STUDY EFFECT OF ALLOYING ELEMENT AND TESTING TEMPERATURE ON AUSTENITIC STAINLESS STEEL USING ARTIFICIAL NEURAL NETWORK

• Main Author: Zaicho Nur Ahmad, Alif
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/61001
• Institution: Institut Teknologi Bandung

Austenitic stainless steel alloys are the most widely used types of stainless steel because they have good corrosion resistance, high temperature creep resistance, impact resistance at low temperatures, and good workability. This alloy most commonly used in chemical reactors, heat exchangers, and other uses. Material failure in one of its uses can be caused by combination of its exposure due to high temperatures and mechanical loads. The current challenge is how to design the desired steel properties to reduce the experimental time and costs of its alloy development. In this study, the effect of alloying element and testing temperature on mechanical properties of the austenitic stainless steel was studied using artificial neural network modeling method. Series of experiments artificial neural network modeling have been carried out to learn the effect of 16 alloying elements and testing temperature as input on austenitic stainless steel mechanical properties such as yield strength, tensile strength, %elongation, and %reduction of area as output. Modeling was carried out in python and it’s libraries with certain modeling parameters and then the optimal number of neurons and hidden layers was determined. Sensitivity test is determined toward mechanical properties and temperature. Based on modeling results, the optimal ANN architectural model for determining yield strength, tensile strength, %elongation, and %reduction of area are 18-77-77-77-77-77-77-77-11, 18-98-98-98-98-98-98-98-11, 18-88-88-88-88-88-88-11, dan 18-65-65-65-65-11 with loss function respectively 7,711, 12,706, 1,232, and 1,780. The effect of alloying element on mechanical cannot be determined because it slightly different from literature. The effect of testing temperature on mechanical properties with ANN modeling is good enough according to the actual data, although there are still errors.