Prediction of Continuous Cooling Transformation Diagrams for Ni-Cr-Mo Welding Steels via Machine Learning Approaches
- PDF / 1,317,545 Bytes
- 9 Pages / 593.972 x 792 pts Page_size
- 43 Downloads / 268 Views
https://doi.org/10.1007/s11837-020-04057-z Ó 2020 The Minerals, Metals & Materials Society
MACHINE LEARNING APPLICATIONS IN ADVANCED MANUFACTURING PROCESSES
Prediction of Continuous Cooling Transformation Diagrams for Ni-Cr-Mo Welding Steels via Machine Learning Approaches XIAOXIAO GENG,1 HAO WANG,1,7 ASAD ULLAH,2 WEIHUA XUE,3 SONG XIANG,4 LI MENG,5 and GUANG MA6 1.—School of Materials Science and Engineering, University of Science and Technology Beijing, Beijing, China. 2.—Department of Mathematical Sciences, Karakoram International University Gilgit, Gilgit-Baltistan 15100, Pakistan. 3.—School of Materials Science and Engineering, Liaoning Technical University, Fuxin, China. 4.—College of Materials and Metallurgy, Guizhou University, Guiyang, China. 5.—Metallurgical Technology Institute, Central Iron and Steel Research Institute, Beijing, China. 6.—State Key Laboratory of Advanced Power Transmission Technology, Global Energy Interconnection Research Institute Co., Ltd, Beijing, China. 7.—e-mail: [email protected]
Continuous cooling transformation diagrams in synthetic weld heat-affected zones (SH-CCT diagrams) are important tools to analyze the microstructure and mechanical properties of the heat-affected zone under certain welding conditions and to evaluate the weldability of steel. In this study, various machine-learning approaches are used to select an appropriate model for prediction of SH-CCT diagrams for Ni-Cr-Mo steels using relevant material descriptors including the chemical compositions and cooling rate. Random forest is the best model to predict the ferrite and bainite transition start temperature accurately, K-nearest neighbors is suitable for predicting the start temperature of martensite transformation, and random committee is used to predict the hardness. These optimal models are used to predict the SHCCT diagrams of five kinds of steels to verify the accuracy. The results show that the predicted values of the optimal models agree well with the experimental data with a strong correlation coefficient and low error value.
INTRODUCTION In welding, problems such as hardening, cold cracking, local embrittlement and re-cracking occur in the heat-affected zone of welding, which leads to weld failure, and these failures are often related directly to the microstructure in the heat-affected zone of welding under the cooling condition.1 Continuous cooling transformation diagrams (CCT diagrams) in welding steels show transformation temperatures and metallographic changes of weld HAZs during cooling from just below the melting temperatures of steels under different rates of cooling as well as the resulting constitution of micro-structures and hardness.2 The CCT diagram of weld HAZs is different from the CCT diagram of heat treatment.3 First, the highest heating temperature of the CCT diagram of the welding heataffected zone is between 1250°C and 1400°C, which
is only applicable to metals near the fusion line.4 However, the maximum heating temperature of the heat treatment CCT diagram is between 859°C and 9
Data Loading...
