A Deep Learning Approach to Detect Anomalies in an Electric Power Steering System.

Area of Science:

• Automotive Engineering
• Artificial Intelligence
• Machine Learning

Background:

• Electrical Power Steering (EPS) systems are increasingly complex, demanding advanced safety and reliability.
• Current anomaly detection methods struggle with novel or unknown system anomalies due to reliance on prior knowledge.

Purpose of the Study:

• To propose a novel deep learning approach for enhanced anomaly detection in EPS sensor data.
• To address the limitations of existing methods in identifying previously unknown anomalies.

Main Methods:

• A two-stage deep learning model combining an autoencoder and Long Short-Term Memory (LSTM) network.
• Autoencoder for feature extraction and dimensionality reduction.
• LSTM for capturing temporal dependencies and reconstructing data for anomaly scoring based on reconstruction loss.

Main Results:

• The proposed model achieved high accuracy (0.99) in anomaly detection.
• Demonstrated superior performance compared to existing methods, evidenced by a higher Area Under the Receiver Operating Characteristic curve.
• Effectiveness validated through experimental data collected using an EPS test jig.

Conclusions:

• The deep learning approach offers a valuable tool for robust anomaly detection in EPS systems.
• The autoencoder-LSTM model effectively identifies both known and unknown anomalies.
• The method significantly improves reliability and safety in complex EPS environments.