Turbine low-pressure cylinder last-stage blade edge abrasion identification method

Abstract

The turbine low-pressure cylinder last-stage blade edge abrasion method comprises the steps that picture information of a turbine last-stage blade is obtained and converted into a grey-scale map; performing block filtering, gradient calculation and the like on the grey-scale map; solving an edge curve and a convex hull line of the blade, and calculating a centroid; calculating a distance and angle sequence formed by the centroid, the edge curve and the convex hull line; obtaining an edge curve starting point serial number and a screening edge curve terminal point serial number by using a filtering and extreme value screening mode; obtaining a first closed curve coordinate sequence and a second closed curve coordinate sequence which are composed of the edge curve, the convex hull line and the centroid; the edge abrasion area is finally obtained through a Gaussian shoelace formula, whether edge abrasion reaches a set alarm threshold value or not is judged, and an alarm is given in time; the method can quickly and accurately identify the edge wear of the last-stage blade of the low-pressure cylinder of the steam turbine, improves the identification efficiency, reduces the investment of manpower and material resources, and is of great significance to the intelligent construction of a power plant.

Description

Technical Field

[0001] The present application belongs to the field of edge wear identification of the last-stage blades of a low-pressure cylinder of a steam turbine, and specifically relates to a method for edge wear identification of the last-stage blades of a low-pressure cylinder of a steam turbine. Background Art

[0002] The last-stage blades of the low-pressure cylinder of a steam turbine are an important core component of the steam turbine. They are long and work in a humid environment. The working environment is relatively complex. In addition, due to the demand for peak load regulation in recent years, the working conditions in the low-pressure last-stage area have become even worse. The main failure form of the last-stage blades of the low-pressure cylinder of a steam turbine is blade edge wear. The main causes of blade edge wear are blade water erosion and foreign body impact. Severe wear may even cause blade breakage. At present, some steam turbine manufacturers have installed endoscope equipment on the last-stage blades to monitor and evaluate the edge wear status of the steam turbine in real time. However, in the existing technology, the identification of edge wear of the last-stage blades of the low-pressure cylinder of a steam turbine is still at the stage of manual identification. The manual identification method is too subjective, has low accuracy, and cannot quantify the amount of wear. The manual identification method also requires a lot of manpower and material resources, which does not meet the needs of intelligent power plants. Summary of the invention

[0003] In order to solve the problem of low efficiency of the prior art method of manually identifying the edge wear of the last stage blade of the low pressure cylinder of a steam turbine, a method for identifying the edge wear of the last stage blade of the low pressure cylinder of a steam turbine is proposed.

[0004] A method for identifying edge wear of a last-stage blade of a low-pressure cylinder of a steam turbine comprises:

[0005] Step 1: Obtain image information of the last-stage blade of the steam turbine, and convert the obtained image information into a grayscale image;

[0006] Step 2: Obtain the image matrix R of the grayscale image obtained in step 1; obtain the edge curve according to the image matrix R, and obtain the convex hull line and the centroid according to the edge curve;

[0007] Step 3: Calculate angle and distance data according to the edge curve, convex hull line and centroid, wherein the angle and distance data include the angle sequence between the convex hull line and the centroid, the angle sequence between the edge curve and the centroid, and the distance sequence between the edge curve and the centroid;

[0008] Step 4: Filter and select the extreme values ​​of the angle and distance data to obtain the sequence number of the angle sequence between the blade edge starting point and the edge curve and the centroid and the sequence number of the angle sequence between the blade edge end point and the edge curve and the centroid;

[0009] Step 5: According to the obtained sequence number of the angle sequence between the edge curve and the centroid of the blade edge starting point and the sequence number of the angle sequence between the edge curve and the centroid of the blade edge end point, the angle sequence between the edge curve and the centroid and the distance sequence between the convex hull and the centroid are intercepted to obtain sequence L1 and sequence L3 respectively; the centroid is added after sequence L1 to obtain the first closed curve coordinate sequence; the starting point of sequence L1, the end point of sequence L1 and the centroid are added after sequence L3 to obtain the second closed curve coordinate sequence;

[0010] Step 6: Calculate the first closed area by using the Gaussian shoelace formula according to the first closed curve coordinate sequence, calculate the second closed area by using the Gaussian shoelace formula according to the second closed curve coordinate sequence, and obtain the edge wear area by subtracting the first closed area from the second closed area;

[0011] Step 7: Determine whether the edge wear area is greater than the set threshold, and if so, issue an alarm.

[0012] Beneficial Effects

[0013] The present invention obtains image information of the last-stage blade of a steam turbine and converts it into a one-dimensional grayscale image; performs block filtering and gradient calculation on the grayscale image to obtain an image matrix, so as to eliminate noise and expose the edge of the metal blade; performs corrosion and expansion calculation on the image matrix, so as to close the metal edge of the blade; obtains the edge curve and convex hull of the blade to calculate the centroid; calculates the distance and angle sequence formed by the centroid and the edge curve and the convex hull; uses filtering and extreme value screening to obtain the edge curve starting point sequence number and the edge curve end point sequence number; obtains the first closed curve coordinate sequence and the second closed curve coordinate sequence composed of the edge curve, the convex hull and the centroid; uses the Gaussian shoelace formula to calculate the closed curve area, finally obtains the edge wear area, and judges whether the edge wear of the last-stage blade of the low-pressure cylinder of the steam turbine reaches the set alarm threshold, and issues an alarm in time; the method can quickly and accurately identify the edge wear of the last-stage blade of the low-pressure cylinder of the steam turbine, improve the recognition efficiency, reduce the investment of manpower and material resources, and is of great significance to the intelligent construction of power plants. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A flow chart of a method for identifying edge wear of a low-pressure cylinder last-stage blade of a steam turbine according to a specific implementation method of the present application;

[0015] Figure 2 The image obtained by performing block filtering on the grayscale image in the specific implementation mode of this application;

[0016] Figure 3 The image obtained after gradient processing for the specific implementation method of this application;

[0017] Figure 4The image obtained after expansion of the specific implementation method of this application;

[0018] Figure 5 For the specific implementation of this application, the edge curve starting point serial number and the edge curve end point serial number position image are screened;

[0019] Figure 6 The image of the blade edge wear area for the specific implementation method of this application;

[0020] Figure 7 This is an image of the unworn area of ​​the blade edge in accordance with the specific embodiment of the present application. DETAILED DESCRIPTION

[0021] The following will be combined Figures 1 to 7 , describing this embodiment, a method for identifying edge wear of a low-pressure cylinder last-stage blade of a steam turbine, comprising:

[0022] Step 1: Obtain image information of the last-stage blade of the steam turbine, and convert the obtained image information into a grayscale image;

[0023] Step 2: Obtain the image matrix R of the grayscale image obtained in step 1; obtain the edge curve according to the image matrix R, and obtain the convex hull line and the centroid according to the edge curve;

[0024] Step 3: Calculate angle and distance data according to the edge curve, convex hull line and centroid, wherein the angle and distance data include the angle sequence between the convex hull line and the centroid, the angle sequence between the edge curve and the centroid, and the distance sequence between the edge curve and the centroid;

[0025] Step 4: Filter and select the extreme values ​​of the angle and distance data to obtain the sequence number of the angle sequence between the blade edge starting point and the edge curve and the centroid and the sequence number of the angle sequence between the blade edge end point and the edge curve and the centroid;

[0026] Step 5: According to the obtained sequence number of the angle sequence between the edge curve and the centroid of the blade edge starting point and the sequence number of the angle sequence between the edge curve and the centroid of the blade edge end point, the angle sequence between the edge curve and the centroid and the distance sequence between the convex hull and the centroid are intercepted to obtain sequence L1 and sequence L3 respectively; the centroid is added after sequence L1 to obtain the first closed curve coordinate sequence; the starting point of sequence L1, the end point of sequence L1 and the centroid are added after sequence L3 to obtain the second closed curve coordinate sequence;

[0027] Step 6: Calculate the first closed area by using the Gaussian shoelace formula according to the first closed curve coordinate sequence, calculate the second closed area by using the Gaussian shoelace formula according to the second closed curve coordinate sequence, and obtain the edge wear area by subtracting the first closed area from the second closed area;

[0028] Step 7: Determine whether the edge wear area is greater than the set threshold, and if so, issue an alarm.

[0029] Furthermore, the method of converting the obtained image information into a grayscale image is:

[0030] grayImage=cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)

[0031] Among them, grayImage is the grayscale image matrix, img is the image information, cv2.cvtColor is the conversion function, and cv2.COLOR_BGR2GRAY is the conversion parameter.

[0032] Specifically, the image information of the last stage blades of the low-pressure cylinder of the turbine is obtained through a high-speed camera or a high-speed video camera. The image information is an img image, and the img image is converted into a grayscale image of grayImage using the cvtColor package in openCV. The grayscale image of grayImage is a two-dimensional matrix with a horizontal length of h and a vertical length of v, and h and v are related to the size of the original image.

[0033] Furthermore, the method of obtaining the image matrix R of the grayscale image obtained in step 1 includes:

[0034] Step 21: Perform block filtering on the grayscale image to obtain the image matrix L;

[0035] L=cv2.boxFilter(grayImage,-1,(F_1,F_2),normalize=0)

[0036] Among them, cv2.boxFilter is the box filter function, grayImage is the grayscale image, normalize = 0 means no normalization, (F_1, F_2) is the kernel of the box filter, F_1 ​​= F_2, and the values ​​of F_1 and F_2 are in the range of 2 to 5;

[0037] Step 22: Perform gradient processing on the obtained image matrix L to obtain the image matrix P;

[0038] Specifically, the grayscale image grayImage is subjected to block filtering and gradient processing in order to eliminate noise and expose the metal edge of the blade.

[0039] Step 23: Corrode the image matrix P to obtain the image matrix Q;

[0040] Q = cv2.erode(P,D,k1)

[0041] Where cv2.erode is the erosion function, D is the kernel, k1 is the number of corrosion iterations, and the value of k1 is in the range of 1-2;

[0042] Step 24: dilate the image matrix Q to obtain the image matrix R;

[0043] R = cv2.dilate(Q,D,k2)

[0044] Among them, cv2.dilate is the dilation function, k2 is the number of dilation iterations, and the value of k2 is in the range of 1-2.

[0045] Specifically, the erosion and dilation processing is performed to close the edges; the image matrix P is first eroded to obtain the image matrix Q, and the method of using is the erode function in opencv (cv2). The image matrix Q is first eroded to obtain the image matrix R, and the method of using is the dilate function in opencv (cv2). The image after erosion and dilation is as follows Figure 4 shown.

[0046] Furthermore, the obtained image matrix L is subjected to gradient processing to obtain an image matrix P, including:

[0047] According to the graphic matrix L, the edge matrix sobelxy is obtained by the sobel operator, and the obtained edge matrix sobelxy is binarized to obtain the image matrix P;

[0048] sobelxy=cv2.Sobel(src=L,ddepth=cv2.CV_64F,dx=dx1,dy=dy1,ksize=z)

[0049] Among them, cv2.Sobel is the edge detection function, src is the calculation object, ddepth = cv2.CV_64F is the calculation depth, dx = dx1, dy = dy1 is to obtain the first-order derivative, ksize is the sobel operator size, and z is 9 or 11;

[0050]

[0051] Among them, a=[1,2...h], b=[1,2...v], h is the horizontal length of the grayscale image, v is the vertical length of the grayscale image, a is the row sequence of the image matrix P, and b is the column sequence of the image matrix P.

[0052] Specifically, the gradient of the graphic matrix L is obtained using the sobel method and the Sobel function of openCV. The obtained matrix is ​​sobelxy; the calculated gradient is negative, so the obtained gradient must be binarized and the edge must be blackened. The processing method is to process the matrix sobelxy. When the element is 0, it becomes 255, and if it is non-0, it becomes 0. The processed matrix is ​​P.

[0053] Further, the edge curve of the leaf is obtained according to the image matrix R, and the convex hull and centroid of the leaf are obtained according to the edge curve of the leaf, including:

[0054] Step 25: Obtain the edge curve H of the leaf through the findContours function according to the image matrix R;

[0055] H=findContours(R,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_NONE)

[0056] Among them, cv2.RETR_EXTERNAL and cv2.CHAIN_APPROX_NONE) are parameters;

[0057] Step 26: Get the convex hull J according to the edge curve H of the leaf through the cv2.convexHull function;

[0058] J = cv2.convexHull(H)

[0059] Step 27: Obtain the centroid A[l,m] according to the edge curve H of the leaf;

[0060]

[0061] Among them, max(·) is the maximum value, min(·) is the minimum value, To round up, c is the edge curve row vector, d is the edge curve column vector, l is the row coordinate of the centroid, and m is the column coordinate of the centroid.

[0062] Furthermore, the angle between the convex hull and the centroid is calculated as follows:

[0063] AnglePoint_T = math.atan2(-(el),(fm))*180 / 3.14, where AnglePoint_T is the angle between the convex hull and the centroid; math.atan is the angle calculation function, e is the convex hull row vector, and f is the convex hull column vector;

[0064] AnglePoint_T represents the angle change curve when the centroid moves counterclockwise along the convex hull line.

[0065] The angle between the edge curve and the centroid is calculated as:

[0066] AnglePoint_r = math.atan2(-(cl),(dm))*180 / 3.14, where AnglePoint_r is the angle between the edge curve and the centroid, c is the row vector of the edge curve, and d is the column vector of the edge curve;

[0067] AnglePoint_r is the curve showing the angle change when the centroid moves counterclockwise along the Bird Garden curve.

[0068] The distance between the edge curve and the centroid is calculated as:

[0069]

[0070] Among them, LengthPoint_r is the distance between the edge curve and the centroid;

[0071] Specifically, LengthPoint_r represents the distance change curve when the centroid moves counterclockwise along the edge curve.

[0072] Furthermore, the angle and distance data are filtered and extreme value screened to obtain the sequence number of the angle sequence between the blade edge starting point and the edge curve and the centroid and the sequence number of the angle sequence between the blade edge end point and the edge curve and the centroid, including:

[0073] Step 41: Select the points whose angles are greater than Angle1 and less than Angle2 in the angle sequence AnglePoint_r between the edge curve and the centroid, and obtain the index start sequence number H_low and the index end sequence number H_up according to the selected points;

[0074] Step 42: According to the index start sequence number H_low and the index end sequence number H_up, the distance sequence LengthPoint_r between the edge curve and the centroid is truncated to obtain a new sequence LPJ, and LPJ is filtered according to the first window size to obtain an LPF signal, and the peak point of the LPF signal is found to obtain LPP, and the coordinate index [LPP_low, LPP_up] is obtained according to LPP;

[0075] LPF=savgol_filter(LPJ,window1,Deg)

[0076] Where savgol_filter is filter function 1, window1 is the first window size, window1 value is in the range of 600 to 800, Deg is the polynomial order, Deg = 3;

[0077] LPP=argrelextrema(LPF,np.greater)

[0078] Among them, argrelextrema is a function for finding extreme values, and np.greater means finding the maximum value;

[0079] Step 43: Filter LPJ according to the second window size to obtain the LPF1 signal, find the peak point of the LPF1 signal to obtain LPP1; find the points in the coordinate index [LPP_low, LPP_up] in LPP1 to form the sequence LPP c , according to the sequence LPP c Get LPP c The minimum LPP clow and LPP c The maximum LPP cup ;

[0080] LPF1=savgol_filter(LPJ,window2,Deg)

[0081] LPP1=argrelextrema(LPF1,np.greater)

[0082] Window2 is the second window size, and the value of window2 is in the range of 40 to 50.

[0083] Step 44: According to LPP c The minimum LPP clow 、LPP c The maximum LPP up And the coordinate index [LPP_low,LPP_up] gets the sequence number of the angle sequence between the edge curve and the centroid of the blade edge starting point Node low Node number of the angle sequence between the edge curve and the centroid of the blade edge end point up ;

[0084] Node low =LPP clow +LPP_low

[0085] Node up =LPP cup +LPP_low.

[0086] Specifically, LPP clow is the serial number of the starting point of the blade edge, LPP cupIt is the serial number of the end of the blade edge. The serial number of the corresponding point is consistent in LengthPoint_r and AnglePoint_r. The serial number of the starting point of the blade edge in LengthPoint_r and AnglePoint_r is Node low , the sequence number of the blade edge end point at LengthPoint_r and AnglePoint_r is Node up .

[0087] Furthermore, the value range of Angle1 is [-135, -90], and the value range of Angle2 is [45, 60].

[0088] Further, according to the obtained sequence number of the angle sequence between the edge curve and the centroid at the starting point of the blade edge and the sequence number of the angle sequence between the edge curve and the centroid at the end point of the blade edge, the angle sequence between the edge curve and the centroid and the distance sequence between the convex hull and the centroid are intercepted to obtain the sequence L1 and the sequence L3, which include:

[0089] Step 51: Find the sequence number Node in the angle sequence AnglePoint_r between the edge curve and the centroid low and sequence number Node up The corresponding angle AnglePoint_r[Node low ] and AnglePoint_r[Node up ];

[0090] Step 52: The distance sequence LengthPoint_r between the edge curve and the centroid is sorted by the sequence number [Node low ,Node up ] is truncated to obtain sequence L1, the starting point of sequence L1 is LengthPoint_r[Node low ], the end point is LengthPoint_r[Node up ];

[0091] Find the angle sequence AnglePoint_T between the convex hull and the centroid. The angle is in AnglePoint_r[Node low ] and AnglePoint_r[Node up ] between the sequence L3;

[0092] Adding the centroid after the sequence L1 to obtain the first closed curve coordinate sequence; adding the starting point of the sequence L1, the end point of the sequence L1 and the centroid after the sequence L3 to obtain the second closed curve coordinate sequence includes:

[0093] Step 53: Add the centroid [l,m] after L1 to form a new sequence L2, L2(xi ,y i ),i∈[1,s1], s1 is the data length of sequence L2, L2 is the first closed curve coordinate sequence, x i is the row coordinate of sequence L2, y i Column coordinates for sequence L2;

[0094] The sequence L3, LengthPoint_r[Node low ]、LengthPoint_r[Node up ] and the centroid A[l,m] are arranged in sequence to obtain a new sequence L4, L4(g j ,h j ),j∈[1,s2], s2 is the data length of sequence L4, L4 is the second closed curve coordinate sequence, g j is the row coordinate of sequence L4, h j Column coordinates for sequence L4.

[0095] Further, a first closed area is calculated by using the Gaussian shoelace formula according to the first closed curve coordinate sequence, a second closed area is calculated by using the Gaussian shoelace formula according to the second closed curve coordinate sequence, and an edge wear area is obtained by subtracting the first closed area from the second closed area, including:

[0096] The first closed area A1 is calculated by using the Gaussian shoelace formula according to the first closed curve coordinate sequence L2;

[0097]

[0098] The second closed area A2 is calculated based on the second closed curve coordinate sequence L4 by using the Gaussian shoelace formula;

[0099]

[0100] The edge wear area A is obtained by subtracting the first closed area A1 from the second closed area A2. cal , A cal =A2-A1.

[0101] When area A cal When the set threshold is exceeded, an alarm is triggered, indicating that the blade edge is currently worn too much and needs to be processed. The image information and intermediate calculation process of the turbine's last-stage blades are stored.

[0102] 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 various dependent claims and features described herein may be combined in a manner different from that described in the original claims. It should also be understood that features described in conjunction with individual embodiments may be used in other described embodiments.

Claims

1. A method for identifying edge wear of the last stage blade of a low pressure cylinder of a steam turbine, characterized in that: include: Step 1: Obtain image information of the last-stage blade of the steam turbine, and convert the obtained image information into a grayscale image; Step 2: Obtain the image matrix R of the grayscale image obtained in step 1; obtain the edge curve according to the image matrix R, and obtain the convex hull line and the centroid according to the edge curve; Step 3: Calculate angle and distance data according to the edge curve, convex hull line and centroid, wherein the angle and distance data include the angle sequence between the convex hull line and the centroid, the angle sequence between the edge curve and the centroid, and the distance sequence between the edge curve and the centroid; Step 4: Filter and select the extreme values ​​of the angle and distance data to obtain the sequence number of the angle sequence between the blade edge starting point and the edge curve and the centroid and the sequence number of the angle sequence between the blade edge end point and the edge curve and the centroid; Step 5: According to the obtained sequence number of the angle sequence between the edge curve and the centroid of the blade edge starting point and the sequence number of the angle sequence between the edge curve and the centroid of the blade edge end point, the angle sequence between the edge curve and the centroid and the distance sequence between the convex hull and the centroid are intercepted to obtain sequence L1 and sequence L3 respectively; the centroid is added after sequence L1 to obtain the first closed curve coordinate sequence; the starting point of sequence L1, the end point of sequence L1 and the centroid are added after sequence L3 to obtain the second closed curve coordinate sequence; Step 6: Calculate the first closed area by using the Gaussian shoelace formula according to the first closed curve coordinate sequence, calculate the second closed area by using the Gaussian shoelace formula according to the second closed curve coordinate sequence, and obtain the edge wear area by subtracting the first closed area from the second closed area; Step 7: Determine whether the edge wear area is greater than the set threshold, and if so, issue an alarm.

2. A method for identifying edge wear of a low-pressure cylinder last-stage blade of a steam turbine according to claim 1, characterized in that: The method of converting the obtained image information into a grayscale image is: grayImage=cv2.cvtColor(img,cv2.COLOR_BGR2GRAY) Among them, grayImage is the grayscale image matrix, img is the image information, cv2.cvtColor is the conversion function, and cv2.COLOR_BGR2GRAY is the conversion parameter.

3. The method for identifying edge wear of the last stage blade of a low pressure cylinder of a steam turbine according to claim 1, characterized in that: The method of obtaining the image matrix R of the grayscale image obtained in step 1 includes: Step 21: Perform block filtering on the grayscale image to obtain the image matrix L; L=cv2.boxFilter(grayImage,-1,(F_1,F_2),normalize=0) Among them, cv2.boxFilter is the box filter function, grayImage is the grayscale image, normalize = 0 means no normalization, (F_1, F_2) is the kernel of the box filter, F_1 ​​= F_2, and the values ​​of F_1 and F_2 are in the range of 2 to 5; Step 22: Perform gradient processing on the obtained image matrix L to obtain the image matrix P; Step 23: Corrode the image matrix P to obtain the image matrix Q; Q = cv2.erode(P,D,k1) Among them, cv2.erode is the erosion function, k1 is the number of corrosion iterations, and the value of k1 is in the range of 1-2; Step 24: dilate the image matrix Q to obtain the image matrix R; R = cv2.dilate(Q,D,k2) Among them, cv2.dilate is the dilation function, k2 is the number of dilation iterations, and the value of k2 is in the range of 1-2.

4. A method for identifying edge wear of a low-pressure cylinder last-stage blade of a steam turbine according to claim 3, characterized in that: The image matrix L is subjected to gradient processing to obtain the image matrix P, including: According to the graphic matrix L, the edge matrix sobelxy is obtained by the sobel operator, and the obtained edge matrix sobelxy is binarized to obtain the image matrix P; sobelxy=cv2.Sobel(src=L,ddepth=cv2.CV_64F,dx=dx1,dy=dy1,ksize=z) Among them, cv2.Sobel is the edge detection function, src = L is the calculation object, ddepth = cv2.CV_64F is the calculation depth, dx = dx1, dy = dy1 is to obtain the first-order derivative, ksize is the sobel operator size, and z is 9 or 11; Among them, a=[1,2...h], b=[1,2...v], h is the horizontal length of the grayscale image, v is the vertical length of the grayscale image, a is the row sequence of the image matrix P, and b is the column sequence of the image matrix P.

5. The method for identifying edge wear of the last stage blade of a low pressure cylinder of a steam turbine according to claim 1, characterized in that: The edge curve of the leaf is obtained according to the image matrix R, and the convex hull and centroid of the leaf are obtained according to the edge curve of the leaf, including: Step 25: Obtain the edge curve H of the leaf through the findContours function according to the image matrix R; H=findContours(R,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_NONE) Among them, cv2.RETR_EXTERNAL and cv2.CHAIN_APPROX_NONE) are the parameters of findContours function; Step 26: Get the convex hull J according to the edge curve H of the leaf through the cv2.convexHull function; J = cv2.convexHull(H) Step 27: Obtain the centroid A[l,m] according to the edge curve H of the leaf; Among them, max(·) is the maximum value, min(·) is the minimum value, is rounded up, c is the edge curve row vector, d is the edge curve column vector, l is the row coordinate of the centroid, and m is the column coordinate of the centroid.

6. A method for identifying edge wear of a low-pressure cylinder last-stage blade of a steam turbine according to claim 5, characterized in that: The calculation method of the angle sequence between the convex hull and the centroid is: AnglePoint_T=math.atan2(-(el),(fm))*180 / 3.14 Among them, AnglePoint_T is the angle sequence between the convex hull and the centroid; math.atan is the angle calculation function, e is the convex hull row vector, and f is the convex hull column vector; The calculation method of the angle sequence between the edge curve and the centroid is: AnglePoint_r=math.atan2(-(cl),(dm))*180 / 3.14 Where AnglePoint_r is the angle sequence between the edge curve and the centroid, c is the edge curve row vector, and d is the edge curve column vector; The distance sequence between the edge curve and the centroid is calculated as: Among them, LengthPoint_r is the distance sequence between the edge curve and the centroid.

7. A method for identifying edge wear of a low-pressure cylinder last-stage blade of a steam turbine according to claim 6, characterized in that: The angle and distance data are filtered and extreme value screened to obtain the sequence number of the angle sequence between the blade edge starting point and the edge curve and the centroid and the sequence number of the angle sequence between the blade edge end point and the edge curve and the centroid, including: Step 41: Select the points whose angles are greater than Angle1 and less than Angle2 in the angle sequence AnglePoint_r between the edge curve and the centroid, and obtain the index start sequence number H_low and the index end sequence number H_up according to the selected points; Step 42: According to the index start sequence number H_low and the index end sequence number H_up, the distance sequence LengthPoint_r between the edge curve and the centroid is truncated to obtain a new sequence LPJ, and LPJ is filtered according to the first window size to obtain an LPF signal, and the peak point of the LPF signal is found to obtain LPP, and the coordinate index [LPP_low, LPP_up] is obtained according to LPP; LPF=savgol_filter(LPJ,window1,Deg) Where savgol_filter is filter function 1, window1 is the first window size, window1 value is in the range of 600 to 800, Deg is the polynomial order, Deg = 3; LPP=argrelextrema(LPF,np.greater) Among them, argrelextrema is a function for finding extreme values, and np.greater means finding the maximum value; Step 43: Filter LPJ according to the second window size to obtain the LPF1 signal, find the peak point of the LPF1 signal to obtain LPP1; find the points in the coordinate index [LPP_low, LPP_up] in LPP1 to form the sequence LPP c , according to the sequence LPP c Get LPP c The minimum LPP clow and LPP c The maximum LPP cup ; LPF1=savgol_filter(LPJ,window2,Deg) LPP1=argrelextrema(LPF1,np.greater) Window2 is the second window size, and the value of window2 is in the range of 40 to 50. Step 44: According to LPP c The minimum LPP clow 、LPP c The maximum LPP up And the coordinate index [LPP_low,LPP_up] gets the sequence number of the angle sequence between the edge curve and the centroid of the blade edge starting point Node low Node number of the angle sequence between the edge curve and the centroid of the blade edge end point up ; Node low =LPP clow +LPP_low Node up =LPP cup +LPP_low。 8. A method for identifying edge wear of a low-pressure cylinder last-stage blade of a steam turbine according to claim 7, characterized in that: According to the obtained sequence number of the angle sequence between the blade edge starting point and the edge curve and the centroid and the sequence number of the angle sequence between the blade edge end point and the edge curve and the centroid, the angle sequence between the edge curve and the centroid and the distance sequence between the convex hull and the centroid are intercepted to obtain the sequence L1 and the sequence L3, including: Step 51: Find the sequence number Node in the angle sequence AnglePoint_r between the edge curve and the centroid low and sequence number Node up The corresponding angle AnglePoint_r[Node low ] and AnglePoint_r[Node up ]; Step 52: The distance sequence LengthPoint_r between the edge curve and the centroid is sorted by the sequence number [Node low ,Node up ] is truncated to obtain sequence L1, the starting point of sequence L1 is LengthPoint_r[Node low ], the end point is LengthPoint_r[Node up ]; Find the angle sequence AnglePoint_T between the convex hull and the centroid. The angle is in AnglePoint_r[Node low ] and AnglePoint_r[Node up ] between the sequence L3; Adding the centroid after the sequence L1 to obtain the first closed curve coordinate sequence; adding the starting point of the sequence L1, the end point of the sequence L1 and the centroid after the sequence L3 to obtain the second closed curve coordinate sequence includes: Step 53: Add the centroid [l,m] after L1 to form a new sequence L2, L2(x i ,y i ),i∈[1,s1], s1 is the data length of sequence L2, L2 is the first closed curve coordinate sequence, x i is the row coordinate of sequence L2, y i Column coordinates for sequence L2; Sequence L3, LengthPoint_r[Node low ]、LengthPoint_r[Node up ] and the centroid A[l,m] are arranged in sequence to obtain a new sequence L4, L4(g j ,h j ),j∈[1,s2], s2 is the data length of sequence L4, L4 is the second closed curve coordinate sequence, g j is the row coordinate of sequence L4, h j Column coordinates for sequence L4.

9. A method for identifying edge wear of a low-pressure cylinder last-stage blade of a steam turbine according to claim 8, characterized in that: The first closed area is calculated by the Gaussian shoelace formula according to the first closed curve coordinate sequence, the second closed area is calculated by the Gaussian shoelace formula according to the second closed curve coordinate sequence, and the edge wear area is obtained by subtracting the first closed area from the second closed area, including: The first closed area A1 is calculated by using the Gaussian shoelace formula according to the first closed curve coordinate sequence L2; The second closed area A2 is calculated based on the second closed curve coordinate sequence L4 by using the Gaussian shoelace formula; The edge wear area A is obtained by subtracting the first closed area A1 from the second closed area A2. cal , A cal =A2-A1.

10. A method for identifying edge wear of a low-pressure cylinder last-stage blade of a steam turbine according to claim 7, characterized in that: The value range of Angle1 is [-135,-90], and the value range of Angle2 is [45,60].

Images