Health State Identification Method of Nuclear Power Main Circulating Pump Based on EEMD and OQGA-SVM

Abstract

Main circulation pump is the only high-speed rotating equipment in primary loop of nuclear power plant. Its function is to ensure the normal operation of primary loop system by controlling the circulating flow of reactor coolant. In order to ensure long-term healthy operation of nuclear power main circulating pump, a method for identifying the health states of nuclear power main circulating pump based on ensemble empirical mode decomposition (EEMD) and support vector machine optimized by optimized quantum genetic algorithm (OQGA-SVM) is proposed. Vibration signal of main circulating pump is decomposed by EEMD. Vibration signal characteristics of nuclear power main circulating pump in healthy state and different fault states are analyzed and target characteristic indexes are put forward. Then, health state identification model of main circulation pump of OQGA-SVM is established, and target characteristic indexes are used as input parameter of the model. Finally, combined with experimental data, the model analysis and validation show that the health state identification method of nuclear power main circulating pump based on EEMD-OQGA-SVM can accurately and effectively identify the states of main circulation pump, has a higher identification accuracy than EEMD-SVM method and is more efficient and accurate than EEMD-QGA-SVM method.

1. Introduction

Nuclear power main circulating pump is an important piece of equipment in primary loop, and also the only high-speed rotating equipment in primary loop. It ensures the normal operation of primary loop system by controlling the circulating flow of reactor coolant. Circulating pump operates in high temperature and pressure environment for a long time, so it is prone to equipment performance degradation and failure, which threatens the safety and reliability of nuclear power plant and even causes serious accidents. At present, the health state identification of circulating pump is often based on vibration measurement signal analysis. Vibration signal measured with high-temperature accelerometer is analyzed to obtain the characteristic value. Based on characteristic value, fault or different states are judged, or fault is automatically diagnosed and classified by machine learning algorithm.

Pump health state identification research has been ongoing. Literature [1] proposes a three-stage fault diagnosis strategy for multistage centrifugal pumps. The three stages are signal noise reduction, feature extraction and feature dimension reduction. The power flow theory (PFT), which has been proposed to analyze the fluid–solid interaction of underwater structures, provides great potential to process underwater acoustic signals. Literature [2] proposes a novel fault diagnosis method based on PFT, deep convolution processing method (DCPM), and genetic algorithm backpropagation neural network (GANN). Literature [3] proposes a novel combined unscented Kalman filter (UKF) and radial basis function (RBF) method based on an adaptive noise factor for fault diagnosis in the pumping unit. Experimental results show the effectiveness and favorable recognition rate in classifying multiple faults. Literature [4] develops a new data augmentation method, i.e., randomized wavelet expansion (RWE), to generate a set of synthesis samples that share similar characteristics with original sample. Synthesis samples are used as training dataset to train a deep convolutional neural network (CNN) for implementing the few-shot fault diagnosis of aviation hydraulic pumps. In literature [5], the training dataset can be enlarged and the increase in the labeled data could further promote the performance of fault diagnosis. In literature [6], scholars investigate sensor acquired vibration, flow rate, and pressure signals to establish a reliable fault prognostics framework of a gear pump subjected to different fuel contamination levels. A bi-directional long short-term memory (BLSTM) neural network is trained and analyzed. In literature [7], the end-ring wear detection through a multicomponent approach is researched through simulation, laboratory results, and the diagnosis of two-field motors showing that new fault alarm levels need to be defined. Literatures [8,9,10,11,12] present pump health state recognition methods based on data fusion, machine learning and probability. Literature [8] proposes a decision fusion method based on Bayesian probability formula, and obtains the effective evaluation result of pump state. Literature [9] proposes an unsupervised learning algorithm named mixture slow feature analysis (MSFA) to timely evaluate the health states. Literature [10] introduces a multi-sensor prognostics approach which merges highly predictable statistical features from vibrational and pressure sensor measurements. In literature [11], results indicate that neural data fusion method is a reliable non-intrusive diagnostic motor testing under normal loading. A multi-sensor data fusion method based on adaptive weighting strategy using analytic hierarchy process algorithm and cross-correlation function fusion algorithm is proposed in literature [12]. In literatures [13,14,15,16], aiming at the problems of data anomalies and high-dimensional data, corresponding state recognition methods are presented. In literature [13], a method is proposed to locate abnormal vibration data based on correlation coefficient. Literature [14] develops a new deep CNN model called a multireceptive field denoising residual convolutional network (MF-DRCN). In literature [15], a fault detection method based on support vector data description (SVDD) is proposed. In literature [16], the influence of high-amplitude blade pass frequency (BPF) vibration on rotor fault detection is analyzed. Literatures [17,18,19,20,21] use single variable analysis or method migration for state identification and fault diagnosis. The purpose of literature [17] is to examine whether Motor current signature analysis (MCSA) can be used to detect faults in a wet-rotor pump. Literature [18] realizes pump fault diagnosis using load torque characteristic analysis. In literature [19], a time-frequency signal analysis method based on the cyclostationary theory is adopted. In literature [20], a Failure Modes and Effects Analysis (FMEA) is used to analyze typical motor control designs with the intent of determining a means of preventing a runaway condition due to failure of any one component. Literature [21] develops the Matlab Simulink model for the identification of bearing fault in induction motor-based pumping systems. Literatures [22,23,24,25,26,27] use the improved machine learning algorithm to identify and diagnose the states of pump loads. Literature [22] proposes an online fault diagnosis method based on a deep transfer convolutional neural network (TCNN) framework. In literature [23], a new CNN based on LeNet-5 is proposed for fault diagnosis. Literature [24] proposes a model based on the ResNet-34 residual network to identify the indicator diagrams. In literature [25], a family of temporal aware Variational Auto-Encoders is proposed in conjunction with a novel semi-supervised training scheme for pump fault diagnosis. In literature [26], a physics-informed domain adaptation network, termed Adaptive Fault attention Residual Network (AFARN), is proposed. Literature [27] presents a new architecture of an artificial neural network (ANN) for detecting the existence of an internal leakage fault as labelled data.

Pump fault diagnosis or state identification by establishing dynamic models or a multiparameter models is described in literatures [28,29,30,31]. Literature [28] discusses a parametric approach on a steady state for fault detection on pumps. Literature [29] presents a method to estimate the remaining useful life (RUL) of multiple-component systems based on health monitoring information regarding each component in the system under consideration. In literature [30], a lumped parameter electro-mechanical model for the Electric submersible pump (ESPs) system is developed to observe its dynamics. The pump dynamics are estimated using a system identification approach in literature [31].

Pump fault diagnosis and state recognition mainly focus on the following aspects: (1) pump state recognition by overcoming data anomalies and high dimensional data; (2) innovation in machine learning to identify pump states; (3) pump fault diagnosis or state identification by establishing dynamic models or a multiparameter models; (4) single variable analysis or method migration for pump state identification; (5) pump state recognition based on data fusion and probability.

However, there are few studies on the state identification of nuclear power main circulating pump. Based on this, this paper proposes a method of state identification of main circulation pump. The health state identification method of main circulating pump proposed in this paper is mainly realized by the following three steps:

(1)

Based on vibration signal analysis, the vibration characteristics of main circulation pump in different states are extracted. First, vibration signals are analyzed by EEMD, and then the energy value of IMF components is calculated. Based on the change of the energy characteristics of each IMF component, a target feature index is proposed.

(2)

Based on target characteristic index, the health state identification model of the main circulation pump of OQGA-SVM is established. In order to improve the accuracy of state identification, an adaptive selection of SVM parameters is performed using OQGA algorithm.

(3)

The EEMD-OQGA-SVM state identification model is analyzed and validated with experimental data, and results are compared with previous method.

The rest of the paper is organized as follows. Section 2 describes target characteristic index calculation of main pump state. Section 3 demonstrates health state identification model of main pump. Section 4 analyzes experimental results and discusses them. Section 5 includes the research conclusions of this study.

2. Target Characteristic Index Calculation of Main Pump State

2.1. EEMD Analysis of Vibration Signal and of Main Pump

Ensemble empirical mode decomposition (EEMD) [32] is an improvement on the empirical mode decomposition (EMD) by addition of white noise components to original signal, and then decomposition and averaging. Its advantage is to overcome the problem of mode aliasing in EMD. The vibration signal collected from the original equipment often contains main signal components, background noise, etc. It is a relatively complex nonlinear signal. Complex signal is decomposed into the sum of several intrinsic modal functions (IMFS) by EEMD. Different intrinsic modal functions contain different frequency components. Analyzing characteristics of IMF component change during the operation of main pump can help to identify the health states of main pump.

Assuming that the vibration signal of main pump is s(t), when white noise is added, the mixed signal is

$$s_i(t)=s(t)+e_i(t),$$

(1)

where e_i(t) is noise signal and i = 1, 2, 3…, q, q is the average number of times at the beginning of decomposition.

The above mixed vibration signals are decomposed by EMD and the sum of a series of IMF components is obtained.

$$s_i(t)={\displaystyle \sum _{n=1}^N{x}_{in}}(t)+d_i(t),$$

(2)

where N is the number of decomposed IMF, d_i(t) is the residue after the first decomposition. x_in(t) is the IMF component that is decomposed after adding white noise for the i-th time.

The above two steps need to be repeated continuously, and new white noise is added each time until decomposition number reaches q times, resulting in the set of components x_1n(t), x_2n(t), x_3n(t),…, x_qn(t) after each decomposition. Final IMF component is obtained by averaging each set of IMF components to eliminate the adverse effects of white noise.

$$x_n(t)=\frac{1}{q}{\displaystyle \sum _{i=1}^q{x}_{in}}(t),$$

(3)

where x_n (t) is the n-th IMF component after decomposition.

The final IMF component of main pump vibration signal is obtained from the above analysis.

$$s(t)={\displaystyle \sum _{n=1}^N{x}_n}(t)+e(t),$$

(4)

where e(t) is residual term.

Figure 1 shows the time domain waveform of vibration signal when main pump blade is abnormal. Figure 2 shows the EEMD decomposition result of vibration signal when main pump blade is abnormal. Figure 3 shows the time domain waveform of vibration signal when main pump rotor system is abnormal. Figure 4 shows the EEMD decomposition result of vibration signal when main pump rotor system is abnormal. Figure 5 shows the time domain waveform of vibration signal when main pump is normal. Figure 6 shows the EEMD decomposition result of vibration signal when main pump is normal. The decomposition results consist of several IMF components and residuals. Generally, the energy in each state is concentrated in the first several IMF components, so the first several IMF components are selected for analysis. In this paper, IMF1~IMF11 is selected for analysis. Similarly, the IMF components of main pump in normal and abnormal blade conditions can be obtained by the above methods.

2.2. Analysis of Main Pump State Target Characteristics

The target characteristics of main pump state are the premise of the identification of main pump health state. Therefore, the frequency domain characteristics of main pump under health state, rotor system fault and blade abnormal vibration state are analyzed first. In order to obtain different frequency domain characteristics of different states, the energy of IMF component is calculated, and then state target characteristics are obtained. Main pump target characteristics are used as the input parameter of main pump health state recognition model.

In different states, the frequency domain components of its vibration signal are different. When the rotor system is in fault state of main pump, frequency domain components included in the vibration signal of main pump include rotation frequency signal, 3 times rotation frequency signal, 5 times rotation frequency signal, 7 times rotation frequency signal etc.; when the blades of main pump vibrate abnormally, the frequency domain components contained in vibration signal are mainly the blade passing frequency and its harmonics, i.e., p × 50 Hz, 2 × p × 50 Hz, 3 × p × 50 Hz, 4 × p × 50 Hz, 5 × p × 50 Hz, etc.; p is the number of blades in impeller. When main pump is in different health states, it contains different frequency components. Different IMF components contain different frequency components, so the input parameters of the main pump state identification model can be obtained by analyzing the characteristics of IMF components in each state.

After the original vibration signal of main pump is decomposed by EEMD, IMF components are obtained, and energy of each IMF component is calculated.

$$E_n={\displaystyle \underset{0}{\overset{T}{\int }}\left|x_n\right|}dt={\displaystyle \sum _{j=0}^k\left|h_{nj}\right|},$$

(5)

where T is sampling time, n represents the n-th IMF component, k represents total sampling points and h represents amplitude of each discrete point of each component.

The energy ratio for each IMF component is R_IMFn.

$$R_{IMFn}=\frac{E_n}{E_{all}},$$

(6)

where n represents the n-th IMF component and E_all represents the total energy.

$$E_{all}={\displaystyle \sum _{n=1}^N{E}_n}.$$

(7)

In order to identify main pump state more accurately, energy deviation feature quantity ε_n is proposed.

$${\epsilon }_n=\left|R_{IMFn}-R_{IMFn}^{\ast }\right|,$$

(8)

where $R_{IMFn}^{\ast }$ is the energy ratio of each IMF component under the healthy condition of main pump.

The input parameters of main pump health states identification model are obtained.

$$x=\{{\epsilon }_1, {\epsilon }_2, {\epsilon }_3,\dots , {\epsilon }_n\}.$$

(9)

Figure 7 shows the R_IMFn of vibration signal when main pump rotor system is abnormal. Similarly, R_IMFn of main pump in normal and abnormal blade condition can be obtained by the above methods.

The main research content of this part is target characteristic index calculation of main pump state. The next part of the study is mainly about health state identification model of main pump. Target characteristic index is input variable of health state identification model of main pump.

3. Health State Identification Model of Main Pump

Figure 8 is the overall block diagram of main pump health state identification. It can be seen that energy deviation characteristic quantity is the input parameter of health state identification model. Main pump health state identification model is established by optimized quantum genetic algorithm (OQGA) optimized support vector machine (SVM).

3.1. Recognition Model of Main Pump Health States

SVM [33] has obvious advantages in classification, especially in solving small sample classification problems and non-linear classification problems. The energy deviation characteristic of the IMF component of main pump vibration signal is input to the model, that is, the input is x = {ε₁, ε₂, ε₃, …, ε_n}.

The basis of support vector machine (SVM) classification is to find a classification hyperplane. Based on the principle of minimizing structural risk, SVM searches for a hyperplane wψ (x) + b = 0 in high space. If the hyperplane can optimally classify the health states of main pump, then the optimal classification problem can be converted into the following objective function and constraints:

$$\{\begin{array}{l}\min \frac{1}{2}{\Vert w\Vert }^2\hfill \\ \begin{array}{ccc}\mathrm{s}.\mathrm{t}.& y_i(w{\psi }_i+b)\ge 1& i=1,2,\dots ,n\end{array}\hfill \end{array},$$

(10)

where w is weight vector and b is the offset vector.

Introducing relaxation factor ξ_i(i = 1, 2, 3,…, n) can make the identification of the health states of main pump more accurate, so the above optimization problem is converted into the following problem:

$$\{\begin{array}{ll}\min \frac{1}{2}{\Vert w\Vert }^2+C{\displaystyle \sum _{i=1}^n{\xi }_i}\hfill & {\xi }_i\ge 0\hfill \\ s.t.\{\begin{array}{l}{y}_i(w{\psi }_i+b)\ge 1-{\xi }_i\hfill \\ C\ge 0\hfill \end{array}\hfill & i=1,2,\dots ,n\hfill \end{array},$$

(11)

where C is penalty factor.

After iteration, the following decision function for classifying the health states of main pump is obtained:

$$f(x)=sign({\displaystyle \sum _{i=1}^n{\alpha }_{{}_i}^{\ast }}{y}_{i}K(x_i,x_j)+b),$$

(12)

where ${\alpha }_{{}_i}^{\ast }$ is a Lagrange multiplier and K(x_i, x_j) is a kernel function.

The common kernel functions used in SVM are sigmoid kernel function, radial basis kernel function, polynomial kernel function and linear kernel function. In this paper, SVM is used to identify the health states of main pump using radial basis functions. The RBF function is as follows:

$$K(x_i,x_j)=\exp (-\gamma {\Vert x_i-x_j\Vert }^2),$$

(13)

where γ is kernel parameter.

Through the above research, it is found that parameters to be determined in SVM model are penalty factor C and kernel function parameters γ. The values of these two parameters directly affect the accuracy of model. In order to optimize the performance of SVM model in identifying main pump state, the penalty factor C and kernel parameter γ of SVM model are determined by optimized quantum genetic algorithm to avoid blindness in choosing parameters artificially. The next section details the optimized quantum genetic algorithm.

3.2. OQGA of Health State Identification Model

3.2.1. QGA Principle

The combination of quantum computation and genetic algorithm produces the quantum genetic algorithm, which is a probability-based evolutionary algorithm. Quantum genetic algorithm [34] is used to adaptively select the parameters of support vector machine (SVM) to improve the identification accuracy of the health state model of main pump.

Assuming the initial population is P = ($p_1^x$, $p_2^x$, $p_3^x$, …, $p_n^x$) where n is the number of chromosomes, $p_i^x$ indicates that the i-th chromosome in population has evolved to generation x. The SVM parameter is considered as one of the chromosomes.

$$p_i^x=\left[\begin{array}{ccccc}{\alpha }_1^x& {\alpha }_2^x& {\alpha }_3^x& \dots & {\beta }_r^x\\ {\beta }_1^x& {\beta }_2^x& {\beta }_3^x& \dots & {\beta }_r^x\end{array}\right].$$

(14)

In the above formula, (α, β) represents the state of a quantum bit, and r is the length of a quantum chromosome. A quantum bit can be expressed as

$$|\phi \rangle =\alpha |0\rangle +\beta |1\rangle .$$

(15)

Each individual in initial population is binary measured, and initial population is transformed into a population O(0) composed of binary strings. The fitness of the obtained binary population O(0) is evaluated, that is, the randomly generated SVM model parameters are converted from binary to decimal, and then each SVM model parameter value is input to SVM to classification. The classification results are compared, and the SVM model parameter under the best classification effect is retained as the best individual.

By operating quantum gate to update the initial population P(0), a new population P(1) is obtained. By selecting quantum rotary gate, the search is carried out in the direction of the optimal solution as follows:

$$\left[\begin{array}{c}{\alpha }_i^{x+1}\\ {\beta }_i^{x+1}\end{array}\right]=\mu ({\theta }_i)\left[\begin{array}{c}{\alpha }_i^x\\ {\beta }_i^x\end{array}\right],$$

(16)

where μ(θ_i) is quantum rotary gate operation, θ_i is the angle of rotation.

$$\mu ({\theta }_i)=\left[\begin{array}{cc}\cos {\theta }_i& -\sin {\theta }_i\\ \sin {\theta }_i& \cos {\theta }_i\end{array}\right].$$

(17)

According to above steps, continuous population updating is carried out to obtain the optimal parameters of support vector machine.

3.2.2. OQGA Principle

The update operation of quantum genetic algorithm relies on the operation of quantum rotating gate. The operation of quantum rotating gate is an important part of quantum genetic algorithm, which directly affects the convergence speed of the algorithm. If rotation angle is not appropriate, convergence speed of the algorithm will be too slow or premature. The rotation angle of quantum rotary gate of traditional quantum genetic algorithm is fixed, which is not conducive to the rapid convergence of algorithm. Therefore, we propose an optimized quantum genetic algorithm (OQGA) to automatically select the rotation angle of quantum rotating gate and make the algorithm converge quickly.

In order to identify the health state of main pump more accurately, the rotation angle in the quantum genetic algorithm is adjusted dynamically based on the accuracy of health state identification model of main pump. Defining the expected recognition accuracy of main pump health states identification model is ξ^∗ and the expected recognition accuracy can be considered as 100%. In current individual, the actual recognition rate of main pump health state identification model is defined as ξ. If the deviation between ξ and ξ^∗ is large, rotation angle is increased; if the deviation between ξ and ξ^∗ is small, rotation angle is decreased. The adjustment strategy of rotation angle is as follows:

$$\mathsf{\Delta }(\theta )=\frac{\left|\xi -{\xi }^{*}\right|}{{\xi }^{*}}\cdot {\theta }_f,$$

(18)

where θ_f is the reference value of the rotation angle. The value of θ_f depends on the value range of rotation angle and the actual health state recognition effect.

3.2.3. Fitness Function of Genetic Algorithm

The accuracy of main pump state identification is as high as possible. In order to judge the result of main pump state identification, a fitness function fit = (v/v_all) × 100% is proposed. The v is the number of correctly classified samples, the v_all is total number of samples. When the EEMD-OQGA-SVM model identifies the states of main pump, the higher fitness function value, the better health state identification model of the main pump. OQGA is used to optimize model parameters and calculate fitness of each individual. The best fitness value corresponds to the final state identification result.

This part mainly introduces health state identification model of main pump, including EEMD-OQGA-SVM and EEMD-QGA-SVM. In Section 4, we will analyze the results of proposed method for identifying the health states of main pump.

4. Analysis of Experimental Results

Main circulating pump in the experiment is a shielded pump with 5 vanes and main pump speed is 3000 r/min. Data of each state of main pump are collected at 260 °C and 13.7 Mp. Figure 9 is experimental data collection system. It can be seen that the main components include upper computer, industrial computer, and data picking box. Figure 10 is an experimental sample of the main circulation pump.

The operating data of the main pump in each state comes from different operating periods of the main pump, and these data are the vibration signals collected in each state. The experimental sample data can be divided into three categories: the first is the normal state sample data of main pump, the second is the abnormal sample data of main pump rotor system, and the third is the abnormal vibration sample data of main pump blade. The main pump normal state sample data label is 1, the main pump blade abnormal vibration sample data label is 2, and the main pump rotor system abnormal sample data label is 3. Details are given in

Table 1. Sample number and type of main pump states.

Data Sample Type	Number of Samples	Number of Training Samples	Number of Test Samples	Sample Label
Main pump in normal state	119	41	78	1
Abnormal state of blade	193	51	91	2
Abnormal state of rotor system	111	45	66	3

.

Figure 11 is the iterative calculation process diagram of main pump health state identification based on EEMD-OQGA-SVM. It can be seen that after 13 iterations, the identification accuracy of health states of main pump reaches the maximum. Figure 12 is the iterative calculation process diagram of main pump health state identification based on EEMD-QGA-SVM. It can be seen that after 33 iterations, the identification accuracy of health states of main pump reaches the maximum. Therefore, we believe that EEMD-OQGA-SVM algorithm is more efficient in state recognition than EEMD-QGA-SVM algorithm.

Figure 13 is the result diagram of main pump health state identification based on EEMD-OQGA-SVM. The diagram shows the state recognition results of 235 state samples of main circulation pump by the EEMD-OQGA-SVM algorithm. Figure 14 is the result diagram of main pump health state identification based on EEMD-QGA-SVM. The diagram shows the state recognition results of 235 state samples of main circulation pump by the EEMD-QGA-SVM algorithm. Figure 15 is the result diagram of main pump health state identification based on EEMD-SVM. The diagram shows the state recognition results of 235 state samples of main circulation pump by the EEMD-SVM algorithm. By comparing Figure 13, Figure 14 and Figure 15, it can be seen that the accuracy of main circulating pump state recognition algorithm based on EEMD-OQGA-SVM is higher than that of main circulating pump state recognition algorithm based on EEMD-QGA-SVM and EEMD-SVM. The data of main pump in various states are monitored during long-term operation. These data come from different periods of operation of main pump. When main pump is in different operation periods, its components of vibration signal are different. The main components of vibration signal of main pump are vibration signal and noise signal. When main pump is in different periods, it bears different noise signals. These noise signals may come from other experimental devices, such as vibration table, supercritical carbon dioxide device, high-power variable frequency power supply, etc. Therefore, due to the interference of a lot of noise, there are some state identification errors in the identification of main pump states.

The accuracy rates and number of iterations of main pump state recognition algorithms based on EEMD-OQGA-SVM, EEMD-QGA-SVM and EEMD-SVM are shown in

Table 2. Comparison of state identification results of main pump.

Method	Total Number of Samples Not Correctly Identified	Number of State 1 Not Correctly Identified	Number of State 2 Not Correctly Identified	Number of State 3 Not Correctly Identified	Recognition Accuracy	Number of Iterations
EEMD-OQGA-SVM	11	0	1	10	95.31	13
EEMD-QGA-SVM	12	0	1	11	94.89	33
EEMD-SVM	28	2	6	20	88.01	/

.

5. Conclusions

This paper presents a method to identify the states of main circulating pump based on EEMD-OQGA-SVM. The vibration signal of main pump is effectively analyzed by using EEMD, and target characteristic index is extracted. The health state identification model OQGA-SVM of main circulation pump is established, and target characteristic index is used as the input parameter of the model. Finally, the model analysis is combined with the experimental data, and the results show that the proposed method can accurately and effectively identify the different states of main circulation pump, and has a higher recognition accuracy than the EEMD-SVM and is more efficient and accurate than EEMD-QGA-SVM method. Similar to the latest research results [35,36], the state identification model of main pump proposed in this paper is based on analysis of operating data characteristics of research objects under specific background, and then this paper proposes a new method based on the optimization of existing algorithms, and improves the efficiency and accuracy. However, these algorithms are all realized under certain backgrounds. Therefore, it is our research objective to further improve the universality of the methods proposed in this paper, for example, further study of this method will enable it to be directly applied to Nuclear-Grade valves or reactor control rod drive mechanisms without change or with little change.

(1)

The method proposed in this paper combines the advantages of quantum computation and uses optimized quantum genetic algorithm to automatically select the parameters of SVM.

(2)

The EEMD-OQGA-SVM method proposed in this paper has a higher identification accuracy than the EEMD-SVM and is more efficient and accurate than EEMD-QGA-SVM method.

(3)

The EEMD-OQGA-SVM state recognition method proposed in this paper can be used not only for nuclear devices but also for devices in other fields. The premise of this is to re-extract the running characteristics of the new application objects.