Research on Steel Strip Quality Detection Model Based on PCA Dimensionality Reduction and Regression Analysis

This research introduces a quality detection model for cold-rolled steel strips that combines Principal Component Analysis (PCA) for dimensionality reduction with regression analysis. The study involves data preprocessing to correct outliers and employs a decision tree regression model to predict hardness, identifying key parameters affecting quality. The model leverages a new feature matrix post feature engineering and selection to enhance accuracy. PCA is applied to reduce dimensionality, and the resultant features are used in various regression models and a BP neural network. The models are evaluated using MSE, RMSE, MAE, and R-squared, with SHAP values providing insights into feature contributions. The integration of PCA and regression techniques proves effective for steel strip quality detection, offering a valuable tool for production optimization.