Nuclear power last-stage long blade fault diagnosis method and system

Abstract

The invention discloses a nuclear power last-stage long blade fault diagnosis method and system, and relates to a mechanical part testing method and device. The method aims at solving the problem that in an existing fault detection method, specific fault types and fault degrees cannot be accurately obtained. The method comprises the following steps that the actual dynamic response of the nuclear power last-stage long blade is collected; performing classification prediction on the actual dynamic response through a classifier, and identifying to obtain a fault type; the fault type comprises a water erosion fault and a crack fault; selecting a corresponding fitting coefficient based on the fault type, establishing a solving objective function through the fitting coefficient, and calculating to obtain a fault parameter; the fault fitting coefficient is preset in a fault diagnosis database; the fault parameters comprise the water erosion length and the water erosion depth generated by the water erosion fault and / or the crack length generated by the crack fault.

Description

Technical Field

[0001] The invention relates to a testing method and device for mechanical parts. Background Art

[0002] Faults such as fracture and water erosion seriously affect the safety of nuclear power unit equipment operation. The existing fault detection method is to detect and calculate measurement values ​​such as frequency and displacement. When the measurement value deviation is large, it is considered that the current blade has a fault and then an alarm is issued. However, only whether the blade has a fault can be obtained from the alarm information, and the specific fault type and fault degree cannot be accurately obtained. Summary of the invention

[0003] The purpose of the present invention is to overcome the problem that the existing fault detection methods cannot accurately obtain the specific fault type and fault degree, and to provide a nuclear power last stage long blade fault diagnosis method and system.

[0004] The method for diagnosing faults of nuclear power last stage long blades of the present invention is as follows:

[0005] Collect the actual dynamic response of the long blades in the last stage of nuclear power;

[0006] The actual dynamic response includes the actual first-order frequency and the actual first-order frequency response component of the displacement at the middle position of the top of the nuclear power last stage long blade;

[0007] The actual dynamic response is classified and predicted by the classifier to identify the fault type;

[0008] Failure types include water erosion failure and crack failure;

[0009] Select the corresponding fitting coefficient based on the fault type, and establish the objective function through the fitting coefficient to calculate the fault parameters;

[0010] The fault fitting coefficients are preset in the fault diagnosis database;

[0011] The fault parameters include the water erosion length and water erosion depth caused by water erosion faults, and / or the crack length caused by crack faults.

[0012] Furthermore, when the fault type is identified as a water erosion fault, a corresponding fitting coefficient is selected based on the fault type, and a target function is established through the fitting coefficient to calculate the fault parameters, including:

[0013] Select the fitting coefficients a1, b1, c1, d1, f1, g1, h1, a2, b2, c2, d2, f2, g2 and h2 of water erosion fault, establish the target functions X and Y for solving water erosion fault, and calculate the water erosion length x and water erosion depth y;

[0014] X=F-(a1x2 +b1x+c1y 2 +d1y+e1+f1x 2 y+g1xy 2 +h1xy)

[0015] Y=U-(a2x 2 +b2x+c2y 2 +d2y+e2+f2x 2 y+g2xy 2 +h2xy)

[0016] Among them, the error range of solving the objective functions X and Y is 1e-20 to 1e-15; F is the actual first-order frequency; U is the actual first-order frequency response component.

[0017] Furthermore, when the fault type is a crack fault, a corresponding fitting coefficient is selected based on the fault type, and a target function is established by using the fitting coefficient to calculate the fault parameters, including:

[0018] Select the fitting coefficients g1, g2, g3, k1, k2, k3 and k4 of the crack fault, establish the target function Z for solving the crack fault, and calculate the crack length z;

[0019] Z=(F-(g1z 2 +g2z+g3)) / s+(U(z)-(k1z 4 +k2z 3 +k3z 2 +k4z+k5))

[0020] Among them, s is the adjustment coefficient, and its value range is 150~250; F is the actual first-order frequency; U is the actual first-order frequency response component.

[0021] Furthermore, it also includes establishing a fault diagnosis database, the process is as follows:

[0022] Fault simulation of the nuclear power last stage long blade based on preset fault parameters;

[0023] Among the fault parameters, the water erosion depth is 0-28mm, the water erosion length is 0-400mm, and the crack length is 0-100mm;

[0024] Obtain the aerodynamic loads of the nuclear power last stage long blades under rated conditions;

[0025] The simulation obtains the dynamic response of the nuclear power last stage long blade under aerodynamic load;

[0026] The simulated dynamic response includes the simulated first-order frequency and the simulated first-order frequency response component of the displacement at the middle position of the top of the last-stage long blade of the nuclear power plant;

[0027] The simulated dynamic response and the preset fault parameters are fitted to obtain the fitting coefficients of the water erosion fault and the fitting coefficients of the crack fault; and the fitting coefficients are set in the fault diagnosis database to complete the establishment of the fault diagnosis database.

[0028] Furthermore, the simulation dynamic response and the preset fault parameters are fitted to obtain the fitting coefficient of the water erosion fault as follows:

[0029] The simulated first-order frequency, the simulated first-order frequency response component and the variation of the simulated first-order frequency with the water erosion length are fitted by the least squares surface:

[0030] F'(x,y)=a1x 2 +b1s+c1y 2 +d1y+e1+f1x 2 y+g1xy 2 +h1xy Formula 1

[0031] And the simulated first-order frequency, the simulated first-order frequency response component and the simulated first-order frequency response component changing with the water erosion depth are fitted by the least squares surface:

[0032] U'(x,y)=a2x 2 +b2x+c2y 2 +d2y+e2+f2x 2 y+g2xy 2 +h2xy Formula 2 Then the fitting coefficient of water erosion fault is calculated by formula 1 and formula 2;

[0033] Where, F' is the simulated 1st-order frequency; U' is the simulated 1st-order frequency response component.

[0034] Further, the process of simulating the fault of the nuclear power last stage long blade based on the fault parameters includes:

[0035] On the finite element model of the nuclear power plant's last-stage long blade, the mesh of the erosion fault area is peeled off according to the erosion length and erosion depth;

[0036] The water erosion failure area includes the inlet side area of ​​the airfoil top.

[0037] Furthermore, the simulation dynamic response and the preset fault parameters are fitted to obtain the fitting coefficient of the crack fault as follows:

[0038] The simulated first-order frequency, the simulated first-order frequency response component and the crack length are fitted with a univariate polynomial regression to obtain the quadratic curve of the simulated first-order frequency with respect to the crack length:

[0039] F'(z)=g1z 2+g2z+g3 formula 3

[0040] And, the fourth-order curve of the simulated first-order frequency response component with respect to the crack length is obtained:

[0041] U'(z)=k1z 4 +k2z 3 +k3z 2 +k4z+k5 Formula 4 is then used to calculate the fitting coefficient of crack failure through Formula 3 and Formula 4;

[0042] Where, F' is the simulated 1st-order frequency; U' is the simulated 1st-order frequency response component.

[0043] Furthermore, the process of fault simulation for the nuclear power last stage long blade based on the fault parameters is as follows;

[0044] On the finite element model of the nuclear power last stage long blade, the cracking operation is performed along the grid lines of the cracking fault area according to the crack length;

[0045] The cracking fault region includes at least one of a Stellite alloy region of the airfoil, a steam outlet side region of a first tooth of a blade root, and a steam inlet side region of a first tooth of a blade root.

[0046] Furthermore, a method for obtaining a classifier is also included:

[0047] The pre-collected fault types and the dynamic responses corresponding to the fault types are used as output and input respectively, and the support vector machine is trained to obtain a classifier.

[0048] The present invention also provides a nuclear power last stage long blade fault diagnosis system, comprising:

[0049] Response collection module, fault classification module and fault parameter calculation module:

[0050] Select the corresponding objective function based on the fault type and calculate the fault parameters;

[0051] Response acquisition module, used to collect the actual dynamic response of the nuclear power last stage long blades;

[0052] The actual dynamic response includes the actual first-order frequency and the actual first-order frequency response component of the displacement at the middle position of the top of the nuclear power last stage long blade;

[0053] The fault classification module is used to classify and predict the actual dynamic response through a classifier and identify the fault type;

[0054] Failure types include water erosion failure and crack failure;

[0055] The fault parameter module is used to select the corresponding fitting coefficient based on the fault type, and establish the target function through the fitting coefficient to calculate the fault parameter;

[0056] The fault fitting coefficients are preset in the fault diagnosis database;

[0057] The fault parameters include the water erosion length and water erosion depth caused by water erosion faults, and / or the crack length caused by crack faults.

[0058] The beneficial effects of the present invention are:

[0059] The nuclear power plant's last-stage long blade fault diagnosis method and system of the present invention simulates the water erosion fault and crack fault of the nuclear power plant's last-stage long blade and mines the fault rules, understands the influence of different fault levels on the response of the nuclear power plant's last-stage long blade, uses the discovered rules to guide practice, and better ensures the safe operation of the nuclear power plant's last-stage long blade. It not only helps to reduce the frequency of faults, but also improves the efficiency and accuracy of fault handling, and ensures the safe operation of nuclear power units. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a flow chart of the method for diagnosing faults of nuclear power last stage long blades of the present invention;

[0061] Figure 2 It is a schematic diagram of the change of the first-order frequency with the water erosion depth and water erosion length in the method for diagnosing the fault of the nuclear power last stage long blade and the system of the present invention;

[0062] Figure 3 It is a schematic diagram of the change of the first-order frequency response component with the water erosion depth and water erosion length in the method for diagnosing the fault of the nuclear power last stage long blade and the system of the present invention;

[0063] Figure 4 A schematic diagram of a fault diagnosis method for a nuclear power plant's last stage long blade and crack locations in the system according to the present invention;

[0064] Figure 5 It is a schematic diagram showing the variation of the first-order frequency corresponding to crack case 1 in the nuclear power last stage long blade fault diagnosis method and system of the present invention with the crack length;

[0065] Figure 6 It is a schematic diagram of the change of the first-order frequency response component corresponding to the crack case 1 in the nuclear power last stage long blade fault diagnosis method and system of the present invention with the crack length;

[0066] Figure 7 It is a schematic diagram of the distribution of characteristics of different fault types in two-dimensional space in the nuclear power last stage long blade fault diagnosis method and system of the present invention;

[0067] Figure 8It is a state space schematic diagram of the classification of the classifier in the nuclear power last stage long blade fault diagnosis method and system of the present invention. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0070] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments, but they are not intended to limit the present invention. Specific implementation method 1

[0072] The method for diagnosing a fault of a nuclear power plant's last stage long blade in this embodiment is characterized in that the method is as follows:

[0073] Collect the actual dynamic response of the long blades in the last stage of nuclear power;

[0074] The actual dynamic response includes the actual first-order frequency and the actual first-order frequency response component of the displacement at the middle position of the top of the nuclear power last stage long blade;

[0075] The actual dynamic response is classified and predicted by the classifier to identify the fault type;

[0076] Failure types include water erosion failure and crack failure;

[0077] Select the corresponding fitting coefficient based on the fault type, and establish the objective function through the fitting coefficient to calculate the fault parameters;

[0078] The fault fitting coefficients are preset in the fault diagnosis database;

[0079] The fault parameters include the water erosion length and water erosion depth caused by water erosion faults, and / or the crack length caused by crack faults.

[0080] Specifically, during the operation of the nuclear power plant's last-stage long blades, actual dynamic responses are collected through sensors and other devices, and the actual dynamic responses are input into a pre-trained classifier for classification and prediction. If the classifier determines that the current state of the nuclear power plant's last-stage long blades is a water erosion fault, then the fault parameters at this time are the water erosion depth x, the water erosion length y, and the known parameters are the first-order frequency F and the first-order frequency response component U of the displacement response. Two equations are constructed using preset fault fitting coefficients, and the number of equations is equal to the unknown number. The fault parameters, water erosion depth x and water erosion length y, can be obtained by solving nonlinear equations.

[0081] If the classifier determines that the current state of the nuclear power plant's last-stage long blade is a crack fault, the fault parameter is the crack depth z, and the known parameters are the first-order frequency F and the first-order frequency response component U of the displacement response. An equation is constructed using the preset fault fitting coefficients. The unknown number is less than the number of equations, and the fault parameter can be obtained by solving the equation using the least squares method. Specific implementation method 2

[0083] This embodiment is a further explanation of the first embodiment. In this embodiment, when the fault type is identified as a water erosion fault, a corresponding fitting coefficient is selected based on the fault type, and a target function is established by the fitting coefficient to calculate the fault parameters, including:

[0084] Select the fitting coefficients a1, b1, c1, d1, f1, g1, h1, a2, b2, c2, d2, f2, g2 and h2 of water erosion fault, establish the target functions X and Y for solving water erosion fault, and calculate the water erosion length x and water erosion depth y;

[0085] X=F-(a1x 2 +b1x+c1y 2 +d1y+e1+f1x 2 y+g1xy 2 +h1xy)

[0086] Y=U-(a2x 2 +b2x+c2y 2 +d2y+e2+f2x 2 y+g2xy 2 +h2xy)

[0087] Among them, the error range of solving the objective functions X and Y is 1e-20 to 1e-15; F is the actual first-order frequency; U is the actual first-order frequency response component.

[0088] The other technical features of this embodiment are exactly the same as those of the first embodiment. Specific implementation method three

[0090] This embodiment is a further explanation of the first embodiment. In this embodiment, when the fault type is a crack fault, a corresponding fitting coefficient is selected based on the fault type, and a target function is established by the fitting coefficient to calculate the fault parameters, including:

[0091] Select the fitting coefficients g1, g2, g3, k1, k2, k3 and k4 of the crack fault, establish the target function Z for solving the crack fault, and calculate the crack length z;

[0092] Z=(F-(g1z 2 +g2z+g3)) / s+(U(z)-(k1z 4 +k2z 3 +k3z 2 +k4z+k5))

[0093] Among them, s is the adjustment coefficient, and its value range is 150~250; F is the actual first-order frequency; U is the actual first-order frequency response component.

[0094] The other technical features of this embodiment are exactly the same as those of the first embodiment.

[0095] This implementation is a further description of implementation one, two or three. In this implementation, a fault diagnosis database is also established. The process is as follows:

[0096] Fault simulation of the nuclear power last stage long blade based on preset fault parameters;

[0097] Among the fault parameters, the water erosion depth is 0-28mm, the water erosion length is 0-400mm, and the crack length is 0-100mm;

[0098] Obtain the aerodynamic loads of the nuclear power last stage long blades under rated conditions;

[0099] The simulation obtains the dynamic response of the nuclear power last stage long blade under aerodynamic load;

[0100] The simulated dynamic response includes the simulated first-order frequency and the simulated first-order frequency response component of the displacement at the middle position of the top of the last-stage long blade of the nuclear power plant;

[0101] The simulated dynamic response and the preset fault parameters are fitted to obtain the fitting coefficients of the water erosion fault and the fitting coefficients of the crack fault; and the fitting coefficients are set in the fault diagnosis database to complete the establishment of the fault diagnosis database.

[0102] The other technical features of this embodiment are exactly the same as those of embodiments one, two or three.

[0103] Specifically, S1. The faults of the last-stage long blades of nuclear power plants are set according to the operating experience of the last-stage long blades of nuclear power plants. The designed fault types of the last-stage long blades of nuclear power plants include crack faults and erosion faults. The crack defects of the last-stage long blades of nuclear power plants are mainly concentrated in the Stellite alloy area of ​​the blade profile, the steam outlet side area of ​​the first tooth of the blade root, and the steam inlet side area of ​​the first tooth of the blade root; the erosion rate is generally the largest in the steam inlet side area of ​​the top of the blade profile, and the set erosion fault implantation is also mainly concentrated on the air inlet side of the top of the blade profile.

[0104] Fault implantation is mainly completed through finite element mesh operations. The method of fault implantation is considered when establishing the mesh. For example, the crack is designed to crack along the mesh line, and the water erosion fault is peeled off the mesh according to the settings. Based on experience, three types of crack faults are designed, namely the Stellite alloy area of ​​the blade, the steam outlet side area of ​​the first tooth of the blade root, and the steam inlet side area of ​​the first tooth of the blade root. The crack fault implanted during simulation is 0-100mm; water erosion faults are mainly concentrated on the intake side. The water erosion depth implanted during simulation is 0-28mm, and the water erosion length is 0-400mm.

[0105] S2. The finite element analysis process is described as follows. First, the aerodynamic load of the nuclear power last stage long blade under rated conditions is calculated using fluid software to obtain preliminary load data. According to the predetermined defect form, the modified grid is implanted to make the model more in line with the actual situation. When performing simulation, it is necessary to reproduce the dynamic response of the nuclear power last stage long blade in the actual working environment, apply mechanical loads and constraints, etc., and analyze its performance under complex working conditions. When calculating the response, first calculate the natural frequencies of the nuclear power last stage long blades of each order, and use transient calculation to obtain the response components of the nuclear power last stage long blades at different frequencies, focusing on the 1st-order frequency and the 1st-order frequency response component at this frequency.

[0106] When extracting data, the main focus is on the quantities that can be measured on site, mainly the first-order frequency response component of the displacement at the middle position of the top of the last-stage long blade of the nuclear power plant. The above quantities can be extracted from the finite element results. When extracting the position, it is necessary to select the nearest node number according to the position coordinates.

[0107] S3. Summarize the simulation result data, find the corresponding relationship between the fault degree and the first-order frequency and the first-order frequency response component under this frequency, use the response surface method to fit the water erosion fault, and use the multivariate polynomial regression method to fit the crack fault. After the fitting coefficient is obtained, it is stored in the fault diagnosis database. Specific implementation method five

[0109] This embodiment is a further description of the fourth embodiment. In this embodiment, the process of fitting the simulated dynamic response and the preset fault parameters to obtain the fitting coefficient of the water erosion fault is as follows:

[0110] The simulated first-order frequency, the simulated first-order frequency response component and the simulated first-order frequency changing with the water erosion length are fitted with the least squares surface:

[0111] F'(x,y)=a1x 2 +b1x+c1y 2 +d1y+e1+f1x 2 y+g1xy 2 +h1xy Formula 1

[0112] And the simulated first-order frequency, the simulated first-order frequency response component and the simulated first-order frequency response component changing with the water erosion depth are fitted by the least squares surface:

[0113] U'(x,y)=a2x 2 +b2x+c2y 2 +d2y+e2+f2x 2 y+g2xy 2 +h2xy Formula 2 Then the fitting coefficient of water erosion fault is calculated by formula 1 and formula 2;

[0114] Among them, F' is the simulated 1st order frequency; U' is the simulated 1st order frequency

[0115] The other technical features of this embodiment are exactly the same as those of the fourth embodiment. Rate response component.

[0116] Specifically, the water erosion fault summarizes the first-order frequency and the first-order frequency response component at this frequency. The above two components change with the water erosion length and water erosion depth, and the least squares fitting surface method is used to obtain various fitting coefficients.

[0117] Among them, Figure 2 As shown in the figure, it is the law of the first-order frequency changing with the water erosion depth and water erosion length, such as Figure 3 The figure shows the law of the first-order frequency response component of the displacement response changing with the water erosion depth and water erosion length. In the above two figures, the points are the points obtained by simulation, and the surfaces are the fitting surfaces of the points. Specific implementation method 6

[0119] This embodiment is a further explanation of the fifth embodiment. In this embodiment, the process of simulating the fault of the nuclear power last stage long blade based on the fault parameters includes:

[0120] On the finite element model of the nuclear power plant's last-stage long blade, the mesh of the erosion fault area is peeled off according to the erosion length and erosion depth;

[0121] The water erosion failure area includes the inlet side area of ​​the airfoil top.

[0122] The other technical features of this embodiment are exactly the same as those of embodiment five. Specific implementation method seven

[0124] This embodiment is a further description of the fourth embodiment. In this embodiment, the process of fitting the simulated dynamic response and the preset fault parameters to obtain the fitting coefficient of the crack fault is as follows:

[0125] The simulated first-order frequency, the simulated first-order frequency response component and the crack length are fitted with a univariate polynomial regression to obtain the quadratic curve of the simulated first-order frequency with respect to the crack length:

[0126] F'(z)=g1z 2 +g2z+g3 Formula 3

[0127] And, the fourth-order curve of the simulated first-order frequency response component with respect to the crack length is obtained:

[0128] U'(z)=k1z 4 +k2z 3 +k3z 2 +k4z+k5 Formula 4 Then the fitting coefficient of crack failure is calculated by formula 3 and formula 4;

[0129] Where, F' is the simulated 1st-order frequency; U' is the simulated 1st-order frequency response component.

[0130] The other technical features of this embodiment are exactly the same as those of embodiment 4.

[0131] Specifically, the crack failure is summarized, and the crack failure is mainly implanted in Figure 4 As shown in the figure, for the three cases, the first-order frequency and the first-order frequency response component of the displacement response of each case can be plotted into a curve about the change of crack depth. A univariate polynomial regression fitting is performed on the parameters about the crack depth, and the highest power of the regression fitting parameters changes from 2 to 5. Finally, it is found that the first-order frequency has a quadratic curve relationship with the crack depth, and the first-order frequency response component of the displacement response has a better fitting effect with the fourth-order curve of the crack depth.

[0132] Among them, as shown in Figure Figure 5 The following is a graph showing the relationship between the first-order frequency and the crack depth in case 1: Figure 6 As shown, it is a graph showing the variation of the first-order frequency response component of the displacement response with the crack depth. The points in the above two figures are the points obtained by simulation, and the curves are the curves obtained by fitting. Specific implementation method eight

[0134] This embodiment is a further explanation of the seventh embodiment. In this embodiment, the process of simulating the fault of the nuclear power last stage long blade based on the fault parameters is as follows:

[0135] On the finite element model of the nuclear power last stage long blade, the cracking operation is performed along the grid lines of the cracking fault area according to the crack length;

[0136] The cracking fault region includes at least one of a Stellite alloy region of the airfoil, a steam outlet side region of a first tooth of a blade root, and a steam inlet side region of a first tooth of a blade root.

[0137] The other technical features of this embodiment are exactly the same as those of embodiment eight. Specific implementation method nine

[0139] This embodiment is a further description of the first to third and fifth to eighth embodiments. In this embodiment, it is characterized in that it also includes a method for obtaining a classifier:

[0140] The pre-collected fault types and the dynamic responses corresponding to the fault types are used as output and input respectively, and the support vector machine is trained to obtain a classifier.

[0141] The other technical features of this embodiment are exactly the same as those of embodiments one to three and five to eight.

[0142] Specifically, support vector machines are used to classify faults and distinguish fault types.

[0143] When classifying faults, the first-order frequency and displacement response components are known to determine the fault type, whether it is water erosion or cracks. Figure 7 As shown in Figure 1, the distribution of the characteristics of different faults in two-dimensional space is shown. The support vector machine is used to set the classifier. Four faults, including water erosion fault and crack faults in Cases 1 to 3, are set. The first-order frequency response components of the first-order frequency and displacement response are set as training inputs. The training labels are defined, the penalty coefficient is selected in the range of 10000-20000, and the kernel function is selected as the linear kernel function. The classification area is as follows: Figure 8 shown. Specific implementation method ten

[0145] The nuclear power last stage long blade fault diagnosis system of this embodiment includes:

[0146] Response collection module, fault classification module and fault parameter calculation module:

[0147] Select the corresponding objective function based on the fault type and calculate the fault parameters;

[0148] Response acquisition module, used to collect the actual dynamic response of the nuclear power last stage long blades;

[0149] The actual dynamic response includes the actual first-order frequency and the actual first-order frequency response component of the displacement at the middle position of the top of the nuclear power last stage long blade;

[0150] The fault classification module is used to classify and predict the actual dynamic response through a classifier and identify the fault type;

[0151] Failure types include water erosion failure and crack failure;

[0152] The fault parameter module is used to select the corresponding fitting coefficient based on the fault type, and establish the target function through the fitting coefficient to calculate the fault parameter;

[0153] The fault fitting coefficients are preset in the fault diagnosis database;

[0154] The fault parameters include the water erosion length and water erosion depth caused by water erosion faults, and / or the crack length caused by crack faults.

[0155] The present invention makes full use of the advantage of the finite element method in simulating complex working environments. The responses of the nuclear power last-stage long blades under different faults are obtained through finite element simulation calculations. Different types and degrees of faults are implanted in the nuclear power last-stage long blades, and the responses are calculated. The simulation result data are summarized to find the corresponding relationship between the fault degree and the result. The correlation between the key focus quantity and the fault degree is found through data mining and an equation is established. The response surface method is used to fit the water erosion fault, and the multi-variable polynomial regression method is used to fit the crack fault. A support vector machine is used to classify the faults and distinguish the fault types. The least squares method is used to solve the fault degree.

[0156] Although the present invention is described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the present invention. It should therefore be understood that many modifications may be made to the exemplary embodiments and that other arrangements may be devised without departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the features of the various dependent claims and herein may be combined in a manner different from that described in the original claims. It should also be understood that the features described in conjunction with the individual embodiments may be used in other embodiments.

Claims

1. A method for diagnosing faults of long blades in the last stage of a nuclear power plant, characterized in that: Here’s how: Collect the actual dynamic response of the long blades in the last stage of nuclear power; The actual dynamic response includes an actual first-order frequency and an actual first-order frequency response component of the displacement of the middle position of the top of the last-stage long blade of the nuclear power plant; Classify and predict the actual dynamic response by using a classifier to identify the fault type; The fault types include water erosion fault and crack fault; Selecting a corresponding fitting coefficient based on the fault type, establishing a target function through the fitting coefficient, and calculating the fault parameters; The fault fitting coefficient is preset in the fault diagnosis database; The fault parameters include the water erosion length and water erosion depth caused by the water erosion fault, and / or the crack length caused by the crack fault.

2. The method for diagnosing faults of nuclear power last stage long blades according to claim 1, characterized in that: When the fault type is identified as a water erosion fault, a corresponding fitting coefficient is selected based on the fault type, and a target function is established by using the fitting coefficient to calculate the fault parameters, including: Select the fitting coefficients a1, b1, c1, d1, f1, g1, h1, a2, b2, c2, d2, f2, g2 and h2 of water erosion fault, establish the target functions X and Y for solving water erosion fault, and calculate the water erosion length x and water erosion depth y; X=F-(a1x 2 +b1x+c1y 2 +d1y+e1+f1x 2 y+g1xy 2 +h1xy) <h2 style=";text-align:left;direction:ltr">Y = U - (a2x<h2 style=";text-align:left;direction:ltr"> 2 <h2 style=";text-align:left;direction:ltr"> +b2x+c2y<h2 style=";text-align:left;direction:ltr"> 2 <h2 style=";text-align:left;direction:ltr"> +d2y+e2+f2x<h2 style=";text-align:left;direction:ltr"> 2 <h2 style=";text-align:left;direction:ltr"> y+g2xy<h2 style=";text-align:left;direction:ltr"> 2 <h2 style=";text-align:left;direction:ltr"> +h2xy) Among them, the error range of solving the objective functions X and Y is 1e-20 to 1e-15; F is the actual first-order frequency; U is the actual first-order frequency response component.

3. The method for diagnosing faults of nuclear power last stage long blades according to claim 1, characterized in that: When the fault type is a crack fault, a corresponding fitting coefficient is selected based on the fault type, and a target function is established by using the fitting coefficient to calculate the fault parameters, including: Select the fitting coefficients g1, g2, g3, k1, k2, k3 and k4 of the crack fault, establish the target function Z for solving the crack fault, and calculate the crack length z; Z=(F-(g1z 2 +g2z+g3)) / s+(U(z)-(k1z 4 +k2z 3 +k3z 2 +k4z+k5)) Among them, s is the adjustment coefficient, and its value range is 150~250; F is the actual first-order frequency; U is the actual first-order frequency response component.

4. The method for diagnosing faults of nuclear power last stage long blades according to claim 1, 2 or 3, characterized in that: It also includes establishing a fault diagnosis database, the process is as follows: Fault simulation of the nuclear power last stage long blade based on preset fault parameters; The fault parameters include water erosion depth of 0 to 28 mm, water erosion length of 0 to 400 mm, and crack length of 0 to 100 mm; Obtaining the aerodynamic load of the nuclear power last stage long blade under rated operating conditions; Simulating and obtaining the dynamic response of the nuclear power last stage long blade under the aerodynamic load; The simulated dynamic response includes a simulated first-order frequency and a simulated first-order frequency response component of the displacement of the middle position of the top of the last-stage long blade of the nuclear power plant; The simulated dynamic response and the preset fault parameters are fitted to obtain the fitting coefficient of the water erosion fault and the fitting coefficient of the crack fault; and the fitting coefficients are set in the fault diagnosis database to complete the establishment of the fault diagnosis database.

5. The method for diagnosing faults of nuclear power last stage long blades according to claim 4, characterized in that: The process of fitting the simulated dynamic response and the preset fault parameters to obtain the fitting coefficient of the water erosion fault is as follows: The simulated first-order frequency, the simulated first-order frequency response component and the variation of the simulated first-order frequency with the water erosion length are fitted by the least squares surface: F'(x,y)=a1x 2 +b1x+c1y 2 +d1y+e1+f1x 2 y+g1xy 2 +h1xy Formula 1 And the simulated first-order frequency, the simulated first-order frequency response component and the simulated first-order frequency response component changing with the water erosion depth are fitted by the least squares surface: U'(x,y)=a2x 2 +b2x+c2y 2 +d2y+e2+f2x 2 y+g2xy 2 +h2xy Formula 2 Then the fitting coefficient of water erosion fault is calculated by formula 1 and formula 2; Where, F' is the simulated 1st-order frequency; U' is the simulated 1st-order frequency response component.

6. The method for diagnosing faults of nuclear power last stage long blades according to claim 5, characterized in that: The process of fault simulation of the nuclear power last stage long blade based on fault parameters includes: On a finite element model of a nuclear power last stage long blade, a mesh stripping operation is performed on a water erosion fault area according to the water erosion length and water erosion depth; The water erosion failure area includes the air inlet side area of ​​the airfoil top.

7. The method for diagnosing faults of nuclear power last stage long blades according to claim 4, characterized in that: The process of fitting the simulated dynamic response and the preset fault parameters to obtain the fitting coefficient of the crack fault is as follows: The simulated first-order frequency, the simulated first-order frequency response component and the crack length are fitted with a univariate polynomial regression to obtain the quadratic curve of the simulated first-order frequency with respect to the crack length: F'(z)=g1z 2 +g2z+g3 Formula 3 And, the fourth-order curve of the simulated first-order frequency response component with respect to the crack length is obtained: U'(z)=k1z 4 +k2z 3 +k3z 2 +k4z+k5 Formula 4 Then the fitting coefficient of crack failure is calculated by formula 3 and formula 4; Where, F' is the simulated 1st-order frequency; U' is the simulated 1st-order frequency response component.

8. The method for diagnosing faults of nuclear power last stage long blades according to claim 7, characterized in that: The process of fault simulation for the nuclear power last stage long blade based on fault parameters is as follows; On a finite element model of a nuclear power last stage long blade, a cracking operation is performed along a grid line of a cracking fault area according to the crack length; The crack fault area includes at least one of the Stellite alloy area of ​​the blade profile, the steam outlet side area of ​​the first tooth of the blade root, and the steam inlet side area of ​​the first tooth of the blade root.

9. The method for diagnosing faults of nuclear power last stage long blades according to any one of claims 1 to 3 and 5 to 8, characterized in that: It also includes methods for obtaining classifiers: The pre-collected fault types and the dynamic responses corresponding to the fault types are used as output and input respectively, and a support vector machine is trained to obtain the classifier.

10. A nuclear power last stage long blade fault diagnosis system, characterized in that: include; Response collection module, fault classification module and fault parameter calculation module: Select a corresponding objective function based on the fault type and calculate the fault parameters; The response acquisition module is used to collect the actual dynamic response of the nuclear power last stage long blade; The actual dynamic response includes an actual first-order frequency and an actual first-order frequency response component of the displacement of the middle position of the top of the last-stage long blade of the nuclear power plant; The fault classification module is used to classify and predict the actual dynamic response through a classifier to identify the fault type; The fault types include water erosion fault and crack fault; The fault parameter module is used to select a corresponding fitting coefficient based on the fault type, and establish a solution objective function through the fitting coefficient to calculate the fault parameter; The fault fitting coefficient is preset in the fault diagnosis database; The fault parameters include the water erosion length and water erosion depth caused by the water erosion fault, and / or the crack length caused by the crack fault.

Images