LSTM-Autoencoder for Vibration Anomaly Detection in Vertical Carousel Storage and Retrieval System (VCSRS).

Area of Science:

• Manufacturing Technology
• Industrial Engineering
• Machine Learning

Background:

• Industry 5.0, or smart factories, leverages advanced data analytics for process optimization.
• Vibration data analysis is crucial for monitoring machinery and detecting anomalies.
• Vertical Carousel Storage and Retrieval Systems (VCSRS) require effective anomaly detection for operational integrity.

Purpose of the Study:

• To optimize sensor placement for accurate anomaly detection in VCSRS.
• To enhance anomaly detection accuracy using machine learning techniques.
• To evaluate the effectiveness of a combined correlation coefficient and LSTM-autoencoder model.

Main Methods:

• Utilized a correlation coefficient model with Fisher Information Matrix (FIM) and Effective Independence (EFI) for sensor placement optimization.
• Employed an LSTM-autoencoder (long short-term memory) model for training and testing vibration data.
• Integrated vibration data analysis with advanced machine learning for anomaly identification.

Main Results:

• Optimized sensor placement for maximum accuracy and reliability in VCSRS.
• Achieved a 97.70% accuracy rate in detecting anomalies within the vertical carousel system.
• Demonstrated the capability of LSTM-autoencoders to identify subtle patterns in vibration data.

Conclusions:

• The combined correlation coefficient and LSTM-autoencoder model significantly enhances anomaly detection in Industry 5.0 manufacturing.
• Optimized sensor placement is critical for reliable performance monitoring of industrial systems like VCSRS.
• Advanced machine learning techniques offer superior anomaly detection compared to traditional methods in smart factories.