Deep Learning-Based Intelligent X-Bar Control Chart for Monitoring Manufacturing Process

The X-bar control chart is a statistical tool widely employed in manufacturing to detect irregularities in product quality and machine deviation over time. This chart helps ascertain whether a process is statistically under control by establishing upper and lower limits derived from the probability distribution of the quality characteristic. However, traditional control chart methods have limitations, particularly in identifying anomalies beyond predefined patterns. The current study explores the application of deep learning methods to enhance the predictiveness of X-bar control charts. Initially, Long Short-Term Memory (LSTM) and Bidirectional Long Short-Term Memory (Bi-LSTM) models are utilized to forecast X-bar values in statistical process control. Subsequently, a case study is conducted by applying the trained models to the injection molding process, focusing particularly on the Cushion melt parameter, a critical aspect of the injection process. The models are trained and fine-tuned by adjusting various hyperparameters such as optimizer type, number of layers, and number of cell units. The LSTM model achieves an R-squared value of 0.739, Mean Squared Error (MSE) of 0.0043, and Mean Absolute Error (MAE) of 0.0507, while the Bi-LSTM model shows slightly lower performance with an R-squared of 0.702, MSE of 0.048, and MAE of 0.501. These outcomes can be utilized in control charts to predict quality status, thereby enhancing anomaly detection in the injection molding process.