Neural networks for control systems: a survey
Automatica (Journal of IFAC)
Optimal combinations of pattern classifiers
Pattern Recognition Letters
Hi-index | 0.00 |
To compensate the flaws of single neural network such as low classification precision and long training time as well as weak generalized ability etc, the paper proposes a novel ensemble neural classifier for stream turbine vibration fault diagnosis with improved PSO-based. The method fully utilizes the advantages of PSO, such as fast seeking speed and easily realizing mode etc, the integrated time of the whole network therefore becomes very short. Meanwhile, more neural networks are used to implement fault diagnosis concurrently, and their results are integrated with improved PSO-based. The studies indicate that the proposed method has higher precision of the classification and seekink speed, and is an ideal pattern classifier. In the end, a simulation experiment in stream turbine vibration fault diagnosis shows the method is extremely effective.