基于深度学习的双相不锈钢应力-应变场预测模型

Prediction Model of Stress-Strain Fields in Duplex Stainless Steel Based on Deep Learning

摘要

通过人工智能技术深度解析金属材料多尺度构效关系，构建基于深度学习的成分-工艺-性能高通量逆向设计范式，在材料研发的过程中具有重要作用．本研究提出了一种基于条件生成对抗网络(CGAN)的端到端深度学习模型，用于研究双相不锈钢微观组织与力学性能之间的关系．该模型结合了博弈论思想，通过整合双相不锈钢微观组织图像及仪器化压痕试验获取的相组织力学性能数据，实现了微观组织-性能关系的直接预测．模型数据库通过基于微观组织的有限元模拟方法构建，确保了训练数据的高保真性．结果表明，该模型能够直接从双相不锈钢的微观组织预测应力-应变场，其预测结果与有限元模拟和实验数据高度吻合．

Abstract

In materials research and development，the in-depth analysis of multiscale constitutive relationships of metallic materials using artificial intelligence technology and the construction of a high-throughput composition-process-property inverse design paradigm based on deep learning play an important role. In this study，an end-to-end deep learning model based on a conditional generative adversarial network(CGAN) is proposed to investigate the relationship between microstructure and mechanical properties of duplex stainless steel. The model incorporates game theory ideas and achieves a direct prediction of the microstructure-property relationship by integrating microstructure images of duplex stainless steel and phase microstructure mechanical property data obtained from instrumented indentation test. The model database is constructed using a microstructure-based finite element simulation method，which ensures high fidelity of the training data. The results show that the model can predict the stress-strain field directly from the microstructure of duplex stainless steel，and its prediction results are highly consistent with the finite element simulation and experimental data.

关键词

双相不锈钢 / 纳米压痕 / 条件生成对抗网络 / 微观组织 / 应力-应变场

Key words

duplex stainless steel / nanoindentation / conditional generative adversarial network / microstructure / stress-strain field

参考文献

[1] Rhines F N. Microstructure—property relationships in materials[J]. Metallurgical Transactions A, 1977, 8:127-133.

[2] Geers M G D, Yvonnet J. Multiscale modeling of microstructure—property relations[J]. MRS Bulletin, 2016, 41(8):610-616.

[3] Nath S D, Irrinki H, Gupta G, et al. Microstructure—property relationships of 420 stainless steel fabricated by laser—powder bed fusion[J]. Powder Technology, 2019, 343:738-746.

[4] Wang J S, Mulholland M D, Olson G B, et al. Prediction of the yield strength of a secondary—hardening steel[J]. Acta Materialia, 2013, 61(13):4939-4952.

[5] Wang C C, Zhang C, Yang Z C, et al. Microstructure analysis and yield strength simulation in high Co—Ni secondary hardening steel[J]. Materials Science and Engineering：A, 2016, 669:312-317.

[6] Ren D, Wei X L, Wang C C, et al. Deep learning—based method for microstructure—property linkage of dual—phase steel[J]. Computational Materials Science, 2023, 227:112285.

[7] Zheng X L, Zheng P, Zheng L, et al. Multi—channel convolutional neural networks for materials properties prediction[J]. Computational Materials Science, 2020, 173:109436.

[8] Yang Z, Yu C H, Buehler M J. Deep learning model to predict complex stress and strain fields in hierarchical composites[J]. Science Advances, 2021, 7(15):eabd7416.

[9] Oliver W C, Pharr G M. Measurement of hardness and elastic modulus by instrumented indentation：Advances in understanding and refinements to methodology[J]. Journal of Materials Research, 2004, 19(1):3-20.

[10] Dao M, Chollacoop N, Van Vliet K J, et al. Computational modeling of the forward and reverse problems in instrumented sharp indentation[J]. Acta Materialia, 2001, 49(19):3899-3918.

[11] Goodfellow I, Pouget—Abadie J, Mirza M, et al. Generative adversarial networks[J]. Communications of the ACM, 2020, 63(11):139-144.

[12] Harms J, Lei Y, Wang T H, et al. Paired cycle GAN based image correction for quantitative cone beam computed tomography[J]. Medical Physics, 2019, 46(9):3998-4009.

[13] Zhang C Z, Gong B M, Deng C Y, et al. Computational prediction of mechanical properties of a C—Mn weld metal based on the microstructures and micromechanical properties[J]. Materials Science and Engineering：A, 2017, 685:310-316.

基金资助

国家自然科学基金资助项目(52375376)