Real-time fault detection for IIoT facilities using GA-Att-LSTM based on edge-cloud collaboration

Area of Science:

• Industrial Internet of Things (IIoT)
• Machine Learning
• Data Science

Background:

• Industrial Internet of Things (IIoT) devices generate vast amounts of spatiotemporally correlated and heterogeneous sensor data.
• Current anomaly detection algorithms face challenges processing this complex, large-scale data.
• The need for efficient and accurate anomaly detection in IIoT facilities is critical.

Purpose of the Study:

• To propose an improved anomaly detection framework for Industrial Internet of Things (IIoT) facilities.
• To enhance the processing of large-scale, complex sensor data.
• To improve the accuracy and efficiency of fault detection in IIoT systems.

Main Methods:

• Developed a Genetic Algorithm-Attention-LSTM (GA-Att-LSTM) framework.
• Implemented an edge-cloud collaboration architecture for real-time data processing.
• Integrated an attention mechanism to focus on critical features and a genetic algorithm for hyperparameter optimization.

Main Results:

• Achieved high performance on a public fault database: 99.6% accuracy, 84.2% F1-score, 89.8% precision, and 77.6% recall.
• Demonstrated superior performance compared to five traditional machine learning methods.
• The edge-cloud collaboration reduced data uploading time to the cloud platform.

Conclusions:

• The proposed GA-Att-LSTM framework effectively detects anomalies in IIoT facilities.
• Edge-cloud collaboration and advanced deep learning techniques significantly enhance anomaly detection capabilities.
• The method offers a robust solution for real-time fault detection in industrial environments.