Electric arc additive manufacturing process parameter optimization design method combining magnetic field and ultrasonic field

Abstract

The invention relates to an electric arc additive manufacturing process parameter optimization design method combining a magnetic field and an ultrasonic field, and belongs to the technical field of electric arc additive manufacturing. The problems of molten pool stability, microstructure control, mechanical property optimization and the like in the electric arc additive manufacturing process are effectively solved. Comprising the following steps: step 1, acquiring process parameters and external field data; step 2, carrying out data normalization processing; 3, machine learning and multi-objective optimization are carried out; and step 4, knowledge base construction and intelligent optimization. According to the method, the dynamic behavior of the molten pool is accurately regulated and controlled by optimizing the matching of process parameters and external field load parameters, so that the deposition quality, the microstructure and the mechanical property of the material are effectively improved.

Description

Technical Field

[0001] The present invention relates to a method for optimizing process parameters by combining a magnetic field and an ultrasonic field, and belongs to the technical field of wire arc additive manufacturing. Background Art

[0002] Wire Arc Additive Manufacturing (WAAM), as an emerging 3D printing technology, has significant advantages such as high material utilization rate, high forming efficiency, and low cost. It has been widely used in the rapid manufacturing of metal parts in the aerospace, shipbuilding, and automotive industries. However, the WAAM process faces many technical challenges, including unstable molten pools, crack and porosity defects caused by excessive heat input, poor interlayer metallurgical bonding, and non-uniform internal microstructure of materials. These problems seriously affect the mechanical properties of parts, especially in terms of strength, ductility, and fatigue performance.

[0003] To address the above problems, existing research has attempted to improve the manufacturing quality by optimizing process parameters (such as arc current, voltage, etc.) or introducing external physical fields (such as magnetic fields, ultrasonic fields). However, the existing technologies have the following deficiencies:

[0004] 1. Limitations of single physical field regulation: For example, the invention with the publication number CN117380975A uses a constant magnetic field to assist laser deposition to improve the metallurgical structure, but its dynamic response ability is weak, and it is difficult to achieve real-time regulation of complex process processes. Although ultrasonic field technology can optimize the molten pool flow by using cavitation effects and acoustic streaming effects, its adaptability to different process parameters is insufficient, and it is difficult to meet the morphological requirements of complex parts.

[0005] 2. Limitations of optimization methods: Traditional WAAM optimization methods mainly rely on empirical adjustment or offline simulation, lacking real-time monitoring and feedback of the actual manufacturing process. For example, the simulation method in "Research Progress on Wire Arc Additive Manufacturing Process and Numerical Simulation" can only provide static optimization, and its ability to control the dynamic evolution of the microstructure is limited.

[0006] 3. Insufficient research on composite physical fields: Existing research has attempted to combine magnetic fields and ultrasonic fields, but mainly uses fixed parameter settings and cannot accurately regulate the dynamic behavior of the molten pool. For example: The single constant magnetic field optimization process (CN118287831A) can reduce hump weld beads and improve the molten pool morphology, but lacks dynamic response ability. The alternating magnetic field + oscillating laser technology (CN118287830A) mainly optimizes arc and droplet behaviors, but still has deficiencies in the forming consistency at the head and tail ends and the control of the microstructure. The ultrasonic forging + magnetic field-assisted synchronization technology (CN118543932A) enhances the ultrasonic effect, but lacks complex dynamic regulation ability, affecting the uniformity of the microstructure.

[0007] Based on the above analysis, there is an urgent need to develop an optimization method for the arc additive manufacturing process that combines magnetic fields and ultrasonic fields. By optimizing process parameters and external fields through intelligent algorithms, real-time control of the molten pool can be achieved to improve the manufacturing quality of WAAM and the mechanical properties of parts. Summary of the Invention

[0008] In order to effectively solve problems such as molten pool stability, microstructure control, and mechanical property optimization in the process of arc additive manufacturing, an optimization design method for the process parameters of arc additive manufacturing that combines magnetic fields and ultrasonic fields is provided. A brief overview of the present invention is given below to provide a basic understanding of certain aspects of the present invention. It should be understood that this overview is not an exhaustive overview of the present invention. It is not intended to identify the key or important parts of the present invention, nor is it intended to limit the scope of the present invention.

[0009] Technical solution of the present invention:

[0010] An optimization design method for the process parameters of arc additive manufacturing that combines magnetic fields and ultrasonic fields includes the following steps:

[0011] Step 1: Acquisition of process parameter and external field data;

[0012] Step 2: Data normalization processing;

[0013] Step 3: Machine learning and multi-objective optimization;

[0014] Step 4: Knowledge base construction and intelligent optimization.

[0015] Preferably: Step 1 includes the following steps:

[0016] Step 1.1: Acquisition of the geometric model data of the target part;

[0017] Step 1.2: Determination of the material type and acquisition of physical properties;

[0018] Step 1.3: Collection of process parameters;

[0019] Step 1.4: Input of external field load parameters;

[0020] Step 1.5: Acquisition of mechanical property data.

[0021] Preferably: In Step 1.1, the three-dimensional geometric model data of the target part is acquired, and the part is divided into several sub-regions according to this data;

[0022] In Step 1.2, according to the material type of the target part, the physical properties of the material are acquired and defined as the set W = {w1, w2, w3,..., w n}, physical properties include but are not limited to specific heat capacity, coefficient of thermal expansion, density, thermal conductivity, yield strength, Young's modulus;

[0023] In step 1.3, collect the process parameters during the arc additive manufacturing process, defined as the set X = {x1, x2, x3…, x n}, process parameters include welding current, voltage, welding speed, arc waveform, arc frequency, arc polarity, preheating temperature, interlayer temperature, interlayer waiting time;

[0024] In step 1.4, input the external field load parameters, defined as the set Y = {y1, y2, y3,…, y m} including magnetic field strength, magnetic field direction, ultrasonic field frequency and ultrasonic field strength;

[0025] In step 1.5, obtain the mechanical property data of the material, defined as the set Z = {z1, z2, z3,…, z p}, including tensile strength, elongation, fatigue performance, impact toughness and high temperature strength characteristics.

[0026] Preferably: In step 2, use the random forest algorithm to screen the key process parameters and external field parameters.

[0027] Preferably: Step 2 includes the following steps:

[0028] Step 2.1, data normalization processing:

[0029] Perform normalization processing on all parameter data, and use normalization technology to standardize the data to ensure data consistency;

[0030]

[0031] Step 2.2, feature importance calculation:

[0032] Through feature importance calculation, screen out the key parameters highly related to the target mechanical properties; the feature importance calculation formula is as follows:

[0033]

[0034] Where, Δgini(f i , n): the reduction in Gini impurity caused by feature f i in the nth tree; T: the total number of trees in the random forest;

[0035] Definition of Gini impurity: For the sample data set in the node, its Gini impurity calculation formula is:

[0036]

[0037] pk : The proportion of samples belonging to the k-th category in the node; K: The total number of categories;

[0038] After data preprocessing, LASSO regression and principal component analysis are used for feature selection and dimensionality reduction of process parameters to identify the key parameters that have the greatest impact on the molten pool quality and final material properties, avoiding interference from redundant information; its objective function is:

[0039]

[0040] Among them, n is the number of samples; y i is the actual mechanical property value corresponding to the i-th sample (such as the tensile strength, elongation rate measured by experiments, etc.); X i is the feature vector of the process parameters and external field parameters corresponding to the i-th sample; β j is the regression coefficient; p is the number of coefficients; λ is the regularization penalty parameter;

[0041] For dimensionality reduction processing, principal component analysis is used to reduce redundant information and extract the main process parameters and external field parameters that have the greatest impact on mechanical properties;

[0042] Principal component analysis:

[0043] PCA extracts the main components affecting performance by decomposing the covariance matrix of the parameters; the formula is as follows:

[0044] PC = W T X

[0045] Among them, W is the feature matrix, X is the original data matrix, and PC is the data after dimensionality reduction.

[0046] Preferably: Step 3 includes the following steps:

[0047] Step 3.1: Multi-objective optimization;

[0048] Step 3.2: Dynamic adjustment and real-time monitoring.

[0049] Preferably: In Step 3.1, machine learning algorithms are used to perform multi-objective optimization on the selected key process parameters and external field parameters to simultaneously meet multiple mechanical property requirements (such as tensile strength and elongation rate); feature selection is achieved through random forest algorithms, LASSO regression, and principal component analysis methods to ensure that the optimization process focuses on key parameters; genetic algorithms or particle swarm optimization algorithms are used to complete global optimization and output the optimal parameter combination that meets multi-objective performance requirements;

[0050] In Step 3.2, during the optimization process, these parameters are applied to the arc additive manufacturing process, and the magnetic field strength, direction of the external magnetic field, and the frequency and amplitude of the ultrasonic field are dynamically adjusted in combination with real-time monitoring data; through a real-time feedback mechanism, the parameters are continuously corrected to maintain the stability of the manufacturing process, and at the same time, the optimization accuracy is further improved through grid search or Bayesian optimization to ensure that the final performance meets the requirements of complex processes.

[0051] Preferably: in Step 3.1, based on the feature selection results, the relationship between process parameters and mechanical properties is established through a multi-output support vector regression (SVR) model;

[0052] SVR objective function:

[0053]

[0054] where w is the weight vector of SVR; ∈ i is the deviation of each sample point; C is the penalty parameter;

[0055] Nonlinear mapping is performed through the RBF kernel function:

[0056]

[0057] where K(x, x′) is the radial basis function kernel; x, x′ are input data points; σ is the parameter of the kernel function;

[0058] Each sample x is a combination of process parameters, for example, x = [x1, x2, x3,..., xn] (such as current, voltage, welding speed, etc.);

[0059] x′ is a combination of process parameters of another sample x′ = [x1′, x2′, x3′,..., xn′];

[0060] The similarity between two samples in the process parameter space is measured by calculating the Euclidean distance ||x - x′||2 of the two process parameter vectors x and x′;

[0061] If the process parameter vectors of two samples are close, the value of the kernel function K(x, x′) is close to 1; if they are far apart, the value of the kernel function approaches 0;

[0062] The genetic algorithm is used to optimize the process parameters and the external field load parameters; the objective function is:

[0063] F(x) = α1·S pool + α2·O micro + α3·P mech

[0064] where S pool represents the stability of the molten pool, O microIndicates the microstructure uniformity, P mech is a mechanical property index, and α1, α2, α3 are weight coefficients;

[0065] Output the optimal process parameter combination X opt , Y opt ;

[0066] In Step 3.2, dynamically adjust the parameters: Apply the optimized parameters to the arc additive manufacturing process, and dynamically adjust the parameters of the magnetic field and ultrasonic field in combination with the real-time monitoring results; Through real-time feedback, ensure the stability of the manufacturing process and the mechanical properties of the parts reach the design goals;

[0067] The influence of the magnetic field on the molten pool flow: The influence of the magnetic field on the molten pool flow can be expressed by the following formula:

[0068]

[0069] where F is the Lorentz force; q is the electric charge; B is the magnetic field strength; ρ is the density of the fluid;

[0070] The influence of ultrasonic waves on the molten pool: The temperature change of the molten pool caused by ultrasonic waves can be expressed as:

[0071]

[0072] In the formula, ΔT is the temperature change; α is the heating coefficient of ultrasonic waves on the metal; P is the power of ultrasonic waves; λ is the thermal conductivity of the metal;

[0073] Real-time feedback regulation formula:

[0074] y adjusted = y initial + Δy

[0075] where y adjusted is the adjusted parameter value, y initial is the initial set value, and Δy is the amount adjusted according to the feedback signal;

[0076] Magnetic field regulation formula:

[0077] B adjusted = B initial + ΔB

[0078] where B adjusted is the adjusted magnetic field strength, B initial is the initial magnetic field strength, and ΔB is the adjustment amount, which is adjusted according to the real-time monitoring and optimization calculation results;

[0079] Ultrasonic field regulation formula:

[0080] P adjusted = P initial+ΔP

[0081] where P adjusted is the adjusted ultrasonic field intensity, P initial is the initial ultrasonic field intensity, and ΔP is the adjustment amount, which is adjusted according to the optimization result;

[0082] According to the real-time monitoring results, the process parameters are dynamically adjusted to keep the process always in the optimized state;

[0083] The dynamic feedback and parameter adjustment include:

[0084] (1) Define the real-time monitoring error: During the additive manufacturing process, the real-time monitoring system continuously provides data on the current state;

[0085] Assume:

[0086] y target is the target process parameter (e.g., target temperature, target molten pool depth, etc.);

[0087] y measured is the process parameter measured in real time by the sensor;

[0088] Then, the error Δy can be calculated as:

[0089] Δy = y measured - y target

[0090] Δy represents the error between the current process parameter and the target parameter;

[0091] (2) Calculate the adjustment amount based on the error: The control strategies include proportional-integral-derivative (PID) control and adaptive control; assume a simple proportional control method is used, then the adjustment amount Δy adjusted is:

[0092] Δy adjusted = K p ·Δy

[0093] where K p is the proportionality coefficient, which is used to adjust the intensity of the control;

[0094] (3) Control strategy: The form of PID control is:

[0095]

[0096] where K p is the proportionality coefficient; K i is the integral coefficient; K d is the derivative coefficient.

[0097] Preferably, Step 4 includes the following steps:

[0098] Step 4.1: Define key concepts and construct a semantic association network;

[0099] Step 4.2: Semantic reasoning and parameter matching;

[0100] Step 4.3: Prediction and optimization of mechanical properties;

[0101] Step 4.4: Verification and optimization iteration.

[0102] Preferably, in Step 4.1, define the key concepts of arc additive manufacturing, including process parameters, material properties, external field parameters, and mechanical properties; construct a semantic association network and establish a knowledge base through ontological methods and a professional term dictionary;

[0103] In Step 4.2, match relevant parameter combinations in the knowledge base, extract recommended process parameters and external field load parameters through a semantic reasoning engine, and optimize them in combination with user requirements to generate initial process parameters and external field parameters that best meet the current manufacturing requirements;

[0104] In Step 4.3, for the optimization results of process parameters and external field parameters, use multi-output support vector regression for mechanical property prediction; construct a regression model through a kernel function to predict multiple mechanical property indicators; use grid search or Bayesian optimization to tune parameters to improve the prediction accuracy of the model;

[0105] In Step 4.4, verify the manufacturing results. If there are deviations, re-optimize and adjust until the design requirements are met; if the performance does not meet the design requirements, re-adjust the optimization model or experimental plan according to the deviation, update the data set, and repeat the optimization process until the performance requirements are met; combine dynamic adjustment and real-time monitoring to further improve the stability and consistency of the manufacturing process; after completing the manufacturing process, comprehensively verify the results to ensure that the performance meets the requirements; store the effective parameter combinations in the knowledge base as a reference for subsequent optimization to form a closed-loop feedback.

[0106] The present invention has the following beneficial effects:

[0107] 1. The present invention improves the stability of the molten pool: Through the synergistic effect of the magnetic field and the ultrasonic field, the flow behavior in the molten pool is optimized, effectively reducing defects such as cracks and pores, and improving the deposition quality and forming accuracy.

[0108] 2. The present invention improves the uniformity of the microstructure: The action of the magnetic field and the ultrasonic field helps to refine grains, improve the distribution of the microstructure of the material, reduce internal segregation, and improve the comprehensive mechanical properties of the material.

[0109] 3. The present invention improves mechanical properties: By jointly optimizing process parameters and external fields, the tensile strength, ductility, fatigue performance and other mechanical indexes of materials can be significantly improved, and the service life and working reliability of parts can be extended.

[0110] 4. The present invention is suitable for complex geometric shapes: The method of the present invention can effectively meet the additive manufacturing requirements of complex parts, especially showing excellent performance in the microstructure control and performance optimization of complex structures.

[0111] 5. The present invention has intelligent optimization design: Using machine learning algorithms and multi-objective optimization strategies, key parameters are automatically screened and globally optimized, reducing manual intervention and improving design efficiency and optimization accuracy.

[0112] 6. Through these innovative technical measures, the present invention can effectively solve the defects in the prior art, provide a more accurate and efficient method for optimizing the arc additive manufacturing process, greatly improve the material properties, deposition quality and part accuracy in the additive manufacturing process, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0113] Figure 1 is the general flow chart of the method for optimizing the arc additive manufacturing process parameters by combining magnetic field and ultrasonic field according to the present invention;

[0114] Figure 2 is the flow chart of data collection and preprocessing process according to the present invention;

[0115] Figure 3 is the schematic diagram of the feature selection process according to the present invention;

[0116] Figure 4 is the schematic diagram of the optimization process according to the present invention;

[0117] Figure 5 is the schematic diagram of real-time feedback and dynamic adjustment according to the present invention;

[0118] Figure 6 is the schematic diagram of knowledge base construction and update according to the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0119] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be described below through specific embodiments shown in the drawings. However, it should be understood that these descriptions are exemplary and are not intended to limit the scope of the present invention. In addition, in the following description, the descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present invention.

[0120] DETAILED DESCRIPTION OF THE INVENTION I: Combining Figure 1-2 To illustrate this embodiment, the method for optimizing the arc additive manufacturing process parameters by combining magnetic field and ultrasonic field in this embodiment includes the following steps:

[0121] Step 1: Obtain process parameters and external field data;

[0122] Step 2: Perform data normalization;

[0123] Step 3: Machine learning and multi-objective optimization;

[0124] Step 4: Knowledge base construction and intelligent optimization;

[0125] Therefore, on the basis of traditional methods, the present invention combines the synergistic effects of magnetic fields and ultrasonic fields to achieve precise control of the molten pool dynamic behavior during the WAAM process; the initial process parameters are optimized through machine learning algorithms (such as random forest, LASSO regression, etc.), and a multi-output support vector regression model (SVR) is constructed to systematically correlate process parameters with material properties, realizing real-time online multi-modal monitoring and dynamic adjustment of external field parameters, thereby comprehensively improving the manufacturing quality, efficiency, and intelligent level; the present invention realizes precise control of the molten pool dynamic behavior by optimizing the coordination of process parameters and external field load parameters, thereby effectively improving the deposition quality, microstructure, and mechanical properties of materials, aiming to improve the deposition quality, forming accuracy, and mechanical properties of materials during the arc additive manufacturing process through external field control and process parameter optimization to meet the needs of high-end industrial applications.

[0126] Specific Embodiment 2: Combination Figure 1-2 Describing this embodiment, the method for optimizing the process parameters of arc additive manufacturing by combining magnetic fields and ultrasonic fields in this embodiment, Step 1 includes the following steps:

[0127] Step 1.1: Obtain the geometric model data of the target part;

[0128] Step 1.2: Determine the material type and obtain physical properties;

[0129] Step 1.3: Collect process parameters;

[0130] Step 1.4: Input external field load parameters;

[0131] Step 1.5: Obtain mechanical property data.

[0132] Specific Embodiment 3: Combination Figure 1-2 Describing this embodiment, in Step 1.1 of the method for optimizing the process parameters of arc additive manufacturing by combining magnetic fields and ultrasonic fields in this embodiment, the three-dimensional geometric model data of the target part can be obtained by using CAD, and the part can be divided into several sub-regions according to this data; this helps with subsequent local process optimization for each sub-region;

[0133] In Step 1.2, according to the material type of the target part, obtain the physical properties of the material. Import the three-dimensional geometric model data into ANSYS Workbench to define the material physical properties, defined as the set W = {w1, w2, w3, …, w n}, and the physical properties include but are not limited to specific heat capacity, coefficient of thermal expansion, density, thermal conductivity, yield strength, Young's modulus, etc.;

[0134] In Step 1.3, collect the process parameters during the arc additive manufacturing process. Embed FORTRAN into ANSYS Workbench, defined as the set X = {x1, x2, x3, …, x n}, and the process parameters include welding current, voltage, welding speed, arc waveform, arc frequency, arc polarity, preheating temperature, interlayer temperature, interlayer waiting time, etc.;

[0135] In Step 1.4, input the external field load parameters in ANSYS Workbench, defined as the set Y = {y1, y2, y3, …, y m}, including magnetic field strength, magnetic field direction, ultrasonic field frequency, and ultrasonic field strength, etc.;

[0136] In Step 1.5, obtain the mechanical property data of the material through ABAQUS, defined as the set Z = {z1, z2, z3, …, z p}, including properties such as tensile strength, elongation, fatigue performance, impact toughness, and high-temperature strength. ANSYS Workbench and ABAQUS support data exchange.

[0137] Specific Embodiment 4: Combine Figure 1-2 To illustrate this embodiment, for the optimization design method of the arc additive manufacturing process parameters combining magnetic field and ultrasonic field in this embodiment, in Step 2, Python can be used to normalize all the obtained parameter data, remove invalid data and fill in missing values to ensure the integrity and high quality of the data, providing a good foundation for the subsequent model establishment. Use the random forest algorithm to screen the key process parameters and external field parameters; by calculating the importance of features, select the parameters highly correlated with the target mechanical properties to ensure the accuracy and feasibility of the optimization goal.

[0138] Specific Embodiment 5: Combine Figure 1-2 To illustrate this embodiment, for the optimization design method of the arc additive manufacturing process parameters combining magnetic field and ultrasonic field in this embodiment, Step 2 includes the following steps:

[0139] Step 2.1: Data normalization processing:

[0140] Normalize all the obtained parameter data, remove invalid data and complete missing values to ensure data integrity and high quality, providing a good foundation for the subsequent establishment of the model;

[0141] Normalize all parameter data,

[0142]

[0143] Remove invalid data and complete missing values to provide high-quality data for the subsequent model establishment;

[0144] Step 2.2, Feature importance calculation:

[0145] Use the random forest algorithm to screen key process parameters and external field parameters; by calculating the importance of features, select parameters highly correlated with the target mechanical properties to ensure the accuracy and feasibility of the optimization goal;

[0146] Through feature importance calculation, screen out key parameters highly correlated with the target mechanical properties (such as tensile strength, fatigue performance, etc.); the formula for feature importance is as follows:

[0147]

[0148] where, Δgini(f i , n): the reduction in Gini impurity caused by feature f i in the nth tree; T: the total number of trees in the random forest;

[0149] Definition of Gini impurity: For the sample data set in a node, its Gini impurity calculation formula is:

[0150]

[0151] p k : the proportion of samples belonging to the kth class in the node; K: the total number of classes;

[0152] After data preprocessing, use LASSO regression and principal component analysis (PCA) for feature selection and dimensionality reduction of process parameters, and identify the key parameters that have the greatest impact on the molten pool quality and final material properties; avoid interference from redundant information; its objective function is:

[0153]

[0154] β j is the regression coefficient in the regression model, which represents the influence degree of feature j (such as welding current, voltage, magnetic field strength, etc.) on the target variable (such as mechanical properties: tensile strength, elongation, etc.);

[0155] where, The estimated value of the regression coefficient is the result obtained by optimizing the objective function; n is the number of samples; y i is the actual mechanical property value corresponding to the i-th sample (such as the tensile strength, elongation rate, etc. measured in experiments); X i is the eigenvector of the process parameters and external field parameters corresponding to the i-th sample (such as parameters like welding current, voltage, magnetic field strength, etc.); β j is the regression coefficient; p is the number of coefficients, that is, the total number of features included in the model; λ is the regularization penalty parameter, which is used to control the sparsity of the coefficients and prevent the model from overfitting;

[0156] LASSO regression can select the features most relevant to the target mechanical properties from all input features, thereby reducing the computational complexity and improving the accuracy of the model;

[0157] Principal component analysis (PCA) performs eigen-decomposition on the covariance matrix of the data to find the most principal components, reduce the dimension of the feature space, thereby reducing redundant information and noise, and improving the accuracy and computational efficiency of the model;

[0158] For dimensionality reduction processing, principal component analysis (PCA) is used to reduce redundant information and extract the main process parameters and external field parameters that have the greatest impact on mechanical properties;

[0159] Principal component analysis (PCA):

[0160] PCA decomposes the covariance matrix of the parameters to extract the main components affecting the performance; the formula is as follows:

[0161] PC = W T X

[0162] where W is the feature matrix, X is the original data matrix, and PC is the data after dimensionality reduction.

[0163] Specific implementation method six: Combine Figure 1-2 To illustrate this implementation method, the method for optimizing the process parameters of arc additive manufacturing combining magnetic field and ultrasonic field in this implementation method, step 3 includes the following steps:

[0164] Step 3.1: Multi-objective optimization;

[0165] Step 3.2: Dynamic adjustment and real-time monitoring;

[0166] In the existing alternating magnetic field and oscillating laser technologies, the present invention has also achieved breakthroughs; the present invention breaks through the limitations of single swing control methods by introducing machine learning-driven dynamic regulation and multi-physical field coupling; the synergistic effect of the magnetic field and ultrasonic field enhances the stirring effect of the molten pool and significantly reduces the porosity and texture anisotropy problems, realizing higher-performance aluminum alloy additive components;

[0167] By introducing multi - physical - field regulation and intelligent optimization technology, the present invention breaks through the limitations of traditional experience - dependent adjustments; through real - time regulation of the synergistic effect of the magnetic field and the ultrasonic field, combined with the residual stress distribution and molten pool behavior feedback system constructed by the machine - learning model, it realizes precise control of process stability and material properties, significantly improving the forming quality and production efficiency;

[0168] Compared with the traditional WAAM optimization method mainly based on numerical simulation research, the present invention combines real - time sensing technology and intelligent algorithm optimization, breaking through the bottleneck of the low calculation efficiency of complex numerical models and the weak regulation ability for the actual manufacturing process. By real - time monitoring the molten pool behavior and dynamically adjusting the magnetic field and ultrasonic field parameters, the present invention significantly improves the efficiency and quality control ability of the manufacturing process;

[0169] The present invention significantly improves the molten pool stability and manufacturing quality through the combined action of multi - physical fields; through the cavitation effect and acoustic streaming effect of the ultrasonic field, which act synergistically on the liquid metal at the edge of the molten pool, it effectively reduces defects while optimizing the grain orientation, thus realizing the adaptive manufacturing of different materials and geometrically complex parts;

[0170] The present invention extends the action of the magnetic field and ultrasonic field to a more complex dynamic regulation field. By real - time optimizing parameters through the machine - learning model, the present invention has more advantages in improving the microstructure uniformity and eliminating additive manufacturing defects. In addition, the use of intelligent algorithms effectively avoids the problem of unstable deposition morphology caused by improper setting of magnetic field and ultrasonic field parameters, thus significantly improving the manufacturing efficiency and the comprehensive performance of parts;

[0171] The present invention also aims at the deficiencies of the existing magnetic field parameter optimization methods in laser additive manufacturing. Based on the characteristics of arc additive manufacturing, it introduces unique molten pool behavior regulation and multi - physical field optimization; traditional laser additive manufacturing methods mostly rely on the partition optimization of finite - element simulation, while the present invention focuses on solving the synergistic problem of grain orientation optimization and mechanical property improvement through real - time monitoring and dynamic adjustment, thus overcoming the defects of hysteresis and insufficient adaptability in traditional magnetic field optimization methods, and further improving the closed - loop control efficiency and performance customization ability of the manufacturing process;

[0172] The advantages of the present invention are particularly obvious. First of all, by introducing the synergistic regulation of the magnetic field and ultrasonic field, the present invention significantly improves the control effect on the dynamic behavior of the molten pool during the additive manufacturing process. The electromagnetic - assisted laser technology mainly uses the magnetic field to refine the microstructure of the molten pool, but its forming area is limited and it is difficult to dynamically adjust. The present invention combines machine - learning to optimize process parameters, real - time adjusts the intensity and frequency of the magnetic field and ultrasonic field, enhances the fluidity and tissue uniformity of the molten pool, and solves the limitations of traditional methods in the manufacturing of complex - shaped components.

[0173] Specific Embodiment Seven: CombineFigure 1-2 Regarding this embodiment, in the arc additive manufacturing process parameter optimization design method combining a magnetic field and an ultrasonic field of this embodiment, in step 3.1, a machine learning algorithm is used to perform multi-objective optimization on the selected key process parameters and external field parameters to simultaneously meet multiple mechanical property requirements (such as tensile strength and elongation); feature selection is achieved through methods such as the random forest algorithm, LASSO regression, and principal component analysis to ensure that the optimization process focuses on the most critical parameters; subsequently, a genetic algorithm or a particle swarm optimization algorithm is used to complete the global optimization and output the optimal parameter combination that meets the multi-objective performance requirements.

[0174] In step 3.2, during the optimization process, these parameters are applied to the arc additive manufacturing process, and the magnetic field strength, direction of the external magnetic field, and the ultrasonic field frequency and amplitude are dynamically adjusted in combination with real-time monitoring data (such as molten pool stability and microstructure uniformity); through a real-time feedback mechanism, the parameters are continuously corrected to maintain the stability of the manufacturing process, and at the same time, the optimization accuracy is further improved through grid search or Bayesian optimization to ensure that the final performance meets the complex process requirements.

[0175] Specific Embodiment VIII: Combining Figure 1-2 Regarding this embodiment, in the arc additive manufacturing process parameter optimization design method combining a magnetic field and an ultrasonic field of this embodiment, in step 3.1, a machine learning optimization algorithm is used to perform multi-objective optimization on the process parameters and external field parameters; the optimal combination of process parameters and external field load parameters is output through the optimization algorithm to maximize the mechanical properties.

[0176] Based on the feature selection results, the relationship between the process parameters and the mechanical properties is established through a multi-output support vector regression (SVR) model; SVR performs non-linear mapping through a kernel function, enabling accurate prediction of material properties even in a complex additive manufacturing environment; by training the SVR model, the mechanical properties under different process parameter combinations, such as strength, ductility, and fatigue properties, can be predicted.

[0177] SVR objective function:

[0178]

[0179] where w is the weight vector of SVR; ∈ i is the deviation of each sample point; C is the penalty parameter used to control the model complexity and prevent overfitting.

[0180] Non-linear mapping is performed through the RBF kernel function:

[0181]

[0182] Among them, K(x, x′) is a radial basis function (RBF) kernel; x and x′ are input data points; σ is a parameter of the kernel function, which controls the similarity between data points;

[0183] Each sample x is a combination of process parameters. For example, x = [x1, x2, x3, …, xn] (such as current, voltage, welding speed, etc.);

[0184] x′ is a combination of process parameters of another sample, x′ = [x1′, x2′, x3′, …, xn′];

[0185] By calculating the Euclidean distance ||x - x′||2 between two process parameter vectors x and x′, the similarity between two samples in the process parameter space is measured;

[0186] If the process parameter vectors of two samples are close, the value of the kernel function K(x, x′) is close to 1; if they are far apart, the value of the kernel function approaches 0;

[0187] Using the RBF kernel function can effectively handle non - linear relationships, especially performing excellently in dealing with complex additive manufacturing processes;

[0188] Use the genetic algorithm to optimize the process parameters and the external field load parameters;

[0189] (1) Determine the weight coefficients based on experimental data analysis. Through experimental results, quantitatively analyze the influence degree of each index. Usually, methods such as regression analysis or sensitivity analysis can be used; including:

[0190] (11) Experimental data collection: Through a series of experiments, respectively obtain the data of pool stability (S pol ), microstructure uniformity (O micro ) and mechanical properties (Pmech);

[0191] (12) Standardization processing: In order to eliminate the influence of dimensions between different indexes, normalize each index:

[0192]

[0193] Among them, min is the minimum, max is the maximum, and norm is the standard;

[0194] Other indexes are processed similarly to ensure that the value ranges of all indexes are between [0, 1];

[0195] (13) Sensitivity analysis: Analyze the influence degree of each index on the final performance (such as mechanical properties). By changing the value of a single index through experimental design, observe the change range of the overall performance, and then quantify the weight;

[0196] (14) Weight calculation: Calculate the contribution ratio of the changes in each index to the overall performance as the weight coefficients α1, α2, α3:

[0197]

[0198] Among them, △F i represents the contribution of the changes in molten pool stability, microstructure uniformity, and mechanical properties to the objective function F; ΔF i represents the change amount of a certain index (such as molten pool stability, microstructure uniformity, or mechanical properties) under a certain optimization condition relative to the reference state; specifically:

[0199] If F i is molten pool stability, ΔF i represents the improvement amount of the optimization process on molten pool stability;

[0200] If F i is microstructure uniformity, ΔF i represents the improvement amount of the uniformity after optimization;

[0201] If F i is mechanical properties, ΔF i represents the improvement of the mechanical properties of the material (such as tensile strength) by the optimization;

[0202] The meaning of ΔF j :

[0203] ΔF j is a general term for ΔF i , representing the change amount of all objective functions, used for normalization;

[0204] The denominator part is the sum of the changes in all objective performance indicators (molten pool stability, microstructure uniformity, mechanical properties), representing the improvement amplitude of the overall performance;

[0205] (2) Determine the weight coefficient based on the machine learning method (LASSO regression): LASSO regression can automatically screen out the features (indicators) that contribute significantly to the objective function and quantify the importance of each indicator by introducing the L1 regularization constraint, and then determine the weight coefficient; including:

[0206] (21) Input data preparation: Organize the experimental data into matrix form, let X be the input variable matrix (including S pool , O micro , P mech ), and y be the target value (such as the overall performance evaluation index):

[0207]

[0208] (22) LASSO regression modeling: Use the LASSO regression model to fit the relationship between the input variables and the objective function:

[0209]

[0210] where β = [α1, α2, α3] is the weight coefficient; λ is the regularization hyperparameter that controls sparsity and ensures the rationality of the weight coefficient;

[0211] (23) Weight coefficient output: Through LASSO regression training, the model will automatically adjust the weight coefficients α1, α2, α3 so that they reflect the contribution of each index to the objective function;

[0212] (24) Result verification: Check the fitting effect of the regression model, such as the mean squared error (MSE) and the coefficient of determination (R 2 ), to ensure the accuracy of the weight coefficient;

[0213] To find the optimal parameter combination X opt and Y opt that can minimize the objective function F(x), an optimization algorithm is introduced, such as:

[0214] (3) Genetic Algorithm (GA) The genetic algorithm simulates the process of natural selection and evolution, and finds the optimal solution through continuous iteration of "selection", "crossover", and "mutation" operations; including:

[0215] (31) Initialization: Randomly generate an initial combination of a set of process parameters X and external field parameters Y;

[0216] (32) Objective function calculation: Substitute each set of (X, Y) into the objective function F(x) and calculate the corresponding fitness value (i.e., the objective function value);

[0217] (33) Selection: Select a parameter combination with better performance according to the fitness value;

[0218] (34) Crossover and mutation: Perform crossover and random mutation on the selected parameter combination to generate a new generation of candidate solutions;

[0219] (35) Iteration: Repeat the calculation of the objective function and iterate until the convergence condition is met (for example, the objective function no longer decreases significantly);

[0220] (36) Output the optimal solution: Finally, obtain the parameter combination X opt , Y opt ;

[0221] The objective function is:

[0222] F(x) = α1S pool+α2·O micro +α3·P mech

[0223] where S pool represents the stability of the molten pool, O micro represents the uniformity of the microstructure, and P mech is a mechanical property index; α1, α2, and α3 are weighting coefficients;

[0224] Output the optimal process parameter combination X opt , Y opt ;

[0225] In step 3.2, dynamically adjust the parameters: Apply the optimized parameters to the arc additive manufacturing process, and dynamically adjust the parameters of the magnetic field and ultrasonic field in combination with the real-time monitoring results; Through real-time feedback, ensure that the stability of the manufacturing process and the mechanical properties of the parts reach the design goals;

[0226] Influence of the magnetic field on the molten pool flow: The magnetic field changes the flow pattern of the molten pool metal through the Lorentz force, thereby affecting the stability of the molten pool; The magnetic field influence on the molten pool flow can be expressed by the following formula:

[0227]

[0228] where F is the Lorentz force; q is the electric charge; B is the magnetic field strength; ρ is the density of the fluid; is the velocity gradient representing the change of the flow velocity in space; represents the acceleration distribution of the velocity changing with position, affecting the flow stability of the molten pool;

[0229] By adjusting the magnetic field strength and direction, the flow pattern of the molten pool can be effectively controlled, and defects such as cracks and pores can be reduced;

[0230] Influence of ultrasonic waves on the molten pool: Ultrasonic waves vibrate particles and generate a bubble effect, making the metal inside the molten pool cool more uniformly and reducing the non-uniformity of the microstructure; The temperature change of the molten pool caused by ultrasonic waves can be expressed as:

[0231]

[0232] In the formula, ΔT is the temperature change; α is the heating coefficient of ultrasonic waves on the metal; P is the power of ultrasonic waves; λ is the thermal conductivity of the metal;

[0233] The adjustment of the frequency and power of ultrasonic waves directly affects the cooling rate and solidification process of the molten pool, thereby improving the microstructure of the final material;

[0234] Apply the optimized parameters to the actual additive manufacturing process. By real-time monitoring the results (such as molten pool temperature, interlayer temperature, etc.), dynamically adjust the magnetic field and ultrasonic field parameters to ensure the stability of the manufacturing process and the performance of the parts;

[0235] Real-time feedback regulation formula:

[0236] y odjusted = y initial +Δy

[0237] Where y adjusted is the adjusted parameter value, y initial is the initial set value, and Δy is the amount adjusted according to the feedback signal;

[0238] Magnetic field regulation formula:

[0239] B adjusted = B initial +ΔB

[0240] Where B adjusted is the adjusted magnetic field strength, B initial is the initial magnetic field strength, and ΔB is the adjustment amount, which is adjusted according to the real-time monitoring and optimization calculation results;

[0241] Ultrasonic field regulation formula:

[0242] P adjusted = P initial +ΔP

[0243] Where P adjusted is the adjusted ultrasonic field strength, P initial is the initial ultrasonic field strength, and ΔP is the adjustment amount, which is adjusted according to the optimization results;

[0244] Monitor the dynamic state of the molten pool (such as temperature, flow pattern, etc.) through high-frequency sensors and a real-time monitoring system; According to the real-time monitoring results, dynamically adjust the process parameters (such as welding speed, current, external field parameters, etc.) to keep the process always in an optimized state;

[0245] Dynamic feedback and parameter adjustment include:

[0246] (1) Define the real-time monitoring error: During the additive manufacturing process, the real-time monitoring system (such as temperature sensors, welding cameras, etc.) continuously provides data on the current state; For example, parameters such as temperature, molten pool shape, deposition rate, etc. can be collected in real time through sensors; In order to calculate the adjustment amount, it is first necessary to define the error (deviation) between the target value and the actual measured value; Assume:

[0247] y target is the target process parameter (for example, target temperature, target molten pool depth, etc.);

[0248] y measured is the process parameter measured in real time by the sensor;

[0249] Then, the error (deviation) Δy can be calculated as:

[0250] Δy = y measured - y target

[0251] Here, Δy represents the error between the current process parameter and the target parameter;

[0252] (2) Calculate the adjustment amount based on the error: According to the magnitude of the error, different adjustment methods can be adopted to correct the current process parameter; the adjustment amount is usually determined based on a certain proportional factor of the error amount; common control strategies include proportional-integral-differential (PID) control and adaptive control;

[0253] Assume that a simple proportional control method is adopted, then the adjustment amount Δy adjusted is:

[0254] Δy adjusted = K p ·Δy

[0255] where K p is the proportional coefficient, which is used to adjust the intensity of the control; the proportional control strategy directly adjusts the process parameter based on the magnitude of the error, and is usually applicable to systems with fast feedback response and small errors;

[0256] The proportional coefficient K p is usually determined through experiments or optimization algorithms (such as grid search, particle swarm optimization, etc.) to achieve the optimal response of the system;

[0257] (3) More complex control strategy (PID control): For more complex dynamic systems, it may be necessary to introduce the PID control algorithm, which can adjust the process parameter more precisely. The form of PID control is:

[0258]

[0259] where K p is the proportional coefficient (adjusted based on the current error); K i is the integral coefficient (used to accumulate past errors and compensate for continuous deviations); K d is the differential coefficient (adjusted according to the error change rate to cope with future changes in advance);

[0260] The physical meaning of dt:

[0261] represents a small change in time;

[0262] In the integral dt represents the cumulative summation of the error Δy(t) over time;

[0263] In the differential term dt represents the rate of change of the error Δy(t) over time (i.e., the rate of change of the error);

[0264] The first term (proportional term): adjusted according to the current error Δy;

[0265] The second term (integral term): the cumulative summation of the error Δy(t) within the time range t, eliminating the persistent bias in the system;

[0266] The third term (differential term): calculates the rate of change of the error (i.e., the rate of change of Δy(t) with respect to time), used to anticipate the error trend in advance;

[0267] The advantage of PID control is that it can comprehensively consider the error, the historical accumulation of the error, and the trend of error change, thereby more precisely controlling the adjustment amount Δy adjusted , and is applicable to relatively complex and dynamically changing additive manufacturing processes.

[0268] Specific Embodiment Nine: Combined with Figure 1-2 This embodiment is described. The method for optimizing the process parameters of arc additive manufacturing by combining magnetic field and ultrasonic field in this embodiment, step 4 includes the following steps:

[0269] Step 4.1: Define key concepts and construct a semantic association network;

[0270] Step 4.2: Semantic reasoning and parameter matching;

[0271] Step 4.3: Mechanical property prediction and optimization;

[0272] Step 4.4: Verification and optimization iteration.

[0273] Specific Embodiment Ten: Combined with Figure 1-2 This embodiment is described. In step 4.1 of the method for optimizing the process parameters of arc additive manufacturing by combining magnetic field and ultrasonic field in this embodiment, the key concepts of arc additive manufacturing are defined, including process parameters, material properties, external field parameters, and mechanical properties; through the ontology method and the professional term dictionary, a semantic association network is constructed and a knowledge base is established;

[0274] In step 4.2, relevant parameter combinations are matched in the knowledge base, and the recommended process parameters and additional field load parameters are extracted through the semantic inference engine, and optimized in combination with user requirements to generate the initial process parameters and additional field parameters that best meet the current manufacturing requirements;

[0275] In step 4.3, for the optimization results of the process parameters and additional field parameters, multi-output support vector regression (SVR) is used for mechanical property prediction; a regression model is constructed through the kernel function to predict multiple mechanical property indexes (such as tensile strength, elongation, etc.); grid search or Bayesian optimization is used to optimize the parameters to improve the prediction accuracy of the model;

[0276] In step 4.4, the manufacturing results are verified. If there are deviations, re-optimization and adjustment are carried out until the design requirements are met; if the performance does not meet the design requirements, the optimization model or experimental plan is re-adjusted according to the deviations, the data set is updated, and the optimization process is repeated until the performance requirements are met; combined with dynamic adjustment and real-time monitoring, the stability and consistency of the manufacturing process are further improved; after the manufacturing process is completed, the results are comprehensively verified to ensure that the performance meets the requirements; the effective parameter combinations are stored in the knowledge base as a reference for subsequent optimization to form a closed-loop feedback.

[0277] Example 1:

[0278] Combined with Figure 1-5 , the process parameter optimization design method for arc additive manufacturing combining magnetic field and ultrasonic field includes the following steps:

[0279] Step 1: Acquisition of process parameters and additional field data

[0280] Step 1.1: Obtain the geometric model data of the target part Obtain the three-dimensional geometric model data of the target part, and divide the part into several sub-regions according to this data. This helps with subsequent local process optimization for each sub-region.

[0281] Step 1.2: Determine the material type and obtain physical properties

[0282] According to the material type of the target part, obtain the physical properties of the material, defined as the set W = {w1, w2, w3, …, w n}, which includes but is not limited to specific heat capacity, thermal expansion coefficient, density, thermal conductivity, yield strength, Young's modulus, etc.

[0283] Step 1.3: Collect process parameters

[0284] Collect the process parameters in the arc additive manufacturing process, defined as the set X = {x1, x2, x3, …, x n} including welding current, voltage, welding speed, arc waveform, arc frequency, arc polarity, preheating temperature, interpass temperature, interpass waiting time, etc.

[0285] Step 1.4: Input the external field load parameters

[0286] Input the external field load parameters, defined as the set Y = {y1, y2, y3, …, y m} including magnetic field strength, magnetic field direction, ultrasonic field frequency, and ultrasonic field strength, etc.

[0287] Step 1.5: Obtain the mechanical property data

[0288] Obtain the mechanical property data of the material, defined as the set Z = {z1, z2, z3, …, z p} including tensile strength, elongation, fatigue performance, impact toughness, and high-temperature strength, etc.

[0289] Step 2: Data processing and key parameter screening

[0290] Step 2.1: Data normalization

[0291] Perform normalization on all the obtained parameter data, remove invalid data and fill in missing values to ensure the integrity and high quality of the data, providing a good foundation for the subsequent model establishment.

[0292] Perform normalization on all the parameter data,

[0293] Remove invalid data and fill in missing values to provide high-quality data for the subsequent model establishment;

[0294] Step 2.2: Use the random forest algorithm to screen key parameters

[0295] Use the random forest algorithm to screen key process parameters and external field parameters. By calculating the importance of features, select the parameters that are highly correlated with the target mechanical properties to ensure the accuracy and feasibility of the optimization goal.

[0296] Through the calculation of feature importance, screen out the key parameters that are highly correlated with the target mechanical properties (such as tensile strength, fatigue performance, etc.). The formula for calculating feature importance is as follows:

[0297] where, gini(f i ) represents the Gini impurity under different features, and N is the number of samples.

[0298] After data preprocessing, use LASSO regression and principal component analysis (PCA) for feature selection and dimensionality reduction of process parameters to identify the key parameters that have the greatest impact on the molten pool quality and the final material properties.

[0299]

[0300] yi: The actual mechanical property value corresponding to the i-th sample (such as the tensile strength, elongation rate, etc. measured by experiments); n is the number of samples;

[0301] Xi: The feature vector of the process parameters and external field parameters corresponding to the i-th sample (such as parameters like welding current, voltage, magnetic field strength, etc.);

[0302] β j is the regression coefficient, and p is the number of coefficients;

[0303] λ is the regularization penalty parameter, which is used to control the sparsity of the coefficients;

[0304] LASSO regression can select the features most relevant to the target mechanical properties from all input features, thereby reducing the computational complexity and improving the accuracy of the model.

[0305] Principal component analysis (PCA) finds the most principal components by performing eigenvalue decomposition on the covariance matrix of the data, reduces the dimension of the feature space, thereby reducing redundant information and noise, and improving the accuracy and computational efficiency of the model.

[0306] For dimensionality reduction processing, principal component analysis (PCA) is used to reduce redundant information and extract the main process parameters and external field parameters that have the greatest impact on the mechanical properties.

[0307] Principal component analysis (PCA): PC = W T X

[0308] where W is the feature matrix, X is the original data matrix, and PC is the data after dimensionality reduction;

[0309] Such as Figure 3 shown, Step 3: Machine learning and multi-objective optimization

[0310] Step 3.1: Multi-objective optimization

[0311] Adopt a machine learning optimization algorithm to perform multi-objective optimization on the process parameters and external field parameters. Through the optimization algorithm, the optimal combination of process parameters and external field load parameters is output, thereby maximizing the mechanical properties.

[0312] Based on the feature selection results, the relationship between the process parameters and the mechanical properties is established through a multi-output support vector regression (SVR) model. SVR performs non-linear mapping through a kernel function, enabling accurate prediction of material properties even in a complex additive manufacturing environment. By training the SVR model, the mechanical properties under different combinations of process parameters can be predicted, such as strength, ductility, and fatigue properties, etc.

[0313] SVR objective function:

[0314] w is the weight vector of SVR; C is the penalty parameter; ∈ i is the deviation of each sample point; C is the penalty parameter, used to control the model complexity and prevent overfitting.

[0315] Perform non - linear mapping through the RBF kernel function:

[0316] Radial basis function (RBF) kernel:

[0317] x, x′ are input data points. Each sample x is a combination of process parameters, for example, x = [x1, x2, x3, …, xn] (such as current, voltage, welding speed, etc.);

[0318] x′ is the combination of process parameters of another sample x′ = [x1′, x2′, x3′, …, x n ′]

[0319] By calculating the Euclidean distance ||x - x′|| between two process parameter vectors x and x′ 2 , measure the similarity of two samples in the process parameter space.

[0320] If the process parameter vectors of two samples are close, the value of the kernel function K(x, x′) is close to 1; if they are far apart, the value of the kernel function approaches 0.

[0321] σ is the parameter of the kernel function, controlling the similarity between data points;

[0322] Using the RBF kernel function can effectively handle non - linear relationships, especially performing excellently in dealing with complex additive manufacturing processes;

[0323] Optimize the process parameters and the external field load parameters using the genetic algorithm, and the objective function is:

[0324] F(x) = α1·S pool +α2·O micro +α3·P mech

[0325] Among them, S pool represents the molten pool stability, O micro represents the microstructure uniformity, P mech is the mechanical property index, and α1, α2, α3 are weight coefficients.

[0326] Output the optimal process parameter combination X opt , Y opt .

[0327] Such asFigure 4 As shown in Figure 4 , Step 3: Dynamically adjust parameters

[0328] Apply the optimized parameters to the arc additive manufacturing process, and dynamically adjust the parameters of the magnetic field and ultrasonic field in combination with the real-time monitoring results. Through real-time feedback, ensure that the stability of the manufacturing process and the mechanical properties of the parts reach the design goals.

[0329] Influence of magnetic field on molten pool flow:

[0330] The magnetic field changes the flow pattern of the molten pool metal through the action of the Lorentz force, thereby affecting the stability of the molten pool. The magnetic field influence on the molten pool flow can be expressed by the following formula:

[0331] F is the Lorentz force; q is the electric charge; B is the magnetic field strength; ρ is the density of the fluid;

[0332] By adjusting the magnetic field strength and direction, the flow pattern of the molten pool can be effectively controlled, and defects such as cracks and pores can be reduced.

[0333] Influence of ultrasonic waves on the molten pool: Ultrasonic waves vibrate particles and produce a bubble effect, making the metal inside the molten pool cool more uniformly and reducing the non-uniformity of the microstructure. The temperature change of the molten pool caused by ultrasonic waves can be expressed as:

[0334] ΔT is the temperature change; α is the heating coefficient of ultrasonic waves on the metal; P is the power of ultrasonic waves; λ is the thermal conductivity of the metal;

[0335] The adjustment of the frequency and power of ultrasonic waves directly affects the cooling rate and solidification process of the molten pool, thereby improving the microstructure of the final material.

[0336] Apply the optimized parameters to the actual additive manufacturing process. Through real-time monitoring results (such as molten pool temperature, interlayer temperature, etc.), dynamically adjust the parameters of the magnetic field and ultrasonic field to ensure the stability of the manufacturing process and the performance of the parts.

[0337] Real-time feedback control formula: y adjusted = y initial + Δy

[0338] Where y adjusted is the adjusted parameter value, y initial is the initial set value, and Δy is the amount adjusted according to the feedback signal.

[0339] Magnetic field adjustment formula: B adjusted = B initial + ΔB

[0340] Where B adjusted is the adjusted magnetic field strength, B initial$B_0$ is the initial magnetic field strength, and $\Delta B$ is the adjustment amount, which is adjusted according to the real-time monitoring and optimization calculation results.

[0341] Ultrasonic field adjustment formula: $P$ adjusted $=$ $P$ initial $+$ $\Delta P$

[0342] where $P$ adjusted is the adjusted ultrasonic field strength, $P$ initial is the initial ultrasonic field strength, and $\Delta P$ is the adjustment amount, which is adjusted according to the optimization results;

[0343] The dynamic state of the molten pool (such as temperature, flow pattern, etc.) is monitored by high-frequency sensors and a real-time monitoring system. According to the real-time monitoring results, the process parameters (such as welding speed, current, external field parameters, etc.) are dynamically adjusted to keep the process in an optimized state all the time;

[0344] As Figure 5 shown: Dynamic feedback and parameter adjustment

[0345] ① Define the real-time monitoring error: During the additive manufacturing process, the real-time monitoring system (such as temperature sensors, welding cameras, etc.) continuously provides data on the current state. For example, parameters such as temperature, molten pool morphology, deposition rate, etc. can be collected in real time by sensors. To calculate the adjustment amount, we first need to define the error (deviation) between the target value and the actual measured value.

[0346] Assume:

[0347] $y$ target is the target process parameter (for example, target temperature, target molten pool depth, etc.).

[0348] $\hat{y}$ measured is the process parameter measured in real time by the sensor.

[0349] Then, the error (deviation) $\Delta y$ can be calculated as:

[0350] $\Delta y = y$ measured $ - \hat{y}$ target

[0351] Here, $\Delta y$ represents the error between the current process parameter and the target parameter.

[0352] ② Calculate the adjustment amount based on the error: According to the magnitude of the error, we can adopt different adjustment methods to correct the current process parameters. The adjustment amount is usually determined based on a certain proportional factor of the error amount. Common control strategies include proportional-integral-derivative (PID) control and adaptive control.

[0353] Assume that we adopt a simple proportional control method, then the adjustment amount $\Delta y$adjusted is: Δy adjusted = K p ·Δy

[0354] where K p is the proportionality coefficient, which is used to adjust the intensity of control. The proportional control strategy directly adjusts process parameters based on the magnitude of the error and is usually applicable to systems with fast feedback response and small errors.

[0355] The proportionality coefficient K p is usually determined through experiments or optimization algorithms (such as grid search, particle swarm optimization, etc.) to achieve the optimal response of the system.

[0356] ③ More complex control strategy (PID control): For more complex dynamic systems, it may be necessary to introduce the PID control algorithm, which can adjust process parameters more precisely. The form of PID control is:

[0357]

[0358] K p : Proportionality coefficient (adjusted based on the current error).

[0359] K i : Integral coefficient (used to accumulate past errors and compensate for persistent deviations).

[0360] K d : Differential coefficient (adjusted according to the rate of change of the error to anticipate future changes).

[0361] The advantage of PID control is that it can comprehensively consider the error, the historical accumulation of the error, and the trend of change of the error, so as to more precisely control the adjustment amount Δy adjusted , and is applicable to relatively complex and dynamically changing additive manufacturing processes.

[0362] 3.2: Verification and optimization iteration Verify the manufacturing results. If there are deviations, re - perform optimization adjustment until the design requirements are met.

[0363] Manufacture test specimens and conduct mechanical property tests, recording indicators such as tensile strength and ductility.

[0364] If the performance does not meet the standards, adjust the optimization weight coefficients α1, α2, α3K p , K i , K d Re - perform optimization iteration.

[0365] Step 4: Knowledge base construction and intelligent optimization

[0366] Step 4.1: Define key concepts and construct a semantic association network

[0367] Define the key concepts of arc additive manufacturing, including process parameters, material properties, external field parameters, and mechanical properties. Construct a semantic association network and establish a knowledge base through ontology methods and a professional term dictionary.

[0368] Step 4.2: Semantic reasoning and parameter matching

[0369] Match relevant parameter combinations in the knowledge base, extract recommended process parameters and external field load parameters through a semantic reasoning engine, and optimize them in combination with user requirements to generate initial process parameters and external field parameters that best meet the current manufacturing requirements.

[0370] Step 4.3: Verification and optimization iteration

[0371] Verify the manufacturing results. If there are deviations, re-optimize and adjust until the design requirements are met. If the performance does not meet the design requirements, re-adjust the optimization model or experimental plan according to the deviations, update the data set, and repeat the optimization process until the performance requirements are met. Combine dynamic adjustment and real-time monitoring to further improve the stability and consistency of the manufacturing process. After completing the manufacturing process, comprehensively verify the results to ensure that the performance meets the requirements. Store the effective parameter combinations in the knowledge base as a reference for subsequent optimization to form a closed-loop feedback.

[0372] It should be noted that in the above embodiments, as long as the technical solutions are not contradictory, they can be arranged and combined. Those skilled in the art can exhaust all possibilities based on the mathematical knowledge of permutations and combinations. Therefore, the present invention will no longer explain the technical solutions after permutation and combination one by one, but it should be understood that the technical solutions after permutation and combination have been disclosed by the present invention.

[0373] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for optimizing the process parameters of arc additive manufacturing by combining magnetic field and ultrasonic field, characterized in that: The following steps are involved: Step 1: Acquisition of process parameters and external field data; Step 2: Data normalization processing; Step 3: Machine learning and multi-objective optimization; Step 4: Knowledge base construction and intelligent optimization.

2. The arc additive manufacturing process parameter optimization design method combining magnetic field and ultrasonic field according to claim 1 is characterized in that: Step 1 includes the following steps: Step 1.1, obtain the geometric model data of the target part; Step 1.2, determine the material type and obtain physical properties; Step 1.3, collect process parameters; Step 1.4, input the external field load parameters; Step 1.5: Obtain mechanical properties data.

3. The arc additive manufacturing process parameter optimization design method combining magnetic field and ultrasonic field according to claim 2 is characterized in that: In step 1.1, the three-dimensional geometric model data of the target part is obtained, and the part is divided into regions according to the data to obtain a plurality of sub-regions; In step 1.2, the physical properties of the material are obtained according to the material type of the target part, which is defined as the set W = {w1,w2,w3,…,w n }, physical properties include, but are not limited to, specific heat capacity, coefficient of thermal expansion, density, thermal conductivity, yield strength, and Young's modulus; In step 1.3, the process parameters of the arc additive manufacturing process are collected and defined as the set X = {x1, x2, x3…, x n }, process parameters include welding current, voltage, welding speed, arc waveform, arc frequency, arc polarity, preheating temperature, interlayer temperature, and interlayer waiting time; In step 1.4, the external field load parameters are input, which are defined as the set Y = {y1, y2, y3, …, y m }Including magnetic field strength, magnetic field direction, ultrasonic field frequency and ultrasonic field strength; In step 1.5, the mechanical properties data of the material are obtained, which is defined as the set Z = {z1,z2,z3,…,z p }, including tensile strength, elongation, fatigue performance, impact toughness and high temperature strength characteristics.

4. The arc additive manufacturing process parameter optimization design method combining magnetic field and ultrasonic field according to claim 3 is characterized in that: In step 2, the random forest algorithm is used to screen key process parameters and external field parameters.

5. The arc additive manufacturing process parameter optimization design method combining magnetic field and ultrasonic field according to claim 4, characterized in that: Step 2 includes the following steps: Step 2.1: Data normalization: All parameter data are normalized; Step 2.2, feature importance calculation: By calculating the feature importance, the key parameters that are highly correlated with the target mechanical properties are screened out; the feature importance calculation formula is as follows: Among them, Δgini(f i , n): feature f i The reduction in Gini impurity caused in the nth tree; T: the total number of trees in the random forest; Gini impurity definition: For the sample data set in a node, the Gini impurity calculation formula is: p k : The proportion of samples belonging to the kth class in the node; K: The total number of categories; After data preprocessing, LASSO regression and principal component analysis were used to perform feature selection and dimension reduction on process parameters to identify the key parameters that have the greatest impact on the melt pool quality and final material properties; Where n is the number of samples; y i is the actual value of the mechanical properties corresponding to the i-th sample; X i is the characteristic vector of process parameters and external field parameters corresponding to the i-th sample; β j is the regression coefficient; p is the number of coefficients; λ is the regularization penalty parameter; For dimensionality reduction, principal component analysis was used to reduce redundant information and extract the main process parameters and applied field parameters that have the greatest impact on mechanical properties; Principal Component Analysis: PC=W T TX Among them, W is the feature matrix, X is the original data matrix, and PC is the data after dimensionality reduction.

6. The arc additive manufacturing process parameter optimization design method combining magnetic field and ultrasonic field according to claim 5, characterized in that: Step 3 includes the following steps: Step 3.1: Multi-objective optimization; Step 3.2: Dynamic adjustment and real-time monitoring.

7. The arc additive manufacturing process parameter optimization design method combining magnetic field and ultrasonic field according to claim 6, characterized in that: In step 3.1, the machine learning algorithm is used to perform multi-objective optimization on the selected key process parameters and external field parameters to simultaneously meet multiple mechanical performance requirements (such as tensile strength and elongation); feature selection is achieved through random forest algorithm, LASSO regression and principal component analysis method to ensure that the optimization process focuses on key parameters; genetic algorithm or particle swarm optimization algorithm is used to complete global optimization and output the optimal parameter combination that meets multi-objective performance requirements; In step 3.2, during the optimization process, these parameters are applied to the arc additive manufacturing process, and the magnetic field strength, direction, and ultrasonic field frequency and amplitude of the external field are dynamically adjusted in combination with real-time monitoring data; through the real-time feedback mechanism, the parameters are continuously corrected to maintain the stability of the manufacturing process, and the optimization accuracy is further improved through grid search or Bayesian optimization to ensure that the final performance meets the complex process requirements.

8. The arc additive manufacturing process parameter optimization design method combining magnetic field and ultrasonic field according to claim 7, characterized in that: In step 3.1, based on the feature selection results, the relationship between process parameters and mechanical properties is established through a multi-output support vector regression (SVR) model; SVR objective function: Where w is the weight vector of SVR; ∈ i is the deviation of each sample point; C is the penalty parameter; Nonlinear mapping through RBF kernel function: Where K(x,x′) is the radial basis function kernel; x,x′ are input data points; σ is the parameter of the kernel function; Each sample x is a combination of process parameters, for example, x = [x1, x2, x3, …, xn] (such as current, voltage, welding speed, etc.); x′ is another sample’s process parameter combination x′=[x1′,x2′,x3′,…,xn′]; By calculating the Euclidean distance ||xx′||2 between two process parameter vectors x and x′, the similarity of two samples in the process parameter space is measured; If the process parameter vectors of two samples are close, the value of the kernel function K(x,x′) is close to 1; if the two are far apart, the value of the kernel function approaches 0; Genetic algorithm is used to optimize process parameters and external field load parameters; The objective function is: F(x)=α1·S pool +α2·O micro +α3·P mech Among them, S pool Indicates the stability of the molten pool, O micro Indicates the uniformity of microstructure, P mech is the mechanical performance index, α1, α2, α3 are weight coefficients; Output the optimal process parameter combination X opt ,Y opt ; In step 3.2, dynamically adjust parameters: apply the optimized parameters to the arc additive manufacturing process, and dynamically adjust the parameters of the magnetic field and ultrasonic field based on the real-time monitoring results; through real-time feedback, ensure that the stability of the manufacturing process and the mechanical properties of the parts meet the design goals; The influence of magnetic field on molten pool flow: The influence of magnetic field on molten pool flow can be expressed by the following formula: Where F is the Lorentz force; q is the charge; B is the magnetic field strength; ρ is the density of the fluid; The influence of ultrasound on the molten pool: The temperature change of the molten pool caused by ultrasound can be expressed as: Where ΔT is the temperature change; α is the heating coefficient of the ultrasonic wave on the metal; P is the power of the ultrasonic wave; λ is the thermal conductivity of the metal; Real-time feedback control formula: y odjusted =y initial +Δy Among them, y adjusted is the adjusted parameter value, y initial is the initial setting value, Δy is the amount adjusted according to the feedback signal; Magnetic field regulation formula: B adjusted =B initial +ΔB Among them, B adjusted is the adjusted magnetic field strength, B initial is the initial magnetic field strength, ΔB is the adjustment amount, which is adjusted according to the real-time monitoring and optimization calculation results; Ultrasonic field adjustment formula: P adjusted =P initial +AP Among them, P adjusted is the adjusted ultrasonic field intensity, P initial is the initial ultrasonic field intensity, ΔP is the adjustment amount, which is adjusted according to the optimization results; According to the real-time monitoring results, the process parameters are adjusted dynamically to keep the process in an optimized state; Dynamic feedback and parameter adjustment include: (1) Definition of real-time monitoring error: During the additive manufacturing process, the real-time monitoring system continuously provides data on the current status; Assumptions: y target is the target process parameter (e.g., target temperature, target molten pool depth, etc.); y measured It is the process parameters measured in real time by sensors; Then, the error Δy can be calculated as: Δy=y measured -y target Δy represents the error between the current process parameters and the target parameters; (2) Calculate the adjustment amount based on the error: Control strategies include proportional-integral-derivative (PID) control and adaptive control; Assuming a simple proportional control method is used, the adjustment amount Δy adjusted for: Δy adjusted =K p ·Δy Among them, K p is the proportional coefficient, which is used to adjust the intensity of control; (3) Control strategy: The form of PID control is: Among them, K p is the proportionality coefficient; K i is the integral coefficient; K d is the differential coefficient.

9. The arc additive manufacturing process parameter optimization design method combining magnetic field and ultrasonic field according to claim 7 or 8, characterized in that: Step 4 includes the following steps: Step 4.1: Define key concepts and construct semantic association networks; Step 4.2: Semantic reasoning and parameter matching; Step 4.3: Mechanical properties prediction and optimization; Step 4.4: Verification and optimization iteration.

10. The arc additive manufacturing process parameter optimization design method combining magnetic field and ultrasonic field according to claim 9, characterized in that: In step 4.1, the key concepts of arc additive manufacturing are defined, including process parameters, material properties, applied field parameters and mechanical properties; a semantic association network is constructed and a knowledge base is established through ontology methods and professional terminology dictionaries; In step 4.2, relevant parameter combinations are matched in the knowledge base, and the recommended process parameters and external field load parameters are extracted through the semantic reasoning engine, and optimized in combination with user needs to generate initial process parameters and external field parameters that best meet current manufacturing needs; In step 4.3, the optimization results of process parameters and external field parameters are used to predict mechanical properties by multi-output support vector regression; a regression model is constructed by kernel function to predict multiple mechanical performance indicators; grid search or Bayesian optimization is used to tune parameters to improve the accuracy of model prediction; In step 4.4, the manufacturing results are verified. If there is a deviation, the optimization is re-adjusted until the design requirements are met. If the performance does not meet the design requirements, the optimization model or experimental plan is readjusted according to the deviation, the data set is updated, and the optimization process is repeated until the performance requirements are met. Combine dynamic adjustment with real-time monitoring to further improve the stability and consistency of the manufacturing process. After completing the manufacturing process, fully verify the results to ensure that the performance meets the requirements. Store effective parameter combinations in the knowledge base as a reference for subsequent optimization, forming a closed-loop feedback loop.

Images