Extreme manufacturing process technological parameter optimization method and system fused with machine learning

Abstract

The invention relates to the technical field of intelligent manufacturing, and discloses an extreme manufacturing process technological parameter optimization method and system fused with machine learning. The method comprises the following steps: acquiring multi-source data from a manufacturing equipment sensor, and fusing to generate a material state vector; inputting a pre-training model to obtain a material coefficient transition trend; judging whether the trend fluctuation amplitude exceeds a preset threshold value or not, and if yes, marking key nodes and extracting feature parameters; for the key nodes, according to the characteristic parameters and the real-time data of the key nodes, a control algorithm is adopted to calculate the parameter adjustment amount; optimizing the control parameters based on the parameter adjustment amount, generating a control instruction sequence and transmitting the control instruction sequence to an actuator; and obtaining adjusted feedback data, comparing the adjusted feedback data with the transition trend, and if the deviation exceeds an allowable range, updating the pre-training model. Through deep fusion of predictive monitoring and intelligent control, accurate optimization and adaptive control of process parameters are realized, the stability of the extreme manufacturing process and the product quality are improved, and the energy consumption and the defect rate are reduced.

Description

Technical Field

[0001] The present application relates to the field of intelligent manufacturing technology, and in particular to a method and system for optimizing process parameters of extreme manufacturing processes by integrating machine learning. Background Art

[0002] In extreme manufacturing processes (such as material processing and forming under extreme conditions such as high temperature, high pressure, and high stress), precise control of process parameters (such as heating rate and pressure gradient) plays a decisive role in product quality, production efficiency, and cost control. As manufacturing technology develops towards high precision and high reliability, the demand for process parameter optimization is becoming increasingly urgent.

[0003] In the early days, process parameter control in extreme manufacturing processes relied primarily on operator experience, with manual observation and adjustments used to maintain process stability. However, this approach, heavily reliant on experience, struggled to cope with complex and changing operating conditions, resulting in low control accuracy and inefficiency.

[0004] With the development of sensor technology, control methods based on data feedback have gradually emerged. Sensors deployed on manufacturing equipment collect real-time data such as temperature and pressure, and adjust process parameters based on preset rules or simple feedback control algorithms. While this method has improved the level of control automation to a certain extent, it still has limitations and cannot handle the nonlinear changes in material properties and the complex coupling relationships between multiple parameters in extreme manufacturing processes.

[0005] In recent years, the development of machine learning technology has brought new opportunities for optimizing process parameters in extreme manufacturing processes. By learning and analyzing historical data, machine learning models can predict changing trends in material conditions, providing a reference for adjusting process parameters. However, existing machine learning-based methods mostly focus on a single prediction function and lack the ability to accurately identify key nodes, effectively integrate multi-source data, and dynamically optimize control parameters. This makes it difficult to achieve precise and adaptive control of process parameters in extreme manufacturing processes.

[0006] The present invention aims to integrate machine learning technology to solve the problems of low control accuracy, poor adaptability, and lack of dynamic optimization capabilities in existing extreme manufacturing process process parameter control methods, achieve precise optimization and adaptive control of process parameters, and improve the quality and efficiency of extreme manufacturing processes. Summary of the Invention

[0007] The present invention provides a method and system for optimizing process parameters of extreme manufacturing processes that integrates machine learning, aiming to address the problems of low process parameter control accuracy, poor adaptability, and insufficient response to dynamic changes in material states in existing extreme manufacturing processes. It can respond to sudden changes in material states in real time, achieve forward-looking adaptive optimization of process parameters, and improve the efficiency and quality of extreme manufacturing processes.

[0008] In a first aspect, the present application provides a method for optimizing process parameters of an extreme manufacturing process by integrating machine learning, the method comprising:

[0009] Step 1: Acquire multi-source data from sensors deployed on manufacturing equipment, and fuse the multi-source data to generate a material state vector;

[0010] Step 2: Inputting a pre-trained model based on the material state vector to obtain a material coefficient transition trend;

[0011] Step 3: determine whether the fluctuation amplitude of the material coefficient transition trend exceeds a preset threshold; if so, mark the key nodes and extract characteristic parameters;

[0012] Step 4: For the marked key nodes, a control algorithm is used to calculate the parameter adjustment amount according to the characteristic parameters and real-time data;

[0013] Step 5: Optimize the control parameters according to the parameter adjustment amount, generate a control instruction sequence, and transmit it to the actuator;

[0014] Step 6: Obtain the adjusted feedback data, compare it with the material coefficient transition trend, and update the pre-trained model based on the comparison result.

[0015] In a second aspect, the present application provides a system for optimizing process parameters of an extreme manufacturing process by integrating machine learning, the system comprising:

[0016] A data acquisition module is used to acquire multi-source data from sensors deployed on manufacturing equipment and fuse the multi-source data to generate a material state vector;

[0017] A trend analysis module, configured to obtain a material coefficient transition trend according to the material state vector input into a pre-trained model;

[0018] A fluctuation detection module is used to determine whether the fluctuation amplitude of the material coefficient transition trend exceeds a preset threshold. If so, it marks the key nodes and extracts characteristic parameters;

[0019] A parameter calculation module is used to calculate the parameter adjustment amount using a control algorithm based on the characteristic parameters and real-time data for the marked key nodes;

[0020] An instruction generation module is used to optimize the control parameters according to the parameter adjustment amount, generate a control instruction sequence, and transmit it to the actuator;

[0021] The model updating module is used to obtain the adjusted feedback data, compare it with the material coefficient transition trend, and update the pre-trained model based on the comparison result.

[0022] Compared with the prior art, the beneficial effects of the technical solution of this application are at least as follows:

[0023] 1. The transition trend of material coefficients is predicted through a neural network model, and the key nodes where fluctuations exceed the limit are accurately identified. The timing of the control system's action is changed from "after-the-fact remediation" to "pre-emptive prevention", which fundamentally solves the problem of control command lag, realizes advanced perception and active intervention of sudden changes in material state, and effectively avoids the occurrence of irreversible defects such as cracks and deformation.

[0024] 2. A hybrid control algorithm that integrates fuzzy reasoning and data fusion is adopted, which comprehensively considers multi-dimensional information such as trend change rate (slope) and real-time status (pressure). The calculated parameter adjustment amount is more scientific and more in line with the nonlinear characteristics of the process, realizing high-precision and strong robust synchronous optimization control of the heating and pressure systems.

[0025] 3. Through the closed-loop optimization mechanism, it can dynamically adapt to fluctuations in different material batches and process conditions, and always stabilize key process parameters in the optimal window, thereby significantly reducing product performance dispersion and scrap rate. At the same time, through optimization and regulation, it reduces ineffective energy loss, improves product quality consistency, overall production efficiency and economic benefits.

[0026] 4. The model is updated online by using the deviation between real-time feedback data and predicted trends, enabling the system to continuously learn new characteristics of the manufacturing process and continuously evolve optimization strategies, thereby maintaining a high-precision control level for a long time. This meets the stringent requirements of extreme manufacturing for long-term, high-stability production and enhances the system's self-learning and self-adaptive capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0028] Figure 1 A flowchart of the method for optimizing process parameters of an extreme manufacturing process by integrating machine learning in this application;

[0029] Figure 2Schematic diagram of the extreme manufacturing process parameter optimization system integrating machine learning in this application. DETAILED DESCRIPTION

[0030] The terms "first," "second," "third," "fourth," and so forth (if any) in the specification and claims of this application and in the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate so that the embodiments described herein can be implemented in an order other than that shown or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus that includes a series of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, product, or apparatus.

[0031] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 FIG. 1 is a flow chart of a method for optimizing process parameters of an extreme manufacturing process by integrating machine learning provided by the present invention. The flow chart specifically includes the following steps:

[0032] Step 1: Obtain multi-source data from sensors deployed on manufacturing equipment and fuse the multi-source data to generate a material state vector.

[0033] In a specific embodiment, the multi-source data includes temperature, pressure, and stress. The process of performing step 1 may specifically include the following steps:

[0034] (1) Data fusion technology is used to integrate temperature data, pressure data, and stress data, and generate an initial material state vector containing elastic modulus and yield strength;

[0035] (2) Preprocess the initial material state vector, remove noise data, and obtain the material state vector.

[0036] Specifically, multi-source data comes from sensors deployed at key locations on manufacturing equipment, such as hot presses or extruder mold surfaces. Temperature, pressure, and stress are measured using thermocouples, piezoresistive sensors, and strain gauges, respectively. These data are collected in real time at a certain frequency (e.g., 10 times per second) to ensure that instantaneous changes in the manufacturing process are captured.

[0037] Multi-source data is fused via a Kalman filter algorithm, which iteratively updates the state estimate based on the covariance matrix of each sensor data to generate a fusion signal, thereby suppressing single sensor errors and improving data robustness. Fusion technology can reduce errors caused by temperature fluctuations, improve vector accuracy, and thus more accurately predict material deformation. The fusion algorithm highlights the dominant effect of thermal stress through weighted distribution, effectively reducing the characterization deviation caused by thermal fluctuations. For example, in a high-temperature environment, the fusion algorithm can set the weight of temperature data to 0.6, pressure to 0.3, and stress to 0.1 to ensure that the vector reflects the material behavior dominated by thermal stress. Based on this, the robustness of the state vector can be improved and subsequent prediction deviations can be reduced.

[0038] Based on the fused signal, the elastic modulus is calculated using a modified form of Hooke's law: the elastic modulus is equal to stress divided by strain, where the strain is derived from the stress data and a temperature compensation factor. Simultaneously, the pressure and stress data are converted to equivalent stress based on the von Mises yield criterion. When the equivalent stress reaches a material-specific threshold, the yield point is marked, from which the yield strength index is calculated. The calculated elastic modulus and yield strength index are combined into a vector form, forming an initial material state vector containing the elastic modulus and yield strength. This vector is represented by a two-dimensional numerical pair [elastic modulus value, yield strength value]. The initial material state vector is preprocessed by first applying a median filter to remove abnormal noise. Subsequently, z-score normalization is used to transform each index into a distribution with a mean of zero and a standard deviation of one, resulting in a standardized material state vector.

[0039] In a preferred embodiment, principal component analysis is performed on the material state vector, the first n principal components are extracted, and they are combined to generate a eigenvector. The Euclidean norm of the eigenvector is calculated and defined as a comprehensive strength index. The eigenvector and the comprehensive strength index are combined to generate a characteristic expression vector.

[0040] Specifically, this technical solution is applied in continuous rolling manufacturing. For example, when processing steel, the feature expression vector can highlight the pattern of uneven stress distribution, which is beneficial for real-time adjustment of rolling force; in the ceramic sintering process, the extracted feature expression vector helps identify the modulus changes caused by temperature gradients and improve product consistency.

[0041] For example, assuming the normalized vector is [1.2, 0.8], the characteristic expression vector after principal component analysis is [0.9, 0.5], and its norm is 1.0. This characteristic expression vector is used as model input and can accurately reflect the transition of the material from elasticity to plasticity, bringing more precise manufacturing control effects.

[0042] For scenarios requiring high precision, such as aviation material processing, feature expression can be further integrated with time series analysis to calculate the vector change rate as a supplementary feature. This expansion can enhance sensitivity to dynamic changes and help avoid material failure.

[0043] Step 2: Input the pre-trained model based on the material state vector to obtain the material coefficient transition trend.

[0044] In a specific embodiment, the process of performing step 2 may specifically include the following steps:

[0045] The material state vector is input into the pre-trained neural network model. The neural network model processes the nonlinear change characteristics of the material coefficient through the activation function, outputs the predicted value of the material coefficient in the future time series, and generates the transition trend of the material coefficient.

[0046] Specifically, the material coefficient transition trend refers to the dynamic change pattern of the key coefficient of the material (i.e., the material coefficient) in a specific process (such as an extreme manufacturing process, covering material processing and forming under extreme conditions such as high temperature, high pressure, and high stress), as the process parameters such as time, temperature, and pressure change. This trend is expressed in the form of time series data or curves, depicting the law of the coefficient's evolution over time. Common material coefficients include but are not limited to elastic modulus, yield strength, thermal expansion coefficient, thermal conductivity, Poisson's ratio, and creep rate. In metal forming or composite material curing scenarios, elastic modulus and yield strength are core coefficients because they directly determine processing deformation and final component strength.

[0047] The neural network model is a multi-layer perceptron trained based on historical manufacturing data, which can capture the nonlinear response of the material during the heating process. The model uses activation functions such as ReLU to process the nonlinear change characteristics of the material coefficient. The input vector is calculated through the hidden layer and the output is the predicted value of the material coefficient in the future time series, thereby forming a material coefficient transition trend curve. This trend curve characterizes the dynamic changes in material properties during the manufacturing process. For example, in high-temperature alloy forging, the input vector reflects the current material state, and the model predicts the change curve of the elastic modulus in the next 10 seconds. For example, the elastic modulus will drop nonlinearly from 200GPa to 180GPa in the next 10 seconds, which can capture the performance degradation trend caused by sudden pressure changes. This predictive capability helps to identify material fatigue risks in advance and improve the stability of the manufacturing process.

[0048] Preferably, the neural network model is trained by a back-propagation algorithm, and the weight parameters are optimized using the mean square error as the loss function to ensure the mapping capability of nonlinear relationships.

[0049] Neural networks process the complex nonlinear patterns in material state vectors, compensating for the lack of accuracy of traditional physical models under extreme conditions. This ability to process and predict nonlinear characteristics enables the model to accurately predict the dynamic changes of materials under complex manufacturing conditions, providing data support for subsequent process parameter optimization. In this way, the problem of accurately predicting the dynamic changes in material states is solved, and precise control of material states in extreme manufacturing processes is achieved, improving processing stability and product quality, and significantly reducing the defect rate and energy consumption during production.

[0050] Step 3: Determine whether the fluctuation amplitude of the material coefficient transition trend exceeds the preset threshold. If so, mark the key nodes and extract the characteristic parameters.

[0051] In a specific embodiment, the process of executing step 3 may specifically include the following steps:

[0052] (1) Based on the preset time length, the material coefficient transition trend is segmented, the fluctuation amplitude of each segment is calculated, the points where the fluctuation amplitude exceeds the preset threshold are identified, and marked as key nodes;

[0053] (2) Perform linear fitting on a section of data centered on the key node in the material coefficient transition trend, calculate the transition slope of the section of data, and use the transition slope as the characteristic parameter corresponding to the key node;

[0054] (3) Generate characteristic description data of key nodes based on characteristic parameters and fluctuation amplitude.

[0055] Specifically, the material coefficient transition trend curve in the future short period of time is divided into multiple data segments according to a preset time length (such as 5 seconds), and the fluctuation amplitude of each segment is calculated based on this. The fluctuation amplitude is obtained by calculating the standard deviation of the trend curve in the specified time window (i.e. the preset time length) or the difference between the peak and the valley value, so as to quantify the degree of change of the material coefficient in each segment.

[0056] Identify points where the fluctuation amplitude exceeds the preset threshold. These points exceeding the threshold mean that the material coefficient has undergone a more significant change during the period. Mark them as key nodes and record the key nodes in the data log based on timestamp and amplitude information to ensure that the system can respond to potential material mutation points in a timely manner, thereby avoiding manufacturing defects such as crack formation.

[0057] After confirming the key node, extract a section of data centered on the key node and perform linear fitting. Preferably, use the least squares method or other methods to fit the trend data points and calculate the transition slope of the data section. The transition slope reflects the dynamic change rate of the material coefficient near the key node and reflects the speed of the material state change.

[0058] The transition slope is used as the characteristic parameter corresponding to the key node. Combined with the fluctuation amplitude calculated previously, feature description data containing information such as the fluctuation amplitude and transition slope is generated. The feature description data is encapsulated in structured data such as JSON format and used as input for subsequent fuzzy logic controllers.

[0059] In metal forging scenarios, this technology can promptly detect key nodes where sudden pressure increases lead to a decrease in elastic modulus. The extracted transition slope and fluctuation amplitude can be used to calculate the heating rate adjustment amount to avoid defects caused by material overstress. This solves the problem of difficulty in accurately grasping the timing and rate of change of material state mutations in existing manufacturing processes, thereby improving the stability of the manufacturing process and product quality.

[0060] For example, in semiconductor wafer manufacturing scenarios, transition trends involve thermal stress changes. If the fluctuation amplitude is 10%, exceeding the threshold of 8%, a critical node is marked. The slope obtained by further calculation is 2% / s, and the generated feature description data is [2, 10]. This is used to adjust the heating rate, which can prevent wafer warping and improve yield. In plastic injection molding scenarios, it can be used to handle stress trends. If the fluctuation amplitude is 15kPa, exceeding the threshold of 12kPa, a critical node is marked. The slope obtained by further calculation is 1.2kPa / s, and the generated feature description data is [1.2, 15]. This is used to support subsequent fuzzy control to adjust pressure and ensure product uniformity. This optimizes the production process and reduces defect rates through real-time feedback.

[0061] Combining trend analysis, fluctuation amplitude detection, and slope extraction effectively captures material state mutation points and provides timely and accurate adjustment information for the control system, avoiding material defects and reduced production efficiency caused by control lag. This efficient feedback mechanism significantly improves manufacturing process stability and product quality, while also reducing energy waste and material loss during production. In high-precision manufacturing, particularly in aerospace and precision alloy processing, this technology offers significant technical advantages, enabling rapid response and effectively preventing the adverse effects of material mutations.

[0062] Step 4: For the marked key nodes, the control algorithm is used to calculate the parameter adjustment amount based on the characteristic parameters and real-time data.

[0063] In a specific embodiment, the real-time data includes pressure data, and the process of executing step 4 may specifically include the following steps:

[0064] (1) Obtain the input variable combination corresponding to the key node;

[0065] (2) Fuzzify the input variable combination and map it into a fuzzy set, and perform fuzzy reasoning on the fuzzy set based on the preset fuzzy rule base;

[0066] (3) Based on the fuzzy inference results, the control output value is calculated by the defuzzification method, and the heating rate adjustment amount is calculated based on the control output value;

[0067] (4) Perform weighted fusion on the transition slope and pressure data to obtain the optimized rate correction value.

[0068] Specifically, an input variable combination refers to a set of data selected and combined for calculating parameter adjustments at a marked key node. For each key node, the relevant input variable combination is extracted from real-time data. These input variables typically include pressure data and material characteristic parameters (such as transition slope, elastic modulus, and yield strength) derived through trend analysis.

[0069] A fuzzy logic controller is used to process input variable combinations. For example, a fuzzy logic controller is a control method based on fuzzy set theory. It fuzzifies input variables, applies fuzzy rules for inference, and ultimately defuzzifies the output to address uncertainty and nonlinearity in the manufacturing process. First, characteristic parameters such as the transition slope and current pressure data are converted into fuzzy variables. For example, the transition slope is divided into three fuzzy sets: low, medium, and high, and the pressure data is divided into three fuzzy sets: small, medium, and large. The fuzzified data effectively handles the uncertainty and nonlinearity of material state changes during the manufacturing process. Then, inference is performed based on a preset fuzzy rule base. The fuzzy rule base contains rules trained based on expert experience or historical data. These rules guide the fuzzy inference process. For example, if the transition slope is high and the pressure is high, the control algorithm will output a larger adjustment. The design of these rules ensures that the algorithm can make appropriate adjustments based on the current material state and real-time pressure data, ensuring timely response to sudden changes in the material. Finally, the results of the fuzzy inference are converted into specific numerical values ​​(i.e., control output values) through defuzzification methods (such as the center average method).

[0070] For example, assume the extracted characteristic parameters are: a transition slope of 3 (reflecting the rate of change of the material coefficient), an elastic modulus of 180 GPa, a yield strength of 400 MPa, and real-time pressure data of 12 MPa. The transition slope is divided into three fuzzy sets: low (0-2), medium (2-4), and high (4-6), with the current transition slope of 3 mapped to the medium fuzzy set. The elastic modulus is divided into three fuzzy sets: low (<150 GPa), medium (150-200 GPa), and high (>200 GPa), with 180 GPa mapped to the medium fuzzy set. The yield strength is divided into three fuzzy sets: low (<300 MPa), medium (300-450 MPa), and high (>450 MPa), with 400 MPa mapped to the medium fuzzy set. The pressure data is divided into three fuzzy sets: small (<10 MPa), medium (10-15 MPa), and large (>15 MPa), with 12 MPa mapped to the medium fuzzy set.

[0071] The pre-set fuzzy rule base contains a series of "if-then" rules, such as "If the transition slope is medium and the pressure is medium, then the heating rate adjustment direction is to appropriately reduce with a small adjustment range." The input fuzzy set is matched with the preconditions of each rule in the rule base. By matching the rules, the conclusions corresponding to the rules that meet the preconditions are extracted, forming a comprehensive fuzzy conclusion set. For each element in the fuzzy conclusion set, such as the description "appropriately reduce the heating rate with a small adjustment range," a membership function curve is constructed. The membership function describes the probability (membership degree) that each heating rate adjustment value belongs to the fuzzy description. The type (e.g., triangular, trapezoidal, etc.) and parameters of the membership function are generally determined based on actual production experience, expert knowledge, or historical data. Assume that a triangular membership function is used to describe the description "appropriately reduce the heating rate with a small adjustment range." The heating rate adjustment ratio is the horizontal axis and the membership degree is the vertical axis. Based on experience, the vertex (maximum membership point) of the triangle membership function corresponds to a heating rate adjustment ratio of 0.9 times the original rate, while the left and right bases correspond to adjustment ratios of 0.8 and 1.0 times the original rate, respectively. This determines how the fuzzy description's membership changes under different heating rate adjustments. Similar methods are used to determine the membership function curves of other fuzzy description elements. Subsequently, the centroid of each membership function curve in the fuzzy conclusion set is calculated, and a weighted average of these centroids is taken to determine the control output value.

[0072] The control output value is multiplied by a preset proportional factor to calculate the heating rate adjustment. For example, when the fuzzy inference result is 0.8, the control system may adjust the heating rate to 80% of the original rate, that is, reduce the heating rate to adapt to the change in material state.

[0073] When weighted fusion is performed on the transition slope and pressure data, different weights are assigned to the transition slope and pressure data. Typically, the transition slope is given a higher weight because it has a more significant impact on material state changes. For example, the transition slope might be weighted 0.6, while the pressure data might be weighted 0.4. Through weighted averaging, the fused rate correction value comprehensively reflects the impact of both, ensuring more precise adjustments. This weighted fusion rate correction value is used for subsequent parameter optimization, such as updating the pressure gradient, to ensure that the material coefficient changes match the pattern.

[0074] In practical applications, the fuzzy rule library can be adapted to different material types based on their properties. For example, for aluminum alloys, the fuzzy rules might emphasize rapid adjustment of the heating rate under low pressure conditions to prevent deformation caused by overheating; for steel, the rules might emphasize reducing the heating rate under high pressure conditions to prevent stress concentration. This diverse control approach enhances the algorithm's adaptability and enables higher adjustment accuracy across the manufacturing process for different materials, reducing energy consumption and improving product quality.

[0075] By integrating fuzzy logic control algorithms with material state characteristic data, not only is the real-time response capability of the manufacturing process enhanced, but it also ensures the effective adjustment of manufacturing parameters in a complex and changing environment, greatly improving manufacturing accuracy and efficiency. Through careful adjustment and optimization, material defects can be effectively avoided, the quality of the finished product can be improved, and energy can be saved.

[0076] In a preferred embodiment, a feature selection mechanism is introduced to dynamically adjust the input variables for fuzzy inference, including the following steps:

[0077] (1) Analyze the importance of input variables based on neural networks, evaluate the contribution of each input variable to the current material state, and adjust the weight of the input variable according to the contribution;

[0078] (2) Use principal component analysis to reduce the dimensionality of multidimensional input data, extract the top K main components that affect the change of material properties, and generate input variable combinations based on the top K main components;

[0079] (3) According to the real-time sensor feedback signal, the real-time importance of each input variable in different process stages and material conditions is evaluated, and the input variable combination is dynamically adjusted based on the real-time importance to ensure the best match between the input variables and the material state.

[0080] Specifically, by introducing a feature selection mechanism, the input variables used for fuzzy inference are dynamically adjusted, thereby optimizing process parameters during the manufacturing process. The feature selection mechanism dynamically analyzes the input variables using multiple data processing methods to ensure the optimal match between the input variables and the material state under different process stages and material conditions, thereby improving the accuracy and adaptability of the control process.

[0081] The neural network model learns the nonlinear relationship between material properties and process parameters through training data, enabling it to assess the impact of each input variable on the current material state. By learning the weights and biases of the neural network, it can accurately reflect the contribution of different input variables (such as temperature, pressure, elastic modulus, and yield strength) in a specific state. Based on these contributions, the system dynamically adjusts the weight of each input variable, assigning higher weights to important variables during the fuzzy inference process, thereby enabling the control system to more accurately reflect the current material state. For example, in a metalworking process, multiple input variables describe the material state, such as temperature (T), pressure (P), stress (S), elastic modulus (E), and yield strength (YS). A neural network analysis of the contribution of each input variable to the current material state yields the following results: temperature (T) 40%, pressure (P) 20%, stress (S) 15%, elastic modulus (E) 10%, and yield strength (YS) 15%. These contribution values ​​indicate that temperature has the greatest impact on the material state. Therefore, temperature is assigned a higher weight during model inference, while elastic modulus, which may have a smaller impact on the current material state, is assigned a lower weight.

[0082] Multidimensional data usually contains redundant information, and PCA, as an effective dimensionality reduction method, converts the original data into a few main components through linear transformation. These components can retain the variance of the original data as much as possible. By extracting the top K main components that affect the change in material properties, the system can eliminate the interference of irrelevant variables and focus on the most relevant features of the material state, thereby generating a more simplified and information-rich input feature combination, which helps to remove unnecessary noise, optimize the selection of input variables, and further improve the efficiency and accuracy of the model. For example, in the plastic injection molding process, the original input variables may include multiple dimensions such as temperature, pressure, and injection speed. Through principal component analysis, the two main components of temperature and pressure are extracted as the input variable combination for subsequent processing, thereby reducing the data dimension and improving computational efficiency.

[0083] On this basis, the system also incorporates real-time sensor feedback to assess the real-time importance of each input variable at different process stages and material conditions. During the manufacturing process, material properties and processing conditions can change dramatically, causing the importance of input variables to fluctuate accordingly. Therefore, the system needs to monitor feedback signals such as temperature and pressure in real time and adjust the weights of input features accordingly. This real-time importance feedback mechanism allows the system to dynamically adjust the input feature combination to ensure that the input variable combination at each moment is optimally matched to the current material state and processing conditions. This allows the control system to better adapt to dynamically changing environments and improve the real-time and accurate decision-making. For example, during the ceramic sintering process, as the sintering temperature increases, the impact of stress on material properties gradually increases. Based on real-time sensor feedback, the system dynamically adjusts the input variable combination, increasing the weight of stress and decreasing the weight of temperature, ensuring that the input variable combination always accurately reflects the current material state.

[0084] For example, in the processing of high-temperature alloys, as the temperature rises, the yield strength and elastic modulus of the material may change significantly. At this time, the system can adjust process parameters such as pressure and heating rate based on real-time feedback signals and the contribution of neural network analysis. In the low-temperature stage, it may rely more on the stress and temperature data of the material and less on the change in elastic modulus. By evaluating the importance of variables in real time and dynamically adjusting the input feature combination, the system can ensure that the process parameters always meet the optimal processing requirements of the material, thereby reducing the defect rate and improving product quality.

[0085] By introducing a feature selection mechanism and adaptive input variable generation logic, it is ensured that under different material and process conditions, the control algorithm can dynamically adjust the input variables based on real-time data and historical feedback, thereby optimizing the fuzzy reasoning process. This solves the problem of insufficient adaptability of fixed input feature combinations in complex and dynamic manufacturing environments, enhances the precise control capability and stability of the manufacturing process, and effectively improves production efficiency and product quality.

[0086] Step 5: Optimize the control parameters according to the parameter adjustment amount, generate a control instruction sequence, and transmit it to the actuator.

[0087] In a specific embodiment, in step 5, the control parameters are optimized according to the parameter adjustment amount to generate a control instruction sequence, including:

[0088] (1) The current pressure gradient parameter is updated based on the rate correction value, and the optimized heating rate is obtained based on the heating rate adjustment amount;

[0089] (2) Using the gradient descent algorithm to iteratively optimize the updated pressure gradient parameters to match the material coefficient transition trend, and obtain the optimized pressure gradient parameters. The gradient descent algorithm calculates the parameter deviation based on the loss function and updates the parameter value.

[0090] (3) Based on the optimized heating rate and optimized pressure gradient parameters, a control instruction sequence is generated to adjust the heating and pressure process parameters during the material manufacturing process.

[0091] Specifically, the rate correction value is multiplied by a preset weight factor and added to the existing pressure gradient parameter to form a preliminary updated value. The product of the heating rate adjustment amount and the current heating rate is used as the adjusted heating rate.

[0092] The gradient descent algorithm evaluates parameter deviations by calculating a loss function, updating the parameter values ​​at each iteration based on the gradient direction. The loss function is calculated based on the deviation between the pressure gradient parameter and the variation pattern of the material coefficient, which is derived from a preset material coefficient transition trend model. Through step-by-step iterative optimization, the gradient descent algorithm aligns the pressure gradient parameter with the actual variation trend of the material, ensuring precise control of material heating and applied pressure during the manufacturing process.

[0093] During this optimization process, by calculating the loss function and updated parameter values ​​after each iteration, an optimized pressure gradient parameter is ultimately obtained. This parameter can more accurately reflect the dynamic changes in material coefficients and adapt to the ever-changing manufacturing environment. For example, during metal heat treatment, the pressure gradient parameter optimized by the gradient descent algorithm can better match the changes in the material's yield strength and elastic modulus, reducing material defects caused by improper pressure control. In plastic extrusion manufacturing, the optimized pressure gradient parameter can effectively address fluctuations in material yield strength caused by temperature changes, ensuring product consistency and quality.

[0094] Based on the optimized heating rate and pressure gradient parameters, the control system generates a corresponding sequence of control instructions, which are then used by the actuator to adjust the heating and pressure parameters during the manufacturing process. The control instruction sequence is transmitted to the actuator via a communication interface, where it performs the corresponding adjustments, such as adjusting the heating rate and applying the pressure gradient, to ensure that the material state remains within the optimal range throughout the production process.

[0095] By combining a gradient descent algorithm with a fuzzy control system, we can achieve dynamic optimization and precise control of process parameters, effectively avoiding material defects, improving product quality, and reducing energy consumption and production costs. We can respond to and adjust process parameters in real time during the material processing process, ensuring precision and consistency at every step, significantly improving production stability and reliability.

[0096] In one embodiment, in step 5, sending a control instruction sequence to an actuator to adjust the manufacturing process includes:

[0097] (1) Formatting the control instruction sequence into instruction data that can be recognized by the actuator and transmitting it to the actuator through the communication interface;

[0098] (2) The actuator applies the command data in real time to adjust the heating rate and pressure gradient;

[0099] (3) Monitor the operating status of the actuator and generate a regulation execution log.

[0100] Specifically, the control instruction sequence, which includes numerical instructions such as the heating rate (temperature increase per second) and the pressure gradient (change per unit time), is converted into a standard protocol format, such as binary encoding, to ensure direct interpretation by the actuator, avoiding transmission errors, improving data compatibility, and ensuring accurate instruction execution. This formatted data packet is sent to the actuator using a wired or wireless interface such as RS-485 or Ethernet, ensuring real-time transmission latency of less than 50 milliseconds and maintaining the continuity of the manufacturing process.

[0101] After receiving the instruction data, the actuator system analyzes the heating rate and pressure gradient parts, activates the heating element to gradually increase the temperature to the target value, and monitors the current temperature to match the predicted material coefficient transition trend. It adjusts the pressure application device to ensure synchronization with the yield strength index of the material state vector to avoid material mutations, verifies the adjustment effect in real time, and uses built-in sensors to feedback the current heating and pressure values. Compared with the instruction, if the deviation exceeds 2%, the execution parameters are fine-tuned.

[0102] Through formatting, transmission and real-time adjustment, the actuator can accurately execute control instructions and achieve optimized adjustment of heating rate and pressure gradient. At the same time, it monitors the operating status and generates logs to form a closed-loop optimization, thereby improving overall prediction accuracy, manufacturing process stability and product quality.

[0103] Step 6: Obtain the adjusted feedback data and compare it with the material coefficient transition trend, and update the pre-trained model based on the comparison results.

[0104] In a specific embodiment, the process of executing step 6 may specifically include the following steps:

[0105] (1) Obtaining feedback data including temperature, pressure, and stress from sensors;

[0106] (2) Compare the feedback data with the material coefficient transition trend and calculate the deviation value;

[0107] (3) Determine whether the deviation value exceeds the preset allowable range. If so, generate a model update signal;

[0108] (4) Adjust the weight parameters of the pre-trained model according to the model update signal to generate an updated pre-trained model.

[0109] Specifically, the feedback data includes but is not limited to temperature, pressure, stress, crystal structure information, part dimensional accuracy, shape error, and surface roughness. The adjusted feedback data is obtained in real time from the sensor array deployed on the manufacturing equipment. These data reflect the actual material state after executing the control instruction sequence, ensuring the timeliness of the feedback data. The obtained feedback data is compared with the material coefficient transition trend predicted by the neural network model. The prediction error is quantified by calculating the data deviation at the corresponding time point. The deviation value can be calculated using methods such as the root mean square error to ensure the accuracy of the quantitative comparison. If the calculated deviation value exceeds the preset allowable range, a model update signal is automatically generated. This signal is used to trigger subsequent adjustments to avoid unnecessary computing overhead.

[0110] Based on the model update signal, the backpropagation algorithm is used to adjust the weight parameters of the pre-trained model. Feature vectors are extracted from the feedback data and input into the backpropagation algorithm. The gradient is calculated and the weight parameters of the neural network are adjusted. The adjustment is iterated until the deviation value falls within the allowable range, generating an updated pre-trained model.

[0111] By dynamically updating the pre-trained model, the model's accuracy in predicting material state changes is improved. The model parameters can be automatically adjusted based on different manufacturing scenarios and material states, enabling precise control of material states during extreme manufacturing processes. This solves the problem of inaccurate material state predictions by pre-trained models and the inability to dynamically adjust based on real-time feedback. This allows for precise control of material states during extreme manufacturing processes, improves processing stability and product quality, and significantly reduces defect rates and energy consumption during production.

[0112] The above describes the extreme manufacturing process parameter optimization method integrating machine learning in the embodiment of the present application. The following describes the extreme manufacturing process parameter optimization system integrating machine learning in the embodiment of the present application. Figure 2 , a schematic diagram of the structure of the extreme manufacturing process parameter optimization system integrated with machine learning provided by this application, the system includes:

[0113] The data acquisition module 10 is used to obtain multi-source data from sensors deployed on the manufacturing equipment and fuse the multi-source data to generate a material state vector.

[0114] The trend analysis module 20 is used to input the pre-trained model according to the material state vector to obtain the material coefficient transition trend.

[0115] The fluctuation detection module 30 is used to determine whether the fluctuation amplitude of the material coefficient transition trend exceeds a preset threshold. If so, it marks the key nodes and extracts characteristic parameters.

[0116] The parameter calculation module 40 is used to calculate the parameter adjustment amount for the marked key nodes according to the characteristic parameters and real-time data using a control algorithm.

[0117] The instruction generation module 50 is used to optimize the control parameters according to the parameter adjustment amount, generate a control instruction sequence, and transmit it to the actuator.

[0118] The model updating module 60 is used to obtain the adjusted feedback data, compare it with the material coefficient transition trend, and update the pre-trained model based on the comparison result.

[0119] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for optimizing process parameters of extreme manufacturing processes by integrating machine learning, characterized in that: include: Step 1: Acquire multi-source data from sensors deployed on manufacturing equipment, and fuse the multi-source data to generate a material state vector; Step 2: Inputting a pre-trained model based on the material state vector to obtain a material coefficient transition trend; Step 3: determine whether the fluctuation amplitude of the material coefficient transition trend exceeds a preset threshold; if so, mark the key nodes and extract characteristic parameters; Step 4: For the marked key nodes, a control algorithm is used to calculate the parameter adjustment amount according to the characteristic parameters and real-time data; Step 5: Optimize the control parameters according to the parameter adjustment amount, generate a control instruction sequence, and transmit it to the actuator; Step 6: Obtain the adjusted feedback data, compare it with the material coefficient transition trend, and update the pre-trained model based on the comparison result.

2. The method according to claim 1, wherein The multi-source data includes temperature, pressure and stress, and step 1 includes: Data fusion technology is used to integrate temperature data, pressure data, and stress data, and generate an initial material state vector including elastic modulus and yield strength; The initial material state vector is preprocessed to remove noise data to obtain the material state vector.

3. The method according to claim 1, wherein The step 2 includes: The material state vector is input into a pre-trained neural network model, and the neural network model processes the nonlinear change characteristics of the material coefficient through an activation function, outputs the material coefficient prediction value in the future time series, and generates the material coefficient transition trend.

4. The method according to claim 1, wherein The step 3 comprises: Segmenting the material coefficient transition trend based on a preset time period, calculating the fluctuation amplitude of each segment, identifying points where the fluctuation amplitude exceeds a preset threshold, and marking them as key nodes; Performing linear fitting on a segment of data centered at the key node in the material coefficient transition trend, calculating a transition slope of the segment of data, and using the transition slope as the characteristic parameter corresponding to the key node; Generate characteristic description data of key nodes based on the characteristic parameters and the fluctuation amplitude.

5. The method according to claim 4, wherein The real-time data includes pressure data, and step 4 includes: Obtaining the input variable combination corresponding to the key node; Fuzzifying the input variable combination and mapping it into a fuzzy set, and performing fuzzy reasoning on the fuzzy set based on a preset fuzzy rule base; A control output value is calculated based on the fuzzy inference result by a defuzzification method, and a heating rate adjustment amount is calculated based on the control output value; The transition slope and the pressure data are weightedly fused to obtain an optimized rate correction value.

6. The method according to claim 5, wherein Introducing a feature selection mechanism to dynamically adjust the input variables for fuzzy inference includes the following steps: Analyze the importance of input variables based on a neural network, evaluate the contribution of each input variable to the current material state, and adjust the weight of the input variable based on the contribution; Using principal component analysis to reduce the dimensionality of multidimensional input data, extract the top K main components that affect material performance changes, and generate input variable combinations based on the top K main components; According to the real-time sensor feedback signal, the real-time importance of each input variable under different process stages and material conditions is evaluated, and the input variable combination is dynamically adjusted based on the real-time importance to ensure the best match between the input variables and the material state.

7. The method according to claim 5, wherein In step 5, optimizing the control parameters according to the parameter adjustment amount to generate a control instruction sequence includes: updating a current pressure gradient parameter based on the rate correction value, and obtaining an optimized heating rate based on the heating rate adjustment amount; The updated pressure gradient parameters are iteratively optimized using a gradient descent algorithm to match the material coefficient transition trend, thereby obtaining the optimized pressure gradient parameters. The gradient descent algorithm calculates the parameter deviation based on the loss function and updates the parameter value. A control instruction sequence is generated based on the optimized heating rate and the optimized pressure gradient parameters to adjust the heating and pressure process parameters in the material manufacturing process.

8. The method according to claim 1, wherein In step 5, the control instruction sequence is sent to the actuator to adjust the manufacturing process, including: formatting the control instruction sequence into instruction data recognizable by the actuator, and transmitting the command data to the actuator via the communication interface; The actuator applies the command data in real time to adjust the heating rate and pressure gradient; Monitor the running status of the executor and generate the adjustment execution log.

9. The method according to claim 1, wherein The step 6 comprises: Acquiring the feedback data including temperature, pressure and stress from sensors; Comparing the feedback data with the material coefficient transition trend to calculate a deviation value; determining whether the deviation value exceeds a preset allowable range, and if so, generating a model update signal; The weight parameters of the pre-trained model are adjusted according to the model update signal to generate an updated pre-trained model.

10. A system for optimizing process parameters of extreme manufacturing processes by integrating machine learning, for implementing the method according to any one of claims 1 to 9, characterized in that: The system comprises: A data acquisition module is used to acquire multi-source data from sensors deployed on manufacturing equipment and fuse the multi-source data to generate a material state vector; A trend analysis module, configured to obtain a material coefficient transition trend according to the material state vector input into a pre-trained model; A fluctuation detection module is used to determine whether the fluctuation amplitude of the material coefficient transition trend exceeds a preset threshold. If so, it marks the key nodes and extracts characteristic parameters; A parameter calculation module is used to calculate the parameter adjustment amount using a control algorithm based on the characteristic parameters and real-time data for the marked key nodes; An instruction generation module is used to optimize the control parameters according to the parameter adjustment amount, generate a control instruction sequence, and transmit it to the actuator; The model updating module is used to obtain the adjusted feedback data, compare it with the material coefficient transition trend, and update the pre-trained model based on the comparison result.

Images