A method for evaluating the health status of electromagnetic coils based on image features

Abstract

The present invention relates to an electromagnetic coil health assessment method based on image features, which uses the electromagnetic coil full life image data to determine the degradation sensitive features; obtains a clear healthy electromagnetic coil image; extracts multiple color features based on the image to construct a high-dimensional feature set; uses the degradation sensitive features to obtain the degradation sensitive feature set; normalizes the degradation sensitive feature set, and obtains the Mahalanobis distance of the electromagnetic coil based on the sensitive feature set; uses Box-Cox transformation to realize the Mahalanobis distance normal distribution, calculates its mean and variance, and defines the degradation threshold and fault threshold; the coil to be tested extracts the sensitive features and calculates the Mahalanobis distance, and then implements Box-Cox transformation to establish the health index of the electromagnetic coil to be tested, and compares the health index of the electromagnetic coil to be tested with the threshold to achieve its health status assessment. The present invention evaluates the actual health status of the electromagnetic coil based on image features. Compared with the traditional method based on electrical signals, it does not require an additional high-frequency signal generator device, and has the advantages of low cost and easy application.

Description

Technical Field

[0001] The invention relates to the field of electromagnetic coil health status monitoring, in particular to an electromagnetic coil health status assessment method based on image features. Background Art

[0002] As a key component of energy conversion equipment, electromagnetic coils are core components of electric motors, generators, transformers, solenoid valves and other equipment. Electromagnetic coils are usually composed of copper wires and polymer insulating materials contained outside the copper wires, and are wound several times with the help of a skeleton to form an electromagnetic coil. The electromagnetic induction principle is used to achieve the conversion of electrical energy and mechanical energy. Electromagnetic coils have been widely used in various industries such as process industries, automobiles, and engineering machinery. Oak Ridge National Laboratory in the United States collected 10 years of failure data on solenoid valves used in the safety protection system of commercial nuclear power plants in the United States. The results show that more than 50% of solenoid valve failures are related to electromagnetic coil insulation failures; data from the International Committee on Large Electric Power Systems (CIGRE) show that 56% of generator failures are related to winding insulation failures.

[0003] The insulation failure of electromagnetic coils usually starts with a turn-to-turn short-circuit fault. Since the application of the turn-to-turn short-circuit fault detection method is based on the premise that the turn-to-turn short-circuit fault has occurred, the electromagnetic coil has entered the late degradation stage at this time, and the turn-to-turn short-circuit fault will quickly spread in the electromagnetic coil and cause the coil to fail completely. For example, relevant research shows that for a 15KW induction motor, the time from the occurrence of the turn-to-turn short-circuit fault to the complete failure of the winding is less than 2 seconds). Therefore, the above method has no practical significance for realizing predictive maintenance of equipment and formulating an overall maintenance plan.

[0004] Since the insulation is in a slow degradation process before a turn-to-turn short-circuit fault occurs, it is more meaningful for predictive maintenance to continuously monitor the turn-to-turn insulation status in the early insulation degradation stage and predict the insulation life based on this. Perisse et al. from the University of Clermont-Ferrand II in France used the thermal accelerated degradation test to study the trend of the electrical parameters of electromagnetic wires during insulation degradation, and found that there was a correlation between the capacitance value of the sample and the breakdown voltage, and then proposed a method to monitor the health status of inter-turn insulation based on the resonant frequency of the winding; Grubic et al. from the Georgia Institute of Technology in the United States constructed a high-frequency parameter winding model, and used simulation technology to study the response of the model impedance, and thus proposed the idea of ​​using high-frequency impedance to monitor the health status of insulation; Sumislawska et al. from Coventry University in the United Kingdom completed the thermal accelerated degradation test of the stator winding and measured the impedance spectrum of the sample, and fitted the parameters of the equivalent circuit model based on the impedance spectrum data to study the influence of thermal stress on the electrical properties of the winding; Zoeller et al. from the Vienna University of Technology in Austria proposed an online monitoring method for the insulation status of the stator winding of the traction motor, by injecting the inverter induced voltage into the traction motor winding, and analyzing the degradation status of the stator winding based on the change of the high-frequency component of its transient current response. The above methods all use certain high-frequency electrical parameters (such as impedance, capacitance, etc.) at a certain characteristic frequency to monitor the insulation degradation of the coil. Injecting high-frequency signals into the tested / monitored equipment is a prerequisite for applying these methods in practice. Therefore, the above research methods all require the development of high-frequency signal generators and their integration into the tested / monitored equipment for health assessment, which will incur additional testing / monitoring costs.

[0005] To this end, the present invention proposes an evaluation method for the health status of electromagnetic coils based on image features, targeting the insulation failure mode and related research of electromagnetic coils. The image features of electromagnetic coils in different degradation states are used to analyze the health status of electromagnetic coils. The proposed technology can not only detect whether the electromagnetic coil can work normally, but also accurately monitor the actual insulation degradation state of the electromagnetic coil. Compared with the traditional method based on electrical signals, it does not require the addition of a high-frequency signal generator device, has the advantages of low cost and easy application, and provides an effective solution for the insulation performance test and health monitoring of electromagnetic coils. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention provides an electromagnetic coil health status assessment method based on image features, which constructs a high-dimensional feature set by extracting multiple image features from the original image, determines the degradation sensitive feature set with the help of a correlation evaluation algorithm, and finally calculates the Mahalanobis distance based on the degradation sensitive feature set, and uses the Mahalanobis distance as the health index to achieve the health evaluation of the electromagnetic coil to be tested. The present invention can not only detect whether the electromagnetic coil can work normally, but also accurately monitor the actual insulation degradation state of the electromagnetic coil, thereby laying the foundation for realizing maintenance based on the health state and improving the accuracy and reliability of the electromagnetic coil degradation monitoring of the corresponding equipment components.

[0007] The technical solution adopted by the present invention to achieve the above-mentioned purpose is: a method for evaluating the health status of electromagnetic coils based on image features, comprising the following steps:

[0008] Image acquisition: acquiring images containing electromagnetic coils; the images include a plurality of life cycle images for representing different states of the electromagnetic coils, and a plurality of healthy electromagnetic coil images that are considered to be healthy electromagnetic coils;

[0009] Establish an evaluation model: implement color space feature preprocessing on the life cycle image of the electromagnetic coil, obtain color feature data information, and construct a historical data set; perform correlation evaluation on the historical data set, determine the degradation sensitive feature vector, and construct the degradation sensitive feature set Q based on the determined degradation sensitive feature vector;

[0010] The degradation sensitive feature set G of the healthy state is established by using the degradation sensitive feature vector, and the degradation sensitive feature set G of the healthy state is normalized by calculating the mean and variance of the degradation sensitive feature set of the healthy state to obtain the normalized degradation sensitive feature set H; according to the data of the healthy degradation sensitive feature set H, the Mahalanobis distance value of each healthy electromagnetic coil image is obtained by using the Mahalanobis distance;

[0011] The Mahalanobis distance value of the healthy electromagnetic coil image is transformed into a normal distribution, the baseline health index of the electromagnetic coil is constructed, and the degradation threshold and the failure threshold are determined;

[0012] Health status assessment: collect images of the electromagnetic coil to be tested, and obtain a baseline health index as the health index of the electromagnetic coil to be tested by establishing an assessment model step. Compare the health index of the electromagnetic coil to be tested with the fault threshold and the degradation threshold to achieve health status assessment of the electromagnetic coil to be tested.

[0013] The method of implementing color space feature preprocessing on the life cycle image of the electromagnetic coil and constructing a historical data set according to the color space feature information includes the following steps:

[0014] According to the p life cycle images collected, a historical data set consisting of color space vectors is obtained;

[0015] The color space vectors of an image include the X-direction vector, Y-direction vector and grayscale vector of the red channel in RGB space, the X-direction vector, Y-direction vector and grayscale vector of the green channel in RGB space, the X-direction vector, Y-direction vector and grayscale vector of the blue channel in RGB space, the X-direction vector, Y-direction vector and grayscale vector of the hue channel in HSV space, the X-direction vector, Y-direction vector and grayscale vector of the saturation channel in HSV space, the X-direction vector, Y-direction vector and grayscale vector of the lightness in HSV space, the X-direction vector, Y-direction vector and grayscale vector of the hue channel in HSL space, the X-direction vector, Y-direction vector and grayscale vector of the saturation channel in HSL space, the X-direction vector, Y-direction vector and grayscale vector of the brightness in HSL space, a total of 27 feature vectors.

[0016] The process of evaluating the correlation of the historical data set, determining the degradation-sensitive feature vector, and constructing the degradation-sensitive feature set Q based on the determined degradation-sensitive feature vector includes the following steps:

[0017] The Spearman rank correlation coefficient ρ was used for correlation evaluation:

[0018]

[0019] Among them, the corresponding element difference d i =x ij -y i , image i=1,2,…,p,x ij is the vector X in the historical data set j =x 1j ,x 2j ,x 3j …x pj The jth eigenvector of the i-th image, j = 1, 2, ..., 27, y i is the time vector element, y i =i; p is the number of life cycle images;

[0020] In the historical data set, m feature vectors with |ρ|≥threshold are extracted to form the degradation sensitive feature set Q={q1,q2,…q m}.

[0021] The degradation sensitive feature set H is obtained by the following steps:

[0022] According to the healthy electromagnetic coil image, the degradation sensitive feature set G = [X1, X2, ... X m ],X1,X2,…X m Respectively represent q1, q2, …q m The eigenvectors of the same type;

[0023] Normalize the feature set G and set the mean Where q = 1, 2, ..., m, n represents the number of healthy electromagnetic coil images, x i,q represents the element of the qth degradation-sensitive feature of the i-th image in the healthy image; let the variance Then the feature set G sensitive to health state degradation is normalized as: H represents the normalized health state degradation sensitive feature set; X m , denote the mth eigenvector of n healthy coil images and the mean of the mth eigenvector of n healthy coil images, σ m represents the variance of the mth eigenvector of n healthy coil images, X′ m is the mth degradation-sensitive eigenvector of the n normalized healthy coil images.

[0024] The method of obtaining the Mahalanobis distance value of each healthy electromagnetic coil image by using the Mahalanobis distance according to the data of the health degradation sensitive feature set H comprises the following steps:

[0025] Where j = 1, 2, ..., n

[0026] Among them, Z j =[z j,1 ,z j,2 ,…z j,m ] are the normalized m degradation-sensitive features of the jth image in the healthy state, corr -1 is the inverse matrix of the covariance matrix of the healthy degradation sensitive feature set H, Z j The transpose of MD j is the Mahalanobis distance value of the jth healthy image, m represents the number of feature vectors of the degradation sensitive feature set of the healthy state, and n represents the number of healthy electromagnetic coil images.

[0027] The method of performing a normal distribution transformation on the Mahalanobis distance value of the healthy electromagnetic coil image to construct a health index of the electromagnetic coil includes the following steps:

[0028] Through the normal distribution transformation method, the baseline health index HI_baseline of the i-th healthy electromagnetic coil image is obtained. i , using the following formula:

[0029] Where i = 1, 2, ..., n

[0030] Among them, λ is the transformation coefficient, which is obtained by maximum likelihood estimation, HI_baseline iis the baseline health index of the i-th healthy electromagnetic coil image, and n represents the number of healthy electromagnetic coil images.

[0031] The determination of the degradation threshold and the failure threshold is specifically as follows:

[0032] According to the mean μ and variance σ of the baseline health index HI_baseline of n images, the fault threshold and degradation threshold are determined:

[0033] Degradation Threshold degradation =|μ±σ|

[0034] Fault Threshold fault =|μ±3σ|.

[0035] The health status assessment comprises the following steps:

[0036] Acquire an image of the electromagnetic coil to be tested;

[0037] For the image of the electromagnetic coil to be tested, a degradation-sensitive feature vector of the image is obtained according to the degradation-sensitive feature vector type;

[0038] The degradation-sensitive feature vector of the image is obtained by averaging the healthy coils and variance σ q Perform normalization processing;

[0039] According to the normalized degradation sensitive feature vector of the electromagnetic coil image to be tested, the Mahalanobis distance value of the electromagnetic coil image to be tested is obtained by using the Mahalanobis distance;

[0040] Performing normal distribution transformation on the Mahalanobis distance value of the electromagnetic coil image to be tested, and constructing the health index of the electromagnetic coil to be tested;

[0041] The health index of the electromagnetic coil to be tested is compared with the fault threshold and the degradation threshold to implement the health status assessment of the electromagnetic coil to be tested.

[0042] The step of comparing the health index of the electromagnetic coil to be tested with the fault threshold and the degradation threshold to evaluate the health status of the electromagnetic coil to be tested comprises the following steps:

[0043] When HI≤Threshold degradation , it is determined that the electromagnetic coil to be tested is in a healthy state, and HI represents the health index of the electromagnetic coil to be tested;

[0044] When Threshold degradation <HI≤Threshold fault , it is determined that the electromagnetic coil to be tested is in a degraded state, and the increase in the HI value represents the severity change of the electromagnetic coil from mild degradation to moderate degradation and finally to severe degradation;

[0045] When HI>Threshold fault , it is determined that the electromagnetic coil to be tested is in a fault state.

[0046] An electromagnetic coil health status assessment system based on image features, comprising:

[0047] An image acquisition module, used to acquire or obtain images containing electromagnetic coils; the images include a plurality of life cycle images for representing different states of the electromagnetic coils, and a plurality of healthy electromagnetic coil images that are considered to be healthy electromagnetic coils;

[0048] An evaluation model module is established to implement color space feature preprocessing of the life cycle image of the electromagnetic coil, obtain color feature data information, and construct a historical data set; a correlation evaluation is performed on the historical data set to determine the degradation sensitive feature vector, and a degradation sensitive feature set Q is constructed based on the determined degradation sensitive feature vector;

[0049] The degradation sensitive feature set G of the healthy state is established by using the degradation sensitive feature vector, and the degradation sensitive feature set G of the healthy state is normalized by calculating the mean and variance of the degradation sensitive feature set of the healthy state to obtain the normalized degradation sensitive feature set H; according to the data of the healthy degradation sensitive feature set H, the Mahalanobis distance value of each healthy electromagnetic coil image is obtained by using the Mahalanobis distance;

[0050] The Mahalanobis distance value of the healthy electromagnetic coil image is transformed into a normal distribution, the baseline health index of the electromagnetic coil is constructed, and the degradation threshold and the failure threshold are determined;

[0051] The health status assessment module is used to collect images of the electromagnetic coil to be tested, and obtain a baseline health index as the health index of the electromagnetic coil to be tested by establishing an assessment model step, and compare the health index of the electromagnetic coil to be tested with a fault threshold and a degradation threshold to achieve health status assessment of the electromagnetic coil to be tested.

[0052] The present invention has the following beneficial effects and advantages:

[0053] 1. The present invention can not only detect whether the electromagnetic coil can work normally, but also accurately monitor the actual insulation degradation state of the electromagnetic coil, thereby laying a foundation for realizing condition-based maintenance.

[0054] 2. The present invention realizes image capture of the camera through the electromagnetic coil, does not need to add an additional high-frequency signal generator device, and has the advantages of low cost and easy application.

[0055] 3. The present invention implements an evaluation of the health status of the electromagnetic coil. The method of the present invention can be used to implement health monitoring for any device that uses the electromagnetic coil as a component, and has a wide range of applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 is a flow chart of the method of the present invention;

[0057] Figure 2 It is the DC resistance collected by the present invention through the laboratory accelerated fatigue test, which is used to determine whether the electromagnetic coil is faulty;

[0058] Figure 3 The present invention obtains the electromagnetic coil life cycle image information by means of laboratory accelerated fatigue test;

[0059] Figure 4 is the Spearman rank correlation coefficient of the high-dimensional feature set;

[0060] Table 1 is the degradation sensitive feature set data of the present invention;

[0061] Figure 5 Shown is the MD value obtained based on the sensitive feature set of the present invention.

[0062] Figure 6 It is shown that the present invention uses the health index HI after Box-Cox transformation, and the degradation threshold and the fault threshold determined by the statistical confidence interval criterion to realize the health status assessment of the electromagnetic coil. DETAILED DESCRIPTION

[0063] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0064] The present invention relates to a method for evaluating the health of electromagnetic coils based on image features, comprising: determining degradation sensitive features using the full life image data of the electromagnetic coil; photographing the healthy electromagnetic coil using a digital camera or a microscope to obtain a clear image of the healthy electromagnetic coil; extracting multiple color features based on the image to construct a high-dimensional feature set; obtaining a degradation sensitive feature set based on the high-dimensional feature set using the degradation sensitive features; normalizing the degradation sensitive feature set, and obtaining the Mahalanobis distance of the electromagnetic coil based on the sensitive feature set; realizing the normal distribution of the Mahalanobis distance by means of Box-Cox transformation, calculating its mean and variance, and defining the degradation threshold and the fault threshold according to statistical theory; extracting the Mahalanobis distance of the coil to be tested by sensitive features, and then implementing Box-Cox transformation to establish the health index of the electromagnetic coil to be tested, and using the comparison between the health index of the electromagnetic coil to be tested and the threshold to achieve its health status evaluation. The present invention evaluates the actual health status of the electromagnetic coil based on image features. Compared with the traditional method based on electrical signals, it does not require the additional addition of a high-frequency signal generator device, has the advantages of low cost and easy application, and provides an effective solution for realizing the insulation performance test and health monitoring of the electromagnetic coil.

[0065] A method for monitoring the health status of an electromagnetic coil based on image features comprises the following steps:

[0066] Modeling preparation, that is, the stage of determining the degradation-sensitive eigenvector:

[0067] Step 1: Obtain life cycle images, including one image each at different stages such as healthy, degradation 1, degradation 2, degradation 3, and failure. The total number of images in each stage is p, and they are named historical state images;

[0068] Step 2: Use MATLAB software to process multiple color space feature data of the historical state image to obtain a high-dimensional data set. This high-dimensional data set is composed of color space vectors, such as RGB space, HSV space, and HSL space. A total of 27 vectors construct a high-dimensional data set, which is a matrix of p rows and 27 columns;

[0069] Step 3: Use the Spearman rank correlation coefficient to evaluate the correlation. The Spearman rank correlation coefficient benchmark model is:

[0070]

[0071] Among them, the corresponding element difference d i =x ij -y i , where i = 1, 2, ..., p, x ij is the vector X in the historical data set j =x 1j ,x 2j ,x 3j …x pj The data of the jth feature vector of the i-th image (where j = 1, 2, ..., 27), y i is the time vector, y i =i, and ρ is the Spearman rank correlation coefficient of this vector. In the historical data set, the degradation sensitive feature of the electromagnetic coil is determined by the size of |ρ|≥0.85. At this time, the degradation sensitive feature is the part of the vectors with |ρ|≥0.85 in the 27 vectors. Let the number of vectors be m, and record Q={q1,q2,…q m}, thus completing the acquisition of degradation sensitive features, that is, selecting m vectors that meet the conditions from 27 color features;

[0072] Model establishment, that is, the stage of building a baseline health index:

[0073] Step 4: Collect n images of healthy electromagnetic coils, use MATLAB software to process the data to obtain the m feature vectors in step 3, and construct the degradation sensitive feature set G of the health state;

[0074] Step 5: Normalize the sensitive feature set G, and let the degraded sensitive feature set G = [X1, X2, ...X m ], normalize the feature set G, and set the mean Where q=1,2,…,m, let variance Then the feature set G sensitive to health state degradation is normalized as:

[0075] Step 6: Obtain the Mahalanobis distance of the healthy electromagnetic coil, which can be expressed as:

[0076] Where j = 1, 2, ..., n

[0077] Where Z j =[z j,1 ,z j,2 ,…z j,m ] are the m degradation-sensitive features of the jth image of the normalized healthy electromagnetic coil, corr -1 is the inverse matrix of the covariance matrix of the healthy degradation sensitive feature set H, Z j The transpose of MD j is the Mahalanobis distance value of the jth healthy image;

[0078] Step 7: Perform Box-Cox transformation on the MD value of the health state using the following formula:

[0079] Where i = 1, 2, ..., n

[0080] Get the baseline health index HI_baseline of the i-th healthy electromagnetic coil image i , where λ is the transformation coefficient, obtained by maximum likelihood estimation, ,HI_baseline i is the baseline health index of the i-th healthy electromagnetic coil image, and the mean μ and variance σ of the baseline health index HI_baseline of n images are calculated at the same time. The method for constructing the fault threshold and degradation threshold is as follows: Define Threshold degradation =|μ±σ| is the degradation threshold, and Threshold is defined fault =|μ±3σ| is the fault threshold;

[0081] Health assessment, that is, the health evaluation stage of the electromagnetic coil to be tested:

[0082] Step 8: Obtain the health index HI of the electromagnetic coil to be tested, and complete it in the following order:

[0083] S1. Use a digital camera or a microscope to photograph the electromagnetic coil and obtain an image;

[0084] S2. MATLAB and other related software should be used to obtain data of degradation-sensitive feature vectors;

[0085] S3. Using the mean and variance and normalization method in step 5, the degradation sensitive feature vector of the electromagnetic coil to be tested is normalized, where the mean and variance σ q are the mean and variance of the corresponding features of the healthy electromagnetic coil;

[0086] In the detection stage, the normalization formula In the above example, substitute the mth feature vector of the electromagnetic coil image to be tested into X m , substitute the variance of the mth eigenvector of the electromagnetic coil image to be tested into σ m ;

[0087] S4. Using the method of calculating MD in step 6, obtain the MD value of the electromagnetic coil to be tested, where corr -1 is the inverse matrix of the covariance matrix of the healthy electromagnetic coil; in the Mahalanobis distance value formula, the normalized m degradation sensitive features of the jth image of the electromagnetic coil to be tested are substituted into Z j =[z j,1 ,z j,2 ,…z j,m ];

[0088] S5. Use the Box-Cox transformation method in step 7 to calculate the health index to obtain the HI of the electromagnetic coil to be tested, where λ is the transformation coefficient determined by the healthy electromagnetic coil.

[0089] Step 9: The health index HI of the electromagnetic coil to be tested is compared with the fault threshold and the degradation threshold to evaluate the health status of the electromagnetic coil to be tested. The following comparison method is adopted:

[0090] When HI≤Threshold degradation , determine that the electromagnetic coil to be tested is in a healthy state,

[0091] When Threshold degradation <HI≤Threshold fault , it is determined that the electromagnetic coil to be tested is in a degraded state, and the increase in the HI value represents the severity change of the electromagnetic coil from mild degradation to moderate degradation and finally to severe degradation;

[0092] When HI>Threshold fault , it is determined that the electromagnetic coil to be tested is in a fault state.

[0093] The health status of the electromagnetic coil under test is evaluated by comparing the range.

[0094] The present invention is a method for evaluating the health of electromagnetic coils based on image features. Figure 1 The method flow chart of the present invention is shown, which is specifically implemented according to the following steps:

[0095] Step 1: Use a digital camera or microscope to take a picture of the electromagnetic coil to obtain a clear image of the current state of the electromagnetic coil. The image should be taken at a fixed position of the electromagnetic coil and the image should be taken at the same position. The purpose is to obtain an image that can effectively characterize the health characteristics of the electromagnetic coil and determine whether it is faulty with the help of the DC resistance of the electromagnetic coil. Figure 2 As shown in FIG. 1 , the DC resistance value of the fifth degradation cycle decreases significantly, thereby determining that the electromagnetic coil is faulty. Based on this, the image of the electromagnetic coil from healthy to degraded to faulty is also determined, as shown in FIG. Figure 3 As shown;

[0096] Step 2: Use image processing methods, such as MATLAB and other tool software, to preprocess the image using color space features such as RGB, HSV, and HSL, obtain data on 27 color features, and construct a historical data set;

[0097] Step 3: With the help of the acquired electromagnetic coil historical data set, the degradation sensitive features are determined using the Spearman rank correlation coefficient; the vector data of the historical data set is evaluated for correlation with the degradation time (such as Figure 4 As shown), to obtain the degradation sensitive feature set Q of the electromagnetic coil state change, that is, the seven vectors in Table 1, thereby realizing the selection of the degradation sensitive feature vector;

[0098] Table 1 Sensitive feature set of the embodiment (|ρ|≥0.85)

[0099]

[0100] Step 4: Collect n images of the healthy electromagnetic coil, use MATLAB software to process the data to obtain the seven feature vectors in step 3, and construct the degradation sensitive feature set G of the health state;

[0101] Step 5: Normalize the sensitive feature set G, let the degraded sensitive feature set G = [X1, X2, ... X7], normalize the feature set G, and set the mean Where q = 1, 2, ..., 7, let the variance Then the feature set G sensitive to health state degradation is normalized as:

[0102] Step 6: Obtain the Mahalanobis distance of the healthy electromagnetic coil, which can be expressed as:

[0103] Where j = 1, 2, ..., n

[0104] Where Z j =[z j,1 ,z j,2 ,…z j,m ] is the degradation sensitive eigenvector, corr -1 is the inverse matrix of the covariance matrix of the healthy degradation sensitive feature set H, Z j The transpose of MD j is the Mahalanobis distance value of the jth healthy image;

[0105] Step 7: Perform Box-Cox transformation on the MD value of the health state using the following formula:

[0106] Where i = 1, 2, ..., n

[0107] Get the electromagnetic coil baseline health index HI_baseline i , where λ is calculated using maximum likelihood estimation, HI_baseline i is the baseline health index of the i-th healthy electromagnetic coil image, and the mean μ and variance σ of the baseline health index HI_baseline of n images are calculated at the same time. The method for constructing the fault threshold and degradation threshold is as follows: Define Threshold degradation =|μ±σ| is the degradation threshold, and Threshold is defined fault =|μ±3σ| is the fault threshold;

[0108] Step 8: Obtain the health index HI of the electromagnetic coil to be tested, and complete it in the following order:

[0109] S1. Use a digital camera or a microscope to photograph the electromagnetic coil and obtain an image;

[0110] S2. MATLAB and other related software should be used to obtain data of degradation-sensitive feature vectors;

[0111] S3. Using the mean and variance and normalization method in step 5, the degradation sensitive feature vector of the electromagnetic coil to be tested is normalized, where the mean and variance σ q are the mean and variance of the corresponding features of the healthy electromagnetic coil;

[0112] S4. Using the method for calculating MD in step 6, obtain the MD value of the electromagnetic coil to be tested, such as Figure 5 As shown, corr -1 is the inverse matrix of the covariance matrix of the healthy electromagnetic coil;

[0113] S5. Use the Box-Cox transformation method in step 7 to calculate the health index to obtain the HI of the electromagnetic coil to be tested, where λ is the transformation coefficient determined by the healthy electromagnetic coil, such as Figure 6 Data points shown in blue.

[0114] Step 9: The health index HI of the electromagnetic coil to be tested is compared with the fault threshold and the degradation threshold to evaluate the health status of the electromagnetic coil to be tested. The following comparison method is adopted:

[0115] When HI≤Threshold degradation , determine whether the electromagnetic coil to be tested is in a healthy state, such as Figure 6 The total health index of the first and second electromagnetic coils is in the healthy range, which means it is in a healthy state;

[0116] When Threshold degradation <HI≤Threshold fault , determine that the electromagnetic coil to be tested is in a degraded state, such as Figure 6 The third and fourth electromagnetic coil health indexes are in the degradation range, that is, the degradation state, and the increase in HI value represents the severity change of the electromagnetic coil from mild degradation to moderate degradation and finally severe degradation;

[0117] When HI>Threshold fault , determine that the electromagnetic coil to be tested is in a fault state, such as Figure 6 As shown, the fifth electromagnetic coil health index is in the fault range, that is, it is in a fault state.

Claims

1. A method for evaluating the health status of electromagnetic coils based on image features, characterized in that: The following steps are involved: Image acquisition: acquiring images containing electromagnetic coils; the images include a plurality of life cycle images for representing different states of the electromagnetic coils, and a plurality of healthy electromagnetic coil images that are considered to be healthy electromagnetic coils; Establish an evaluation model: implement color space feature preprocessing on the life cycle image of the electromagnetic coil, obtain color feature data information, and construct a historical data set; perform correlation evaluation on the historical data set, determine the degradation sensitive feature vector, and construct the degradation sensitive feature set Q based on the determined degradation sensitive feature vector; The degradation-sensitive feature set G of the healthy state is established by using the degradation-sensitive feature vector, and the degradation-sensitive feature set G of the healthy state is normalized by calculating the mean and variance of the degradation-sensitive feature set of the healthy state to obtain the normalized degradation-sensitive feature set H; According to the data of the health degradation sensitive feature set H, the Mahalanobis distance value of each healthy electromagnetic coil image is obtained by using the Mahalanobis distance; The Mahalanobis distance value of the healthy electromagnetic coil image is transformed into a normal distribution, the baseline health index of the electromagnetic coil is constructed, and the degradation threshold and the failure threshold are determined; Health status assessment: collect images of the electromagnetic coil to be tested, and obtain a baseline health index as the health index of the electromagnetic coil to be tested by establishing an assessment model step. Compare the health index of the electromagnetic coil to be tested with the fault threshold and the degradation threshold to achieve health status assessment of the electromagnetic coil to be tested.

2. The method for evaluating the health status of electromagnetic coils based on image features according to claim 1, characterized in that: The method of implementing color space feature preprocessing on the life cycle image of the electromagnetic coil and constructing a historical data set according to the color space feature information includes the following steps: According to the p life cycle images collected, a historical data set consisting of color space vectors is obtained; The color space vectors of an image include the X-direction vector, Y-direction vector and grayscale vector of the red channel in RGB space, the X-direction vector, Y-direction vector and grayscale vector of the green channel in RGB space, the X-direction vector, Y-direction vector and grayscale vector of the blue channel in RGB space, the X-direction vector, Y-direction vector and grayscale vector of the hue channel in HSV space, the X-direction vector, Y-direction vector and grayscale vector of the saturation channel in HSV space, the X-direction vector, Y-direction vector and grayscale vector of the lightness in HSV space, the X-direction vector, Y-direction vector and grayscale vector of the hue channel in HSL space, the X-direction vector, Y-direction vector and grayscale vector of the saturation channel in HSL space, the X-direction vector, Y-direction vector and grayscale vector of the brightness in HSL space, a total of 27 feature vectors.

3. The method for evaluating the health status of electromagnetic coils based on image features according to claim 1, characterized in that: The process of evaluating the correlation of the historical data set, determining the degradation-sensitive feature vector, and constructing the degradation-sensitive feature set Q based on the determined degradation-sensitive feature vector includes the following steps: The Spearman rank correlation coefficient ρ was used for correlation evaluation: Among them, the corresponding element difference d i =x ij -y i , image i=1,2,…,p,x ij is the vector X in the historical data set j =x 1j , x 2j , x 3j …x pj The jth eigenvector of the i-th image, j = 1, 2, ..., 27, y i is the time vector element, y i =i; p is the number of life cycle images; In the historical data set, m feature vectors with |ρ|≥threshold are extracted to form the degradation sensitive feature set Q={q1, q2, …q m }.

4. The method for evaluating the health status of electromagnetic coils based on image features according to claim 1, characterized in that: The degradation sensitive feature set H is obtained by the following steps: According to the healthy electromagnetic coil image, the degradation sensitive feature set G = [X1, X2, ...X m ], X1, X2, … X m Respectively represent q1, q2, …q m The eigenvectors of the same type; Normalize the feature set G and set the mean Where q = 1, 2, ..., m, n represents the number of healthy electromagnetic coil images, x i，q represents the element of the qth degradation-sensitive feature of the i-th image in the healthy image; Assume variance Then the feature set G sensitive to health state degradation is normalized as: H represents the normalized health state degradation sensitive feature set; X m , denote the mth eigenvector of n healthy coil images and the mean of the mth eigenvector of n healthy coil images, σ m represents the variance of the mth eigenvector of n healthy coil images, X′ m is the mth degradation-sensitive eigenvector of the n normalized healthy coil images.

5. The method for evaluating the health status of electromagnetic coils based on image features according to claim 1, characterized in that: The method of obtaining the Mahalanobis distance value of each healthy electromagnetic coil image by using the Mahalanobis distance according to the data of the health degradation sensitive feature set H comprises the following steps: Where j = 1, 2, ..., n Among them, Z j =[z j，1 , z j,2 , …z j，m ] are the normalized m degradation-sensitive features of the jth image in the healthy state, corr -1 is the inverse matrix of the covariance matrix of the healthy degradation sensitive feature set H, Z j The transpose of MD j is the Mahalanobis distance value of the jth healthy image, m represents the number of feature vectors of the degradation sensitive feature set of the healthy state, and n represents the number of healthy electromagnetic coil images.

6. The method for evaluating the health status of electromagnetic coils based on image features according to claim 1, characterized in that: The method of performing a normal distribution transformation on the Mahalanobis distance value of the healthy electromagnetic coil image to construct a health index of the electromagnetic coil includes the following steps: Through the normal distribution transformation method, the baseline health index HI_baseline of the i-th healthy electromagnetic coil image is obtained. i , using the following formula: Where i = 1, 2, ..., n Among them, λ is the transformation coefficient, which is obtained by maximum likelihood estimation, HI_baseline i is the baseline health index of the i-th healthy electromagnetic coil image, and n represents the number of healthy electromagnetic coil images.

7. The method for evaluating the health status of electromagnetic coils based on image features according to claim 1, characterized in that: The determination of the degradation threshold and the failure threshold is specifically as follows: According to the mean μ and variance σ of the baseline health index HI_baseline of n images, the fault threshold and degradation threshold are determined: Degradation Threshold degradation =|μ±σ| Fault Threshold fault =|μ±3σ|.

8. The method for evaluating the health status of electromagnetic coils based on image features according to claim 1, characterized in that: The health status assessment comprises the following steps: Acquire an image of the electromagnetic coil to be tested; For the image of the electromagnetic coil to be tested, a degradation-sensitive feature vector of the image is obtained according to the degradation-sensitive feature vector type; The degradation-sensitive feature vector of the image is obtained by averaging the healthy coils and variance σ q Perform normalization processing; According to the normalized degradation sensitive feature vector of the electromagnetic coil image to be tested, the Mahalanobis distance value of the electromagnetic coil image to be tested is obtained by using the Mahalanobis distance; Performing normal distribution transformation on the Mahalanobis distance value of the electromagnetic coil image to be tested, and constructing the health index of the electromagnetic coil to be tested; The health index of the electromagnetic coil to be tested is compared with the fault threshold and the degradation threshold to implement the health status assessment of the electromagnetic coil to be tested.

9. The method for evaluating the health status of electromagnetic coils based on image features according to claim 1 or 8, characterized in that: The step of comparing the health index of the electromagnetic coil to be tested with the fault threshold and the degradation threshold to evaluate the health status of the electromagnetic coil to be tested comprises the following steps: When HI≤Threshold degradation , it is determined that the electromagnetic coil to be tested is in a healthy state, and HI represents the health index of the electromagnetic coil to be tested; When Threshold degradation <HI≤Threshold fault , it is determined that the electromagnetic coil to be tested is in a degraded state, and the increase in the HI value represents the severity change of the electromagnetic coil from mild degradation to moderate degradation and finally to severe degradation; When HI>Threshold fault , it is determined that the electromagnetic coil to be tested is in a fault state.

10. A system for evaluating the health status of electromagnetic coils based on image features, characterized in that: include: An image acquisition module, used to acquire or obtain images containing electromagnetic coils; the images include a plurality of life cycle images for representing different states of the electromagnetic coils, and a plurality of healthy electromagnetic coil images that are considered to be healthy electromagnetic coils; An evaluation model module is established to implement color space feature preprocessing of the life cycle image of the electromagnetic coil, obtain color feature data information, and construct a historical data set; a correlation evaluation is performed on the historical data set to determine the degradation sensitive feature vector, and a degradation sensitive feature set Q is constructed based on the determined degradation sensitive feature vector; The degradation-sensitive feature set G of the healthy state is established by using the degradation-sensitive feature vector, and the degradation-sensitive feature set G of the healthy state is normalized by calculating the mean and variance of the degradation-sensitive feature set of the healthy state to obtain the normalized degradation-sensitive feature set H; According to the data of the health degradation sensitive feature set H, the Mahalanobis distance value of each healthy electromagnetic coil image is obtained by using the Mahalanobis distance; The Mahalanobis distance value of the healthy electromagnetic coil image is transformed into a normal distribution, the baseline health index of the electromagnetic coil is constructed, and the degradation threshold and the failure threshold are determined; The health status assessment module is used to collect images of the electromagnetic coil to be tested, and obtain a baseline health index as the health index of the electromagnetic coil to be tested by establishing an assessment model step, and compare the health index of the electromagnetic coil to be tested with a fault threshold and a degradation threshold to achieve health status assessment of the electromagnetic coil to be tested.

Images