MULTIVARIATE STATISTICAL TECHNIQUES FOR QUALITY CONTROL AND PROCESS MONITORING

Abstract

This study applies a comprehensive multivariate statistical framework to monitor and predict product quality in a manufacturing process. Using data from 400 production observations, the analysis integrates descriptive statistics, correlation assessment, Principal Component Analysis (PCA), Hotelling’s T² control charting, and logistic regression modeling. PCA effectively reduced data dimensionality, revealing key latent factors explaining over half of total process variance. Hotelling’s T² analysis identified multivariate outliers, indicating occasional deviations from normal operating conditions. The logistic regression classifier demonstrated moderate accuracy but limited sensitivity, highlighting the trade-off between model interpretability and defect detection capability. Overall, the integrated framework enhances understanding of process variability, supports early fault detection, and strengthens data-driven decision-making in industrial quality control. The study underscores the value of combining traditional multivariate statistics with predictive analytics for intelligent manufacturing and continuous process improvement.