Fatigue Strength Prediction using Machine Learning in Steels

Fatigue strength is an intrinsic mechanical characteristic that predicts the endurance of materials under repeated cyclic stresses over a period.Predicting fatigue behaviour accurately is important to avoid early failure in essential applications like aerospace, automotive, and structural engineering.This project focuses on the significance of estimating rotating bending fatigue strength on the basis of material composition, heat treatment characteristics, and microstructural inclusion properties.In order to capture the multifaceted, non-linear dependencies among these variables and fatigue performance, various regression models have been employed, i.e., Gradient Boosting, XGBoost , Random Forest (MS), and Linear Regression.All these models reflect different insights of learning patterns from the dataset and help in mapping the prediction landscape.Missing or undefined user input is handled through K-Nearest Neighbours (KNN) imputation so fatigue strength can be estimated even from incomplete data.Users input known parameters like carbon content, alloying elements, heat treatment information, and inclusion information and the model estimate the fatigue strength accordingly.This model facilitates flexible, data-driven decision-making in material design and evaluation procedures.