Quenching-Tempering 조건이 탄소강 플랜지의 미세조직과 기계적 물성에 미치는 영향과 후열처리 공정 최적화를 위한 머신러닝용 데이터 엔지니어링

Abstract

It was investigated to optimize the post-heat treatment process for flange with high strength and high toughness. Machine learning was introduced for efficiency using the data gathered by experiments. The flange was divided into 40 locations according to cooling rate at quenching and data were constructed for 12 types of tempering conditions of 40 locations. Thermal profiles of each flange location were measured by FEM simulation and the fraction of second phase was characterized by in-house code automatically. Tensile and hardness tests were carried out at room temperature and Charpy impact test was performed at –46℃. A predictive model by CNN was learned by using total of 2,880 data collected by the experiments. This model showed meaningful possibility for the prediction of mechanical properties only with a small amount of data consisting of time-series thermal profile. It was investigated the effect of three locations selected by the cooling rate and various tempering conditions on the mechanical properties and microstructure of flange. It was observed that all specimens showed the distinct multiphase microstructure consisted of matrix and second phases which were decomposed during various tempering conditions. With the decrease of cooling rate of quenching, hardness and strength were decreased while elongation was increased and low temperature toughness was not significantly influenced. After tempering, hardness and strength were decreased while low temperature toughness was improved. However, the effect of tempering on the elongation was inconsistent.