent training sets and test sets; the divided data sets are then normalized using a sigmoid function (as shown in Formula 5) to ensure data consistency and model training efficiency.

[0056] (5)

[0057] The GWO-DE-BP neural network method is used to construct the inverse mapping model between welding performance and parameters. First, the population is initialized. The parameters of the gray wolf algorithm and the differential evolution algorithm are set. Expressed as the weight and bias of the BP neural network, assuming that there are N individuals and a D-dimensional solution vector (corresponding to the number of parameters of the BP neural network), then , (i=1,2, …, N), each Both represent a combination of weights and biases, expressed as parameters of the neural network.

[0058] The second step is to calculate the fitness. The fitness function generally uses the error function of the BP neural network, and the formula is:

[0059] (6)

[0060] Where n is the number of samples in the training set; , is the true value of the i-th sample, is the output predicted by the BP neural network; the fitness value f(x i ) is the error of the corresponding network on the training set.

[0061] The third step is to conduct a global search of the gray wolf. The global search of the gray wolf is carried out through encirclement and hunting behaviors, which is mainly guided by three main gray wolves, namely Alpha (𝛼), Beta (𝛽), and Delta (𝛿). Assume that the current individual position formula is:

[0062] (7)

[0063] in, , and They are the positions of the top three optimal solutions, is the updated position of the current gray wolf. When the gray wolf updates its position, its position is adjusted according to the distance from Alpha, Beta and Delta. The position update formula is:

[0064] (8)

[0065] in, and is the coefficient vector, , , and is a random vector between 0 and 1, It decreases linearly from 2 to 0.

[0066] Then, differential evolution search is performed. Differential evolution is further optimized mainly through mutation, crossover and selection operations. First, the mutation operation is performed by randomly selecting three individuals and calculating the differential vector for mutation. The formula is:

[0067] (9)

[0068] in, , and is a randomly selected individual, is the scaling factor, usually taken as .

[0069] Then, a crossover operation is used to generate new individuals. The crossover formula is:

[0070] (10)

[0071] in, is the crossover probability, usually taking the value , is a randomly chosen index.

[0072] Finally, the selection operation is used to select new individuals according to the fitness function (Equation (6)) Or old individuals The more advantageous individuals enter the next iteration. The selection formula is:

[0073] (11)

[0074] Next, the network is trained. The optimized weights and biases are assigned to the BP neural network for training, and an inverse prediction model is constructed between performance indicators such as tensile strength, high temperature oxidation resistance and welding depth and input parameters such as laser power, welding speed, oscillation frequency and focal distance.

[0075] After building the model, the error calculation is performed. Based on formula (6), the difference between the predicted value of the current model and the true value is measured.

[0076] Finally, the termination condition is judged. When the maximum iteration condition is reached or the error converges to the preset value, the algorithm is stopped and the prediction model is output. If the termination condition is not met, the network weights and biases are updated and the network training step is returned. The error back propagation algorithm of the BP network updates the weights and biases in each round of iteration, and its formula is:

[0077] (12)

[0078] in, is the learning rate, is the loss function, is the weight.

[0079] After the reverse mapping model is trained through the laser welding process data set to determine the reverse prediction model of the laser welding process parameters, the method further includes: calculating at least one model evaluation index according to the true value and the predicted value corresponding to the individual with the smallest fitness value, so as to evaluate the reverse prediction model of the laser welding process parameters based on the at least one model evaluation index, and the model evaluation index includes any one of the model determination coefficient, the model mean absolute error and the model mean deviation error; wherein, , R² is the coefficient of determination of the model, and N is the number of samples; It is the true value corresponding to the individual with the smallest fitness value; is the predicted value corresponding to the individual with the smallest fitness value, , MAE is the mean absolute error of the model, , MBE is the model mean bias error.

[0080] In one embodiment of this specification, in order to comprehensively evaluate the performance of the model, three key evaluation indicators, namely, coefficient of determination (R²), mean absolute error (MAE), and mean bias error (MBE), are introduced. At the same time, the model is rigorously tested and verified using a ten-fold cross-validation method to ensure the accuracy and reliability of the evaluation results.

[0081] (13)

[0082] (14)

[0083] (15)

[0084] Where N is the number of samples; It is the true value corresponding to the individual with the smallest fitness value; It is the predicted value corresponding to the individual with the smallest fitness value. R² evaluates the ability of the model to explain data variation. A value close to 1 indicates a high degree of fit. MAE measures the average error between the prediction and the actual value. The smaller the error, the more accurate it is. MBE indicates the direction of the prediction deviation. A positive number indicates that the value is lower than the true value, while a negative number indicates that the value is higher than the true value.

[0085] In one embodiment of the present specification, the constructed GWO-DE-BP neural network is applied to the actual niobium alloy laser welding process parameter prediction process, the performance requirements of the weld are input into the constructed GWO-DE-BP neural network, and the parameter prediction configuration is performed using the trained model. After obtaining the predicted data, the predicted value is re-verified using the experimental verification method, and the relative error between the predicted value and the experimental value is calculated using formula (16).

[0086] (16)

[0087] Among them, R e is the relative error between the predicted value and the experimental value, y exp is the experimental value, y pred is the predicted value.

[0088] Through the above technical scheme, by directly predicting welding parameters according to specific performance requirements such as high-temperature oxidation resistance, welding depth, welding width-to-depth ratio, heat-affected zone width, tensile strength and shear strength, it can be ensured that the welding results accurately meet the design requirements, avoiding a large number of repeated experiments and cost waste caused by the traditional trial and error method, and improving production efficiency; a back propagation (BP) neural network model optimized based on the grey wolf optimization (GWO) and differential evolution (DE) algorithms is constructed. The model can reversely predict the optimal welding process parameters according to the required welding performance, optimize the welding parameters based on a large amount of data, ensure the stability of the welding process, reduce welding defects, and significantly improve the performance consistency between welded parts by precisely controlling the welding parameters to meet high-quality production requirements; the model can automatically predict the optimal welding parameters according to performance requirements, reduce manual intervention, improve the intelligence level of the production process, and solve the problem of difficult performance and parameter matching in the laser welding process of niobium alloys.

[0089] The embodiment of this specification also provides another reverse prediction method for niobium alloy laser welding process, which specifically includes the following steps:

[0090] S1. Welding data collection. Taking the welding of 0.8 mm niobium alloy thin plate as an example, the process parameters such as laser power, welding speed and oscillation frequency used in the actual production process and the post-welding performance test data are collected. A total of 200 sets of data are collected. The following table is an example table of a data set provided in the embodiments of this specification.

[0091]

[0092] S2. Analysis of welding data characteristics. According to the data collection results, the most important performance indicators are selected based on practical applications. The embodiments of this specification select tensile strength, high temperature oxidation resistance and welding depth as performance indicators. Taking laser power, welding speed, oscillation frequency, focal distance and shielding gas flow as examples, the fuzzy-decision laboratory analysis method is used to determine the three most critical factors affecting welding.

[0093] S21. Five experts were invited to score the impact of process parameters using fuzzy triangular numbers. The scoring matrix of one expert is as follows:

[0094] (16)

[0095] S22. The scores of all experts are integrated by weighted average to obtain a comprehensive fuzzy matrix , and then use the centroid method to defuzzify it to obtain an influence matrix with clear values , and normalize it at the same time to get the normalized matrix Among them, comprehensive fuzzy , Impact Matrix And the normalized matrix N is:

[0096] (17)

[0097] (18)

[0098] (19)

[0099] S23. Calculate the total influence matrix T according to formula (3).

[0100] (20)

[0101] S24, calculate the influence degree between each process parameter based on the full influence matrix and impact , They are 3.16, 2.75, 2.53, 2.70 and 2.23 respectively. They are 0.16, 0.18, 0.05, -0.17 and -0.22 respectively.

[0102] S25, according to + Sort the process parameters. + The larger the value, the higher the importance of process parameter i, and the ranking of the influence of process parameters can be obtained accordingly; + and - To realize the visualization of the mutual influence relationship between process parameters for the horizontal and vertical axes, Figure 3 A schematic diagram for visualizing the influence of process parameters provided in the embodiments of this specification, such as Figure 3 The results show that the importance of process parameters is ranked as laser power, welding speed and focal distance.

[0103] S3. Dataset division and normalization: According to the 500 sets of experimental data imported from the process database, the training set and the test set are divided into a ratio of 8:2, and normalized using formula (5).

[0104] S4. Construct and train the GWO-DE-BP neural network inverse model. The GWO-DE-BP neural network method is used to construct the inverse mapping model between welding performance and parameters. The model uses tensile strength, high temperature oxidation resistance and welding depth as inputs and laser power, welding speed and focal distance as outputs, revealing the complex nonlinear relationship between performance and process parameters.

[0105] S41. Set the initial population size of gray wolves to 100, the maximum number of iterations to 200, the mutation factor of differential evolution to fluctuate between 0.3 and 0.7, and the crossover probability to 0.5.

[0106] S42. Calculate the fitness function according to formula (6).

[0107] S43. Perform a global search for gray wolves based on equations (7) and (8).

[0108] S44. Perform differential evolution search according to equations (9), (10) and (11).

[0109] S45, Figure 4 A schematic diagram of a BP neural network structure provided in the embodiment of this specification is used as follows Figure 4 As shown in the network structure diagram, an inverse prediction model between tensile strength, high temperature oxidation resistance and welding depth and laser power, welding speed and focal distance is constructed. The number of training times is set to 50 times, the target error is 10-4 and the learning rate is 0.01, and continuous training is performed.

[0110] S46, error calculation. The difference between the predicted value of the current model and the true value is measured based on formula (6).

[0111] S47, termination condition judgment. When the maximum iteration condition is reached or the error converges to the preset value, the algorithm is stopped and the reverse prediction model is output. When the termination condition is not met, the network weights and biases are updated and then returned to S45. The error back propagation algorithm of the BP network updates the weights and biases in each round of iteration, and its formula is shown in formula (12).

[0112] S5. Model evaluation. Three key evaluation indicators were introduced: coefficient of determination (R²), mean absolute error (MAE), and mean bias error (MBE). The results of the ten-fold cross validation showed that the model had a high R², and small MAE and MBE, proving the accuracy and reliability of the model results. The evaluation results are shown in the following table:

[0113] R2 MAE MBE First Fold 0.9897 0.4459 -0.3946 Second Fold 0.9796 0.2639 -0.2987 …… …… …… …… 10th Fold 0.9981 0.8231 0.5689 average 0.9912 0.3549 0.0035

[0114] S6. Reverse prediction. The constructed GWO-DE-BP neural network is applied to the actual niobium alloy laser welding process parameter prediction process. The tensile strength, high temperature oxidation resistance and welding depth requirements of the weld are input into the constructed GWO-DE-BP neural network, and the trained model is used to predict the parameters of laser power, welding speed and focal distance. Five groups of prediction values ​​were selected in the embodiments of this specification and experimentally verified. The errors were all within 5%, indicating that the constructed performance-driven niobium alloy laser welding process reverse prediction method is reliable. At the same time, software for this method is also designed based on reverse prediction.

[0115] The technical solution provided in the embodiments of this specification aims to solve the problem of difficult performance and parameter matching during niobium alloy laser welding. A back propagation (BP) neural network model based on grey wolf optimization (GWO) and differential evolution (DE) algorithm optimization is constructed. The model can reversely predict the optimal welding process parameters according to the required welding performance, thereby reducing unnecessary repeated experiments and significantly improving welding quality and efficiency; using a large amount of niobium alloy laser welding data, the welding parameters can be accurately predicted through the prediction model according to performance requirements, thereby improving the stability of the welding process parameters; the data set integrates common welding performance and matching welding process parameters, and the constructed prediction model can be adjusted according to different welding performance requirements, with strong flexibility and adaptability, reducing the cost and time waste caused by the traditional trial and error method; the reverse prediction method can extract key welding process parameters according to the welding performance, and realizes the intelligent prediction of welding parameters by integrating the GWO-DE-BP neural network model, promoting the automation and intelligentization of the welding process.

[0116] The present specification also provides a reverse prediction device for niobium alloy laser welding process, such as Figure 5As shown, the device includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned inverse prediction method of the niobium alloy laser welding process.

[0117] The embodiment of the present specification also provides a non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions are configured to: execute the above-mentioned reverse prediction method for the niobium alloy laser welding process.

[0118] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment, and non-volatile computer storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0119] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0120] The devices and media provided in the embodiments of this specification correspond one-to-one to the methods. Therefore, the devices and media also have similar beneficial technical effects as the corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0121] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, this specification may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, this specification may 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.

[0122] This specification is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0123] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0124] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0125] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0126] Memory may include non-permanent storage in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0127] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0128] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0129] The above description is only one or more embodiments of this specification and is not intended to limit this specification. For those skilled in the art, one or more embodiments of this specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of one or more embodiments of this specification shall be included in the scope of the claims of this specification.

Claims

1. A reverse prediction method for niobium alloy laser welding process, characterized in that: The method comprises: Obtaining welding process performance requirement information corresponding to niobium alloy laser welding, wherein the welding process performance requirement information includes any one or more of a high temperature oxidation resistance index, a welding depth index, a welding width-to-depth ratio index, a heat-affected zone width index, a tensile strength index, and a shear strength index; According to the welding process performance requirement information, a pre-constructed laser welding process parameter inverse prediction model is used to determine the welding process parameters corresponding to the niobium alloy laser welding, wherein the laser welding process parameter inverse prediction model is a back propagation neural network model based on grey wolf optimization and differential evolution algorithm optimization.

2. The reverse prediction method of niobium alloy laser welding process according to claim 1 is characterized in that: According to the welding process performance requirement information, before determining the welding process parameters corresponding to the niobium alloy laser welding using a pre-built laser welding process parameter reverse prediction model, the method further includes: Acquire an initial niobium alloy laser welding data set, wherein the initial niobium alloy laser welding data set includes welding process parameters of niobium alloys of multiple thicknesses and corresponding welding performance indicators; Performing data feature analysis on the initial niobium alloy laser welding data set to determine at least one main welding performance index and at least one key welding process parameter to construct a laser welding process data set; Based on the back propagation neural network method optimized by gray wolf optimization and differential evolution algorithm, a reverse mapping model between welding performance index and welding process parameter is constructed, wherein the input of the reverse mapping model is the at least one main welding performance index, and the output is the at least one key welding process parameter; The reverse mapping model is trained using the laser welding process data set to determine the laser welding process parameter reverse prediction model.

3. The reverse prediction method of niobium alloy laser welding process according to claim 2 is characterized in that: Performing data feature analysis on the initial niobium alloy laser welding data set to determine at least one main welding performance index and at least one key welding process parameter, specifically including: Acquiring actual application environment parameters of the niobium alloy laser welded formed part, and determining at least one main welding performance index corresponding to the niobium alloy laser welded formed part based on the actual application environment parameters; At least one key welding process parameter in the welding process is identified based on the at least one main welding performance indicator using a fuzzy-decision laboratory analysis method.

4. The reverse prediction method for niobium alloy laser welding process according to claim 3 is characterized in that: Using the fuzzy-decision laboratory analysis method, at least one key welding process parameter in the welding process is identified according to the at least one main welding performance indicator, specifically including: Using fuzzy triangular numbers, the influences between multiple welding process parameters are scored, multiple expert scoring data are obtained, and based on each of the expert scoring data, a corresponding fuzzy influence matrix is ​​constructed; Aggregating the plurality of fuzzy influence matrices by weighted average to obtain a comprehensive fuzzy matrix, defuzzifying the comprehensive fuzzy matrix by using a centroid method, and performing normalization processing on the comprehensive fuzzy matrix to determine a normalized matrix; According to the normalized matrix, a total influence matrix between the plurality of welding process parameters is calculated, so as to respectively calculate the influence degree between each of the welding process parameters and the influence degree between each of the welding process parameters based on the total influence matrix; The plurality of welding process parameters are sorted according to the influence between each of the welding process parameters and the influence between each of the welding process parameters to determine at least one key welding process parameter.

5. The reverse prediction method of niobium alloy laser welding process according to claim 4, characterized in that: According to the normalized matrix, a total influence matrix between the plurality of welding process parameters is calculated, specifically comprising: According to the normalized matrix, according to the formula , calculate the total influence matrix between the multiple welding process parameters, where T is the total influence matrix and I is the unit matrix.

6. The reverse prediction method for niobium alloy laser welding process according to claim 4, characterized in that: Sorting the plurality of welding process parameters according to the influence between each welding process parameter and the influence between each welding process parameter to determine at least one key welding process parameter, specifically including: Obtaining an importance coefficient of each welding process parameter according to the influence degree between each welding process parameter and the sum of the influence degree between each welding process parameter; The plurality of welding process parameters are sorted in order from large to small according to the importance coefficient of each welding process parameter, so as to determine at least one key welding process parameter among the plurality of welding process parameters.

7. The reverse prediction method for niobium alloy laser welding process according to claim 2, characterized in that: Based on the back propagation neural network method optimized by grey wolf optimization and differential evolution algorithm, a reverse mapping model between welding performance indicators and welding process parameters is constructed, including: Set the parameters of the gray wolf optimization algorithm and the differential evolution algorithm, and represent each individual as the weight and bias of the neural network; The fitness is calculated through the error function of the BP neural network to determine the fitness value of each individual, where the fitness value of each individual is used to represent the error of each weight and bias combination on the training set; Perform global search through the encirclement and hunting behavior of the gray wolf algorithm, use three gray wolves to guide the search, and update the position of the gray wolves; When performing mutation operation, three individuals are randomly selected, and the difference vector between them is calculated to realize the mutation of individuals. New individuals are generated by crossover operation, and the dominant individuals are selected according to the fitness value of new and old individuals. The BP neural network is trained using the optimized weights and biases to construct an inverse prediction model between performance indicators and input parameters.

8. The reverse prediction method for niobium alloy laser welding process according to claim 7, characterized in that: After the reverse mapping model is trained by the laser welding process data set to determine the reverse prediction model of the laser welding process parameters, the method further includes: Calculate at least one model evaluation index according to the true value and the predicted value corresponding to the individual with the smallest fitness value, so as to evaluate the laser welding process parameter inverse prediction model based on the at least one model evaluation index, wherein the model evaluation index includes any one of a model determination coefficient, a model mean absolute error, and a model mean deviation error; in, , R² is the coefficient of determination of the model, and N is the number of samples; It is the true value corresponding to the individual with the smallest fitness value; is the predicted value corresponding to the individual with the smallest fitness value, , MAE is the mean absolute error of the model, , MBE is the model mean bias error.

9. A reverse prediction device for niobium alloy laser welding process, characterized in that: The device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the inverse prediction method for the niobium alloy laser welding process according to any one of claims 1-8.

10. A non-volatile computer storage medium storing computer executable instructions, characterized in that: The computer executable instructions are configured to execute the reverse prediction method for the niobium alloy laser welding process according to any one of claims 1 to 8.

Images