Evaluation Method for CRDM Roller State Based on Evaluation Function and BP Network

In this paper, an evaluation method for CRDM roller state based on the evaluation function and error back propagation training (BP) network is proposed for the non-stationary and strong noise distortion signals in the control rod drive mechanism (CRDM) vibration signals. The signal is denoised by semi-soft threshold, the feature vectors are extracted by local mean decomposition (LMD), and the sample set composed of feature vectors is identified by BP network for state recognition. An evaluation function is introduced to evaluate the results of state recognition. The distorted samples are removed according to the evaluation results, and the new sample set is retained for state recognition. The results show that this method can effectively identify the defect state of the rollers and effectively solve the difficulty in accurate identification of the rollers state of the control rod drive mechanism.