A Robust Multivariate Time Series Classification Approach Based on Topological Data Analysis for Channel Fault...

Area of Science:

• Artificial Intelligence
• Machine Learning
• Data Science
• Mechanical Engineering
• Industrial IoT

Background:

• Vibration monitoring of rotating equipment is crucial for reliable industrial operations in manufacturing, power generation, and aerospace.
• AI-based predictive maintenance relies heavily on complete and reliable sensor data for anomaly detection.
• Sensor data loss, especially in multi-sensor systems, significantly degrades AI model performance and reduces predictive maintenance reliability.

Purpose of the Study:

• To develop a robust artificial intelligence (AI) model for vibration monitoring of rotating equipment.
• To address the challenge of sensor data loss in AI-based predictive maintenance systems.
• To enhance the reliability and efficiency of maintenance strategies through improved AI model performance.

Main Methods:

• Introduction of topological data analysis (TDA) to create a robust AI model.
• TDA analyzes the topological structure of sensor data to generate consistent feature vectors.
• The method ensures stable predictions by capturing intrinsic data characteristics, even with missing sensor channels.

Main Results:

• The proposed AI model demonstrates high performance resilience under partial sensor data loss.
• Consistent feature vectors generated by TDA maintain predictive accuracy despite missing data.
• The model effectively supports reliable operation of rotating equipment across various industries.

Conclusions:

• The TDA-based AI model significantly enhances the reliability of AI-driven predictive maintenance systems.
• This approach mitigates the impact of sensor failures, ensuring more dependable equipment monitoring.
• The study contributes to establishing more efficient and robust industrial maintenance strategies.