Servo torque fault diagnosis implementation for heavy-legged robots using insufficient information

Area of Science:

• Robotics
• Mechanical Engineering
• Control Systems

Background:

• Servo faults in Heavy-Legged Robots (HLR) are critical for reliability.
• Traditional diagnostics depend on fragile sensors, limiting robustness.
• Mechanical wear, environmental factors, and electrical issues cause servo faults.

Purpose of the Study:

• To develop a sensor-independent framework for HLR servo-fault diagnosis.
• To mitigate the reliance on delicate sensors in diagnostic systems.
• To enable real-time servo torque fault diagnosis.

Main Methods:

• Utilizing joint position and permanent magnet synchronous motor (PMSM) information.
• Developing a model connecting PMSM currents to HLR motion.
• Implementing an improved sliding mode observer (ISMO) for unmeasured outputs.
• Employing a Fourier expansion model for fault-free disturbance characterization.
• Using a dual-line particle filter (DPF) for disturbance prediction.
• Integrating DPF outputs as feedforward to ISMO for real-time diagnosis.

Main Results:

• A novel framework for HLR servo-fault diagnosis was established.
• The proposed method effectively diagnoses servo faults using limited sensor data.
• Real-time fault diagnosis was achieved even with unavailable servo outputs.
• Simulations and experiments validated the framework's accuracy and effectiveness.

Conclusions:

• The developed framework enhances HLR diagnostic reliability by reducing sensor dependency.
• The integration of PMSM data, ISMO, and DPF offers a robust solution for servo-fault detection.
• This approach provides a practical and validated method for real-world robotic applications.