An Improved Control Chart Pattern Classification using a Transfer Learning-Based VGG-16 Network

Control chart pattern recognition is a crucial statistical process control tool used to determine whether a process operates within intended parameters or exhibits abnormal behavior. Accurate and automated recognition is essential for manufacturing companies to maintain high-quality production. Recent studies have combined machine learning with control chart pattern recognition, yet these methods often assume identical feature spaces and statistical distributions between training and test data—a condition rarely met in practice. Retraining models from scratch with new data is time-consuming. Transfer learning offers a more efficient alternative by leveraging knowledge from pre-trained models. This study employs the pre-trained VGG-16 deep convolutional neural network to classify control chart patterns, significantly improving recognition performance without extensive retraining. Through Monte Carlo simulations and a real-world case study, the proposed method achieved a recognition accuracy of 99.3%, outperforming conventional MLP, 1D-CNN, and CNN models. Sensitivity analysis demonstrated that the use of ReLU activation, batch normalization, and dropout techniques substantially enhanced accuracy and model robustness. The results confirm the potential of the proposed transfer learning-based recognizer to improve intelligent quality control in manufacturing processes by effectively handling limited training data and complex feature extraction.