Hybrid Deep Learning Model For Early Fault Detection In Energy-intensive Tablet Press Equipment: Mlp–1d Cnn Fusion For Emis Applications

Abstract

Fault Detection In Pharmaceutical Tablet Press Equipment Is Crucial For Ensuring Product Quality, Minimizing Downtime, And Reducing Energy Consumption In Energy-intensive Manufacturing Environments. This Study Presents A Hybrid Deep Learning Model, Namely MLP–1D CNN FaultNet, Which Integrates A Multilayer Perceptron (MLP) And A One-dimensional Convolutional Neural Network (1D CNN). The Architecture Employs Parallel Branches To Capture Both Global Statistical Dependencies And Localized Feature Patterns. The MLP Branch Models Global Feature Interactions, While The 1D CNN Branch Extracts Spatial Correlations Through Convolutional Operations. The Learned Representations Are Fused In A Dedicated Layer And Further Refined Using Dense Layers With Dropout And Batch Normalization To Improve Generalization. The Final Classification Layer Performs Fault Detection Effectively. Experimental Results Indicate That The Hybrid Model Outperforms Standalone MLP And CNN Models In Terms Of Accuracy, Precision, Recall, And F1-score. Therefore, The Architecture Is Suitable For Real-time Monitoring, Predictive Maintenance, And Energy-aware Fault Management In Pharmaceutical Manufacturing Systems.

Keywords