Toufique-TextileAI/warp_tension: v1.0.0 — Official Code Release for Warp Tension ML Models

Overview This release provides the official codebase and dataset for the mathematical analysis and machine learning-based prediction of warp tension in textile weaving. The project successfully integrates traditional mechanical principles—specifically backrest and shed geometry—with modern supervised and unsupervised learning architectures. Core Features & Implementation Mathematical Modeling: Mathematical Modeling: Includes the derivation of the $T \propto \sin \theta \cdot \tan \theta$ relationship, validating the critical impact of backrest height on tension dynamics. Predictive Regression: Implementation of four major algorithms with optimized hyperparameters: Random Forest: 300 estimators (Best performer, $R^2 \approx 96%$). Decision Tree: Standard implementation ($R^2 \approx 94%$). ANN: 6-layer deep learning model with a (100-400-300-100-50-1) architecture ($R^2 \approx 87%$). SVM: Linear kernel implementation using standard feature scaling ($R^2 \approx 50%$). Data Preprocessing: Integration of StandardScaler for distance-based and gradient-based models (SVM/ANN) to ensure convergence and accuracy. Unsupervised Analysis: PCA-driven dimensionality reduction to visualize variable clustering and variance (Dim1 and Dim2 explaining $\approx 69%$ of total variance). Repository Contents ML_for_warp_tension_regression.ipynb: Full regression pipeline. ML_for_warp_tension_classification.ipynb: Tension level grouping and classification. ML_for_warp_tension_Unsupervised_learning.ipynb: PCA and clustering logic. Tension_data_set_2024.csv: The primary dataset containing 288 experimental entries. How to Use Ensure you have scikit-learn, pandas, and tensorflow installed. Load the dataset directly into the provided Jupyter Notebooks. The models are configured to predict tension based on the 6 primary loom parameters identified in the study.