Online prediction of composite material drilling quality based on multi-sensor fusion

The study introduces a novel online prediction method using a multi-sensor fusion approach for assessing the drilling quality of composite materials in real-time. The Multi-sensor Fusion Long Short-Term Memory (MFLSTM) model, which incorporates a Stacked Sparse Autoencoder (SSAE) within a Bayesian deep learning framework, was developed to manage the uncertainty inherent in composite material processing. Experimental validation, utilizing a specifically constructed dataset from multi-sensor data including force, temperature, and vibration measurements, demonstrates that our approach significantly enhances the predictability of hole quality during drilling. The MFLSTM model outperformed traditional machining process monitoring techniques by reducing prediction errors by over 25%, offering both accurate point predictions and reliable interval estimates. This method not only advances the intelligence of composite component manufacturing but also facilitates its industrial application through the development of supportive software.