Quantum-enhanced LSTM for predictive maintenance in industrial IoT systems

Area of Science:

• Quantum Computing
• Artificial Intelligence
• Industrial Internet of Things (IIoT)

Background:

• Predictive maintenance in IIoT systems is crucial for operational efficiency.
• Analyzing high-dimensional industrial sensor data presents significant challenges.
• Existing methods struggle to preserve temporal relationships in complex time series data.

Purpose of the Study:

• To propose an innovative solution for predictive maintenance in IIoT systems.
• To combine quantum computing capabilities with Long Short-Term Memory (LSTM) neural networks.
• To develop a hybrid quantum-classical architecture for enhanced time series analysis.

Main Methods:

• A hybrid quantum-classical architecture is utilized.
• Quantum computing handles high-dimensional sensor data analysis.
• Quantum channels are designed to minimize temporal dependencies in sensor measurements.
• Classical LSTMs model sequential data, preserving temporal relationships.

Main Results:

• QE-LSTM demonstrated improved F1 scores by 4-5 percentage points on SECOM for bearing failure detection.
• Reduced Root Mean Square Error (RMSE) and NASA Score on C-MAPSS.
• Consistent performance gains were observed on IMMD datasets.

Conclusions:

• The proposed QE-LSTM model offers a novel approach to predictive maintenance in IIoT.
• The hybrid architecture effectively leverages quantum computation for complex industrial data.
• QE-LSTM shows significant improvements in accuracy and efficiency for failure detection tasks.