Chaotic extension neural network theory-based XXY stage collision fault detection using a single accelerometer sensor

Area of Science:

• Mechanical Engineering
• Control Systems
• Signal Processing

Background:

• Collision detection in motion stages is critical for preventing damage and ensuring operational integrity.
• Existing methods may require multiple sensors or complex setups, increasing costs and implementation challenges.

Purpose of the Study:

• To propose and validate a novel, cost-effective collision fault detection system for XXY stages.
• To utilize chaos synchronization dynamics for identifying collision events.
• To develop a fault diagnosis model based on signal trajectory analysis.

Main Methods:

• Extracting stage characteristic signals and filtering vibratory magnitude.
• Implementing master-slave chaos error systems for signal processing.
• Analyzing trajectory diagrams of chaos synchronization dynamic error signals (E1 and E2).
• Utilizing an extension neural network and matter-element model for fault classification.
• Real-time analysis using dSPACE and an accelerometer sensor.

Main Results:

• Successfully identified three fault statuses: normal, Y collision, and X collision.
• Achieved a diagnosis rate of at least 75% for collision faults.
• Demonstrated the effectiveness of using trajectory diagram characteristics (center of gravity distances, max/min distances) for fault recognition.

Conclusions:

• The proposed method offers an effective and accurate solution for collision fault detection in XXY stages.
• The system's reliance on a single sensor significantly reduces hardware costs.
• This approach provides a foundation for implementing robust and economical fault diagnosis systems.