Fault Detection for Nonlinear Process With Deterministic Disturbances: A Just-In-Time Learning Based Data Driven...

Area of Science:

• Industrial process monitoring and control
• Data-driven fault detection methodologies
• Nonlinear system analysis

Background:

• Data-driven fault detection is crucial for systems with unknown physical models.
• Deterministic disturbances are inherent in processes but often overlooked in nonlinear fault detection.
• Existing data-driven methods have limitations in addressing nonlinear dynamics with deterministic disturbances.

Purpose of the Study:

• To propose a novel just-in-time learning-based data-driven (JITL-DD) fault detection method.
• To address the challenge of fault detection in nonlinear processes affected by deterministic disturbances.
• To provide a robust and adaptive solution for industrial fault diagnosis.

Main Methods:

• Utilizing a just-in-time learning (JITL) scheme for process description.
• Employing local model structures to capture complex process dynamics and nonlinearity.
• Developing a data-driven framework for real-time fault detection and adaptation.

Main Results:

• The proposed JITL-DD method effectively detects faults in nonlinear systems with deterministic disturbances.
• The method demonstrates inherent online adaptation capabilities.
• High accuracy in fault detection was achieved on both a numerical example and a sewage treatment process benchmark.

Conclusions:

• The JITL-DD method offers a viable data-driven solution for fault detection in challenging nonlinear industrial environments.
• The approach effectively handles process nonlinearity and deterministic disturbances.
• The study validates the method's effectiveness and accuracy through practical examples.