Elevator fault precursor prediction based on improved LSTM-AE algorithm and TSO-VMD denoising technique.

Area of Science:

• Engineering
• Artificial Intelligence
• Data Science

Background:

• Elevator operation data often suffers from feature redundancy and noise, hindering accurate fault prediction.
• Traditional methods struggle with the complexity and volume of real-time elevator data.
• Predictive maintenance is crucial for ensuring elevator safety and operational efficiency.

Purpose of the Study:

• To develop an advanced algorithm for predicting elevator faults using a novel combination of deep learning and signal processing techniques.
• To address challenges of data noise and feature redundancy in elevator operational data.
• To improve the accuracy and stability of fault precursor prediction for elevators.

Main Methods:

• The proposed VMD-BILSTM-AEAM algorithm integrates Variational Mode Decomposition (VMD), Bidirectional Long Short-Term Memory (BILSTM), and an Autoencoder with an Attention Mechanism (AEAM).
• Attribute Correlation Density Ranking (ACDR) was used for effective feature selection.
• TSO-optimized VMD was employed for data denoising to enhance data quality.

Main Results:

• The VMD-BILSTM-AEAM algorithm achieved a mean True Positive Rate (TPR) of 0.919 (95% CI: 0.915-0.924).
• A mean False Positive Rate (FPR) of 0.090 (95% CI: 0.087-0.092) and a mean Area Under the Curve (AUC) of 0.919 (95% CI: 0.915-0.923) were recorded.
• Performance metrics demonstrated significant improvements over traditional and other deep learning models.

Conclusions:

• The VMD-BILSTM-AEAM algorithm offers superior accuracy and stability for elevator fault precursor prediction.
• The model effectively processes noisy time-series data, showcasing its potential for broader predictive maintenance applications.
• This advanced approach enhances the reliability and safety of elevator systems through improved fault detection.