CN-121995877-A - Intelligent decision-making method for forging-heat treatment quality energy efficiency synergistic technological parameters

CN121995877ACN 121995877 ACN121995877 ACN 121995877ACN-121995877-A

Abstract

The invention provides an intelligent decision method for forging-heat treatment quality energy efficiency synergistic process parameters, which relates to the technical field of heat processing process management and control, and comprises the steps of collecting data of a forging-heat treatment process to construct a sample library containing process parameters, quality indexes and energy efficiency indexes; the method comprises the steps of constructing a multi-objective evaluation model with a mechanism model and a data driving model fused on the basis of a sample library, modeling a forging-heat treatment process as a Markov decision process, constructing a reinforcement learning environment with technological parameters acting and quality indexes and energy efficiency indexes as multi-objective returns, generating a plurality of reference points in a multi-objective space, decomposing a multi-objective optimization task into a plurality of sub-problems, associating each reference point with at least one parameterization strategy, carrying out parallel sampling and strategy updating on strategies corresponding to each reference point, screening a current technological parameter solution by utilizing a non-dominant relation and a reference point neighborhood criterion, updating elite technological parameter files and outputting Pareto technological parameter solution sets.

Inventors

HUA LIN
HU ZHILI
WANG RUI

Assignees

武汉理工大学

Dates

Publication Date: 20260508
Application Date: 20260105

Claims (10)

1. The intelligent decision-making method for the forging-heat treatment quality energy efficiency synergistic process parameters is characterized by comprising the following steps of: collecting historical production data, test data and simulation data of a forging-heat treatment process, and constructing a sample library containing process parameters, quality indexes and energy efficiency indexes; constructing a multi-objective evaluation model fused with a mechanism model and a data driving model based on the sample library so as to represent the mapping relation between the technological parameters, the quality index and the energy efficiency index; Modeling the forging-heat treatment process as a Markov decision process, defining a state space, an action space and a multi-target return vector, and constructing a reinforcement learning environment taking the technological parameters as actions and the quality index and the energy efficiency index as multi-target returns; determining a multi-target space based on the quality index and the energy efficiency index output by the multi-target evaluation model, generating a plurality of reference points in the multi-target space, decomposing a multi-target optimization task into a plurality of sub-problems, and associating each reference point with at least one parameterization strategy; Adopting a deep reinforcement learning algorithm to sample and update strategies corresponding to each reference point in parallel, performing scalar quantization on multi-target return by using scalar functions based on the reference points, and introducing strategy update constraint; after each iteration, screening the current process parameter solution by utilizing a non-dominant relationship and a reference point neighborhood criterion, and updating an elite process parameter file with the synergistic quality and energy efficiency; And when training converges or a preset termination condition is reached, outputting a Pareto process parameter solution set with multiple targets of quality and energy efficiency from the elite process parameter file so as to guide the process scheme formulation of the forging-heat treatment process.
2. The intelligent decision-making method for the forging-heat treatment quality energy efficiency synergistic process parameters according to claim 1, wherein the multi-objective evaluation model is obtained by adopting a mechanism model and a data-driven model to be weighted and fused, and the calculation formula is as follows: ; Wherein: as a result of the prediction of the mechanism model, The data is output from the data-driven model, For the fusion weight coefficient, a is a technological parameter vector, and m is a target sequence number.
3. The intelligent decision-making method for the forging-heat treatment quality energy efficiency collaborative process parameters according to claim 1, wherein the multi-objective payback vector comprises at least one quality objective and one energy efficiency objective, wherein the quality objective is at least one of geometric dimension, yield strength, grain size, defect rate, tissue deviation degree or mechanical property deviation, and the energy efficiency objective is integrated energy consumption of a unit piece.
4. The intelligent decision-making method of forging-heat treatment quality energy efficiency collaborative process parameters according to claim 1, wherein the steps of generating a plurality of reference points in the multi-objective space, decomposing a multi-objective optimization task into a plurality of sub-problems, and associating each reference point with at least one parameterized strategy specifically comprise: generating a group of uniformly distributed reference vectors in the multi-target space, and carrying out normalization processing on the reference vectors so that the sum of elements of each reference vector is 1; defining scalar evaluation functions for each reference vector, wherein the scalar evaluation functions are obtained by calculating weighted maximum deviation between the performance of technological parameters on each target and ideal points and converting multi-target evaluation into single-target scalar values; each reference vector is associated with a parameterized decision strategy such that the optimization objective of the parameterized decision strategy is to minimize the expected value of its scalar function.
5. The intelligent decision-making method for the forging-heat treatment quality energy efficiency collaborative process parameters according to claim 4, wherein the steps of adopting a deep reinforcement learning algorithm to sample and update policies corresponding to each reference point in parallel, scaling multiple target rewards by scalar function based on the reference point, and introducing policy update constraints specifically comprise: constructing a strategy network and a state value network for each parameterized decision strategy; For each reference vector, calculating single-step scalar return according to the scalar evaluation function of the reference vector, and further calculating time sequence difference error and dominant function estimation; the strategy gradient method with clipping constraint is adopted to update the strategy parameters, and the strategy loss function for the kth reference vector is as follows: ; Wherein θ is a policy parameter, k is a reference point number, As a probability ratio of the new strategy to the old strategy, As a corresponding dominance function for the kth reference point, The clip is a cut-off function, and is a preset constraint coefficient; the total loss obtained by averaging the policy losses of all reference vectors updates the policy parameters and updates the value network parameters by minimizing the prediction error of the value network.
6. The intelligent decision-making method for the forging-heat treatment quality energy efficiency collaborative process parameters according to claim 1, wherein the step of screening the current process parameter solution using non-dominant relationships and reference point neighborhood criteria specifically comprises: Judging the dominant relation between the technological parameter solutions by adopting a non-dominant sorting method, and limiting the number of solutions in each reference point neighborhood based on the reference point neighborhood so as to keep the distribution uniformity and diversity of Pareto solution sets in a multi-target space.
7. The intelligent decision-making method of the forging-heat treatment quality energy efficiency synergistic process parameters according to claim 1, wherein when the forging-heat treatment process of the aluminum alloy automobile steering knuckle is optimized, the process parameters comprise at least one of blank heating temperature, heating and heat preserving time, preforming parameters, die forging deformation speed, die preheating temperature, solid solution treatment temperature and time, and artificial aging temperature and time; When optimizing the forging-heat treatment process of the aircraft landing gear forging, the process parameters comprise at least one of isothermal forging temperature, die forging reduction, normalizing temperature, quenching heating temperature and holding time, tempering or graded tempering temperature and time.
8. The intelligent decision-making system for the forging-heat treatment quality energy efficiency synergistic process parameters is characterized by comprising the following components: The sample library construction module is used for collecting historical production data, test data and simulation data of the forging-heat treatment process and constructing a sample library containing technological parameters, quality indexes and energy efficiency indexes; The evaluation model construction module is used for constructing a multi-target evaluation model fused with the mechanism model and the data driving model based on the sample library so as to represent the mapping relation between the technological parameters, the quality index and the energy efficiency index; The reinforcement learning environment construction module is used for modeling the forging-heat treatment process into a Markov decision process, defining a state space, an action space and a multi-target return vector, and constructing a reinforcement learning environment with technological parameters as actions and quality indexes and energy efficiency indexes as multi-target returns; The strategy association module is used for determining a multi-target space based on the quality index and the energy efficiency index output by the multi-target evaluation model, generating a plurality of reference points in the multi-target space, decomposing a multi-target optimization task into a plurality of sub-problems, and associating each reference point with at least one parameterized strategy; the strategy updating module is used for carrying out parallel sampling and strategy updating on strategies corresponding to all the reference points by adopting a deep reinforcement learning algorithm, carrying out scalar quantization on multi-target returns by utilizing scalar functions based on the reference points, and introducing strategy updating constraint; The elite solution set screening module is used for screening the current process parameter solution by utilizing a non-dominant relationship and a reference point neighborhood criterion after each iteration, and updating an elite process parameter file with the synergistic quality and energy efficiency; And the optimal solution set output module is used for outputting a Pareto process parameter solution set with multiple targets of quality and energy efficiency from the elite process parameter file when training converges or a preset termination condition is reached so as to guide the process scheme formulation of the forging-heat treatment process.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the forging-heat treatment quality energy efficiency co-process parameter intelligent decision method of any one of claims 1 to 7.
10. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor performs the steps of the forging-heat treatment quality energy efficiency co-process parameter intelligent decision method of any one of claims 1 to 7.

Description

Intelligent decision-making method for forging-heat treatment quality energy efficiency synergistic technological parameters Technical Field The invention relates to the technical field of thermal processing process management and control, in particular to an intelligent decision method for forging-thermal treatment quality energy efficiency synergistic process parameters. Background The forging-heat treatment process is a core manufacturing link of key bearing components in the fields of aviation, automobiles, energy sources and the like, and the technical parameters directly determine the geometric accuracy, the organization performance and the service reliability of the parts. Meanwhile, the energy consumption ratio in the forging heating, heat preservation and heat treatment processes is high, and the method is a typical high-energy-consumption manufacturing link, and is a key problem to be solved urgently how to reduce energy consumption and improve energy utilization efficiency on the premise of guaranteeing quality and performance under the environment of green manufacturing. At present, the optimization of forging-heat treatment process parameters mainly depends on the following technical schemes. Firstly, a process engineer performs optimizing screening in a limited parameter combination by means of experience knowledge, orthogonal test, response surface method and the like based on engineering experience and test design. And secondly, predicting forging deformation, a temperature field and tissue evolution by using models such as finite element analysis, thermodynamic calculation and the like based on a physical mechanism model and a numerical simulation method, and determining a process window by using parameter scanning or a simple optimization algorithm. Thirdly, based on the traditional intelligent optimization algorithm or the single-target reinforcement learning method, the search technology such as a genetic algorithm, a particle swarm algorithm and the like is adopted, or the reinforcement learning is combined with a single performance index to carry out strategy learning. However, the traditional scheme has the limitations of strong dependence on artificial experience, low optimization efficiency and difficulty in obtaining a balanced optimal solution set, namely, firstly, the method based on experience and experiment is seriously dependent on expert knowledge and has high trial-and-error cost, secondly, the numerical simulation method has higher precision but huge calculation time and is difficult to support large-scale optimization in a high-dimensional parameter space, thirdly, the traditional intelligent optimization and single-target reinforcement learning method cannot effectively cope with the inherent conflict and trade-off relation between a plurality of targets such as quality, energy efficiency and the like, and only can often obtain a single-target local optimal solution or a subjectively weighted trade-off solution, so that the pareto optimal solution set with uniform distribution is difficult to systematically generate. Disclosure of Invention The invention aims to provide an intelligent decision method for forging-heat treatment quality energy efficiency synergistic process parameters, which aims to solve the problems that the traditional scheme in the background art has strong dependence on artificial experience, low optimization efficiency and is difficult to obtain the limitation of balanced optimal solution set. The invention provides a forging-heat treatment quality energy efficiency collaborative process parameter intelligent decision-making method, which comprises the following steps of collecting historical production data, test data and simulation data of a forging-heat treatment process, constructing a sample library comprising process parameters, quality indexes and energy efficiency indexes, constructing a multi-objective evaluation model by fusing a mechanism model and a data driving model based on the sample library so as to characterize a mapping relation between the process parameters and the quality indexes and the energy efficiency indexes, modeling the forging-heat treatment process into a Markov decision process, defining a state space, an action space and a multi-objective return vector, constructing a reinforcement learning environment by taking the process parameters as actions and taking the quality indexes and the energy efficiency indexes as multi-objective return, determining a multi-objective space based on the quality indexes and the energy efficiency indexes output by the multi-objective evaluation model, generating a plurality of reference points in the multi-objective space, decomposing the multi-objective optimization task into a plurality of sub-problems, associating each reference point with at least one parameterized strategy, adopting a depth reinforcement learning algorithm to conduct parallel sampling and updating on the corresponding