Information Fusion Fault Diagnosis Method for Deep-Sea Human Occupied Vehicle Thruster Based on Deep Belief Network

Area of Science:

• Marine Engineering
• Robotics
• Artificial Intelligence

Background:

• Deep-sea human occupied vehicles (HOVs) rely on thrusters for navigation and control.
• Accurate and timely fault diagnosis is critical for HOV safety and mission success.
• Existing fault diagnosis methods struggle with the complex, dynamic, and uncertain fault patterns in HOV thrusters.

Purpose of the Study:

• To propose a novel multisensor information fusion fault diagnosis method for deep-sea HOV thrusters.
• To introduce a deep belief network (DBN) for identifying uncertain and continuously changing fault patterns.
• To evaluate the effectiveness and accuracy of the proposed DBN-based method.

Main Methods:

• Developed a multisensor information fusion model incorporating a deep belief network (DBN).
• Utilized thruster control voltage, feedback current, and rotational speed as inputs.
• Defined a fault degree parameter (s) as the output to indicate fault pattern and severity.

Main Results:

• Conducted pool experiments simulating various fault conditions.
• The DBN-based method successfully diagnosed continuously changing, uncertain, and unknown thruster faults.
• Achieved higher identification accuracy compared to traditional artificial neural network-based methods.

Conclusions:

• The proposed DBN information fusion fault diagnosis method is effective for deep-sea HOV thrusters.
• This approach offers superior accuracy in identifying complex and dynamic fault patterns.
• The DBN method represents a significant advancement in ensuring the reliability of deep-sea exploration vehicles.