Related Experiment Video
Updated: Jun 16, 2026
05:30
Large Scale Energy Efficient Sensor Network Routing Using a Quantum Processor Unit
Published on: September 8, 2023
Entropy space quantum behaved dung beetle optimized long short-term memory network for industrial process fault
Shuai Ao1, Bo Ma2, Haorui Liu3
1College of Mechanical and Electrical Engineering, Beijing University of Chemical Technology, Beijing 100029, China; Mechanical and Electrical Management Department, National Energy Group Guoshen Company Sandaogou Coal Mine, Yulin, Shanxi 719407, China.
ISA Transactions
|June 13, 2026
Summary
This study introduces an enhanced LSTM model for industrial fault diagnosis, improving feature extraction and hyperparameter optimization for complex data. The new method achieves high accuracy on benchmark datasets.
Area of Science:
- Artificial Intelligence
- Machine Learning
- Industrial Process Control
Background:
- Traditional Long Short-Term Memory networks (LSTM) struggle with high-dimensional, nonlinear industrial data due to limitations in feature learning and hyperparameter sensitivity.
- Existing methods often fail to capture critical high-order statistical information present in complex process data.
Purpose of the Study:
- To propose a novel hierarchical fault diagnosis method, Entropy Space Quantum Behaved Dung Beetle optimized LSTM (ES-QLSTM), to overcome the limitations of traditional LSTMs.
- To enhance feature extraction and intelligently optimize hyperparameters for complex industrial fault diagnosis.
Main Methods:
- Kernel Entropy Component Analysis (KECA) is used for feature extraction in the entropy space (ES), preserving high-order statistical information.
- A Quantum Behaved Dung Beetle Optimization (QDBO) algorithm is developed to optimize LSTM hyperparameters, addressing empirical setting instability.
- A synergistic framework integrates ES feature extraction with QDBO for dimensionality reduction and hyperparameter optimization.
Main Results:
- The ES-QLSTM method demonstrated superior performance in complex fault diagnosis compared to traditional and improved models.
- Experimental validation on the Tennessee Eastman Process (TEP) and grid-connected photovoltaic system (GPVS) datasets yielded average diagnostic accuracies of 93.70% and 83.91%, respectively.
- The method effectively compensates for insufficient feature mining and unstable performance issues associated with standard LSTMs.
Conclusions:
- The proposed ES-QLSTM method offers a robust solution for fault diagnosis in complex industrial processes.
- The integration of entropy space feature purification and intelligent hyperparameter optimization significantly enhances diagnostic accuracy and stability.
- This approach provides a valuable advancement for intelligent monitoring and control in industrial systems.