CONDITION MONITORING OF CENTRIFUGAL PUMP IN NUCLEAR POWER PLANT BASED ON IMPROVED VMD AND SVM

抄録

Aiming at the lack of data on centrifugal pumps in nuclear power plants and the inability to provide data support for the subsequent development of centrifugal pump status monitoring systems in nuclear power plants, this paper designs and builds a centrifugal pump test bench. On this basis, the impeller cavitation wear fault of the centrifugal pump is set up, and the acceleration signal of the normal state and the fault state of the centrifugal pump is collected by the acceleration sensor. Since the collected acceleration signal will be mixed with a certain amount of environmental noise, which will affect the monitoring effect, it is necessary to denoise the signal before performing feature extraction on the acceleration signal. This paper first aims at the difficulty of selecting the K value and α penalty factor of the Variational Mode Decomposition (VMD) algorithm, proposes to use the Whale Optimization Algorithm (WOA), and uses the envelope entropy as the fitness function to optimize the VMD model. On the basis of the above work, the collected acceleration signal is decomposed, the environmental noise is eliminated, and the time domain characteristic parameters are extracted from the recombined signal to complete the normalization process of the time domain characteristic parameters. Then the preprocessed time-domain characteristic parameters obtained by calculation are brought into the Support Vector Machines (SVM) model to complete the model training, and then complete the development of the nuclear power plant centrifugal pump condition monitoring system. The final experimental results show that the nuclear power plant centrifugal pump condition monitoring system developed in this paper has a high monitoring accuracy.