Prediction Of Microstructure And Properties Of Stainless Steel Based On Physical Metallurgical Model And Machine Learning

Stainless steel,known for its superior properties such as high strength,good corrosion resistance and wear resistance,high-temperature oxidation resistance,and ease of processing,plays a vital role in various fields including medical,food,transportation,construction,and harsh industrial environments.However,as the performance and quality demand for stainless steel become increasingly multiplicity in all walks of life,the traditional research model based on a large number of destructive experiments and the establishment of mathematical statistical models faces issues such as long cycles,high costs,and low efficiency.Therefore,it is particularly important to improve the level of intelligent research on stainless steel and enhance the accuracy of prediction models.This article presents a comprehensive investigation into the microstructure and properties of stainless steel using experimental data and machine learning methods.By analyzing and preprocessing both experimental and computer simulation data,we explore the physical metallurgy models,machine learning techniques,and the integration of physical metallurgy and machine learning,high-precision prediction of the microstructure and properties of stainless steel has been achieved.This has a positive guiding effect on the optimization and control of stainless steel properties.Here are the primary research contents covered in this article:(1)Through the collection,transformation,and integration of data generated from experimental and theoretical calculations,the collected data were analyzed and processed using data preprocessing techniques to ensure the quality and accuracy of the database,providing a sufficiently rich dataset for machine learning-based prediction of the microstructure and properties of stainless steel.(2)The crystal plasticity finite element method is based on the study of the deformation behavior of materials at the microscopic level,which has the disadvantages of complex modeling and large computational workload.To address this,a prediction framework combining machine learning and CPFEM was proposed to study the microstructure evolution of stainless steel,using the dataset obtained from CPFEM simulations for training and testing.The findings indicate that the suggested prediction framework has significant computational advantages,high flexibility,and high prediction accuracy,providing favorable conditions for materials research in the era of big data.(3)In response to the fact that the traditional mathematical statistics methods may not be sufficient to accurately capture the nonlinear relationship between rheological stress and process parameters,a thermal deformation prediction model combining BP neural network and particle swarm optimization algorithm was proposed.The results show that the PSO-BP prediction model proposed in this study accurately predicts the rheological stress and thermal processing map,and has strong reliability and applicability.It provides theoretical guidance and technical support for the forging,rolling and other processes of stainless steel.(4)To balance the high accuracy of machine learning prediction models and the basic principles of physical metallurgy,a machine learning prediction model guided by physical metallurgy was proposed.Physical metallurgy parameters are calculated and used as additional input features to transform the original data space,fully extracting the implicit relationships between data.Using stainless steel 304 as an example,the results show that the put forward PM-PSO-LSSVM model has powerful dependability and adaptability,providing theoretical guidance for optimizing the composition,process parameters,and mechanical properties of stainless steel.