Research And Application Of Fault Diagnosis And Residual Life Prediction Technology For Key Active Equipment In The Primary Circuit

Nuclear reactors are widely used in the fields of national defense technology and civil economy,and are important facilities in national development.In the field of national defense technology,power plants with nuclear reactors as the core play a crucial role in national defense construction,which can significantly enhance national defense capabilities;In the field of civil economy,nuclear reactors are important production facilities for clean energy and can make significant contributions to national economic development.The reliable operation of key active equipment in the primary circuit of a reactor is a prerequisite for ensuring the safe operation of nuclear facilities.Once the key active equipment in the primary circuit malfunctions,it will directly lead to reactor shutdown and even cause major accidents.Therefore,this article intends to carry out research and application of fault diagnosis and residual life prediction technology for key active equipment in the reactor primary circuit.The specific research content is arranged as follows:(1) Aiming at the problem that it is difficult to accurately identify the status of active equipment in the primary circuit,a health state recognition method based on Ensemble Empirical Mode Decomposition (EEMD),Optimized Quantum Genetic Algorithm(OQGA),and Support Vector Machine (SVM) is proposed.The operating vibration signals often contain equipment health status information.EEMD is used to decompose the vibration signals into modes,and based on the obtained Intrinsic Mode Functions(IMF),the vibration signal characteristics under normal and different fault states are analyzed.From this,the IMF component energy deviation is used to establish target feature indicators to characterize the equipment status.Then,an OQGA model is established by optimizing gene crossover operations and quantum rotation gate operations.By utilizing the parameter optimization ability of OQGA,SVM is optimized to establish a health status recognition model based on OQGA-SVM,with target feature indicators as input parameters.Finally,method validation and application are conducted,and the proposed method is applied to the health status recognition of the main circulation pump.The final results showed that the health status recognition method based on EEMDOQGA-SVM can accurately and effectively identify the health status of the main circulation pump,and the state recognition results are better than the EEMD-SVM method and EEMD-QGA-SVM method.(2) Aiming at the problems of weak early fault of primary circuit active equipment and difficult to detect the critical point of fault,two fault state critical point identification methods are proposed.The fault critical state identification method based on multiprocess variable fusion analysis is as follows:Firstly,Shannon entropy is used to describe the power spectrum of vibration signal related to equipment operation process,and then the mean value of power spectrum entropy is used as indirect process variable.The difference between the indirect process variables in the health state of the equipment and the indirect process variables in the different fault degree state is further studied.In addition,the power signal is used as the direct process variable,and the characteristics of the direct process variable under normal state and different fault degree are studied.By analyzing the changes of two process variables before and after the fault,the identification method of fault critical point is established.The fault critical state identification method based on signal periodic characteristic frequency extraction is as follows:Firstly,the running vibration signal is denoised based on double tree complex wavelet transform to highlight the periodic characteristics of time domain signal;Secondly,the vibration signal after noise reduction is preprocessed,and the vibration signal is clipped to obtain the vibration signal of stable operation.Then the envelope curve of primary active device operating vibration signal is extracted based on cubic spline interpolation.Then the local peak sequence is calculated,and the reference periodic frequency interval is determined according to the local peak sequence,so as to analyze the fault critical characteristics,obtain the fault critical characteristic deviation value,and determine the fault critical point according to the fault critical characteristic deviation value.Finally,the method is verified and applied,and the two methods are applied to the identification of the fault critical point of the nuclear gate valve.The results show that the two methods can effectively identify the fault critical point of the nuclear gate valve in the case of small samples.(3) Aiming at the problem of weak early faults and high background noise in primary circuit active equipment,as well as difficulty in extracting fault frequencies,an incipient fault feature recognition method is proposed based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN),Target Intrinsic Mode Functions (TIMF) and Amplitude Frequency Feature Enhancement (AFFEHT).The proposed method can overcome the difficulty of extracting fault characteristic frequency caused by background noise interference and weak early fault signal.Firstly,CEEMDAN is used to perform mode decomposition of the vibration signal under early faults to obtain multiple IMF components.Secondly,the spectrum of all IMF components is analyzed,and the TIME components are selected according to the spectrum characteristics.Then,the AFFEHT method is proposed to filter the interference frequency of TIMF component,and highlight the amplitude characteristics of fault frequency.Then,the ampliturefrequency characteristic enhanced Hilbert spectrum of TIMF component is obtained to analyze the early fault characteristics of primary circuit active equipment.Finally,the method was verified and applied.After simulation analysis,the proposed method was applied to the early fault diagnosis of the main circulation pump.The results show that the performance of CEEMDAN-TIME-AFFEHT method is outstanding,which is better than CEEMDAN-TIME-HT,HT methods and Short-Time Fourier transform.(4) In order to solve the problem that it is difficult to accurately predict the residual life of active equipment in the primary circuit,the residual life prediction method based on Complete Ensemble Empirical Mode Decomposition with Adaptive Correction Noise(CEEMDACN) and Optimal Quantified Health Indicators (OQHI) is proposed.Firstly,according to the structural characteristics of the equipment and the degradation law,the formation principle of the degradation characteristic frequency (target characteristic frequency) is analyzed.On this basis,the white noise intensity is modified adaptively according to the target characteristic frequency to obtain the CEEMDACN method.Secondly,combined with the target characteristic frequency,the CEEMDACN method and spectrum analysis is used to denoise and reconstruct the vibration signal,and the denoised reconstructed signal containing the target characteristic frequency to the maximum extent is obtained.Then,the multi-feature extraction is performed on the denoised reconstructed signal,and the multi-quantitative health indicators are established and smoothed by using the Moving-Average (MA).Finally,the OQHI is obtained based on the monotonicity evaluation index and the linear evaluation index.Furthermore,based on the extracted OQHI curve,the residual life prediction model based on polynomial is established.Finally,the method is verified and applied based on the ultimate life experiment of Control Rod Drive Mechanism (CRDM).The results show that the proposed life prediction method is effective and has significant performance.