https://github.com/VishalGaud17/Steel-Surface-Fault-Classification-using-Machine-Learning

Overview

This project builds a machine learning–based system to classify surface faults in steel manufacturing using sensor and process data. The goal is to automate fault identification, reduce manual inspection effort, and support faster quality control decisions. The solution uses Python and Scikit-learn, applying robust preprocessing and a tuned Random Forest model to handle complex patterns and class imbalance in industrial data.

Responsibilities

• Designed the end-to-end machine learning workflow for multiclass fault classification
• Processed and transformed multiple fault indicator columns into a single target variable
• Built and trained a Random Forest classification model using Scikit-learn
• Implemented feature preprocessing using ColumnTransformer and RobustScaler
• Handled class imbalance using balanced class weights
• Performed hyperparameter tuning with GridSearchCV and cross-validation
• Evaluated model performance using classification metrics
• Exported trained model and preprocessing artifacts as reusable .pkl files
• Published model artifacts on Hugging Face for easy access and reuse

Outcome

• Automated the classification of steel surface faults, reducing dependency on manual inspection
• Improved consistency and reliability in fault identification across multiple defect types
• Delivered a reusable and deployment-ready model pipeline