A Novel Ensemble Adaptive Sparse Bayesian Transfer Learning Machine for Nonlinear Large-Scale Process Monitoring

Area of Science:

• Engineering
• Computer Science
• Data Science

Background:

• Effective process monitoring is crucial for the safe and stable operation of large-scale industrial equipment.
• Nonlinear fault diagnosis presents challenges due to complex system dynamics and data variability.
• Existing methods may struggle with insufficient training data in real-world industrial settings.

Purpose of the Study:

• To propose a novel data-driven process monitoring framework for nonlinear fault diagnosis.
• To enhance the capabilities of Bayesian methods by incorporating transfer learning for improved data utilization.
• To address the challenge of limited training data in industrial process monitoring.

Main Methods:

• Development of the ensemble adaptive sparse Bayesian transfer learning machine (EAdspB-TLM).
• Re-derivation of the probabilistic relevance vector machine (PrRVM) within a Bayesian framework for forecasting operating conditions.
• Integration of transfer learning into a sparse Bayesian learning framework to enable knowledge transfer from source domains.
• Re-utilization of source domain data to mitigate issues related to insufficient training data.

Main Results:

• The proposed EAdspB-TLM framework demonstrated effective application in monitoring a real wastewater treatment process (WWTP).
• The framework was also successfully applied to monitor the Tennessee Eastman chemical process (TECP).
• Experimental results validated the feasibility and effectiveness of the EAdspB-TLM for nonlinear fault diagnosis.

Conclusions:

• The EAdspB-TLM offers a robust and feasible solution for data-driven process monitoring and nonlinear fault diagnosis.
• The integration of transfer learning and sparse Bayesian methods enhances the ability to handle complex industrial processes with limited data.
• The framework shows significant potential for improving the safety and stable operation of industrial equipment.