CN108921812B - Intelligent evaluation method for fatigue state of breaker spring based on image recognition - Google Patents

Intelligent evaluation method for fatigue state of breaker spring based on image recognition Download PDF

Info

Publication number
CN108921812B
CN108921812B CN201810450389.8A CN201810450389A CN108921812B CN 108921812 B CN108921812 B CN 108921812B CN 201810450389 A CN201810450389 A CN 201810450389A CN 108921812 B CN108921812 B CN 108921812B
Authority
CN
China
Prior art keywords
spring
circuit breaker
fatigue state
fatigue
image
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201810450389.8A
Other languages
Chinese (zh)
Other versions
CN108921812A (en
Inventor
黄辉敏
苏毅
杨健
芦宇峰
夏小飞
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Electric Power Research Institute of Guangxi Power Grid Co Ltd
Original Assignee
Electric Power Research Institute of Guangxi Power Grid Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Electric Power Research Institute of Guangxi Power Grid Co Ltd filed Critical Electric Power Research Institute of Guangxi Power Grid Co Ltd
Priority to CN201810450389.8A priority Critical patent/CN108921812B/en
Publication of CN108921812A publication Critical patent/CN108921812A/en
Application granted granted Critical
Publication of CN108921812B publication Critical patent/CN108921812B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
    • G01B11/00—Measuring arrangements characterised by the use of optical techniques
    • G01B11/16—Measuring arrangements characterised by the use of optical techniques for measuring the deformation in a solid, e.g. optical strain gauge
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • G06T7/251—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments involving models
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30164—Workpiece; Machine component

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Physiology (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Quality & Reliability (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
  • Investigating Strength Of Materials By Application Of Mechanical Stress (AREA)

Abstract

The invention discloses an intelligent evaluation method for a fatigue state of a spring of a circuit breaker based on image recognition, and relates to the technical field of circuit breaker detection. According to the intelligent evaluation method for the fatigue state of the spring of the circuit breaker based on image recognition, a basic database for evaluating the spring state of the circuit breaker is established through multiple off-line test sample statistical analysis and classification, a key motion target position in the spring deformation process is detected and analyzed by adopting a high-speed image sequence and an NCC algorithm, the spring deformation detection in the normal opening and closing process of a high-voltage circuit breaker is realized, a curve representing the fatigue state of the spring is obtained, and therefore a characteristic parameter vector of the fatigue state of the spring is obtained; and analyzing the obtained spring fatigue characteristic parameter vector by adopting a GA-SALBP model to obtain the fatigue state value and the stress relaxation condition of the spring of the circuit breaker, thereby realizing the purpose of monitoring the state of the spring of the high-voltage circuit breaker.

Description

Intelligent evaluation method for fatigue state of breaker spring based on image recognition
Technical Field
The invention belongs to the technical field of circuit breaker detection, and particularly relates to an intelligent evaluation method for a fatigue state of a circuit breaker spring based on image recognition.
Background
The operation reliability of the high-voltage circuit breaker is crucial to the protection and control of a power grid, most faults of the high-voltage circuit breaker are faults of an operating mechanism according to statistics, a circuit breaker spring is used as an important component of the circuit breaker operating mechanism, and the reliability of the circuit breaker spring determines whether the circuit breaker can be normally switched on or off by influencing the circuit breaker operating mechanism. High voltage circuit breaker springs are subjected to varying loads during normal operation, and the failure modes are mostly fatigue failures. The spring is suddenly broken due to long-term fatigue work, and further, the fault accident of the circuit breaker is frequently caused, so that the fatigue state monitoring of the operating mechanism spring is very significant to research.
At present, more monitoring of the high-voltage circuit breaker operating mechanism is focused on the overall mechanical performance of the operating mechanism, a moving contact opening and closing stroke curve, a moving contact opening and closing coil current and the like are measured in a direct or indirect mode, and few researches on a spring monitoring and state evaluation method are carried out. In practice, springs with possible problems can be found only through regular inspection, but the problems of the springs are difficult to judge by naked eyes, human resources are consumed, time and labor are wasted, and efficiency is not high. The fatigue testing machine for the spring is used for calculating a fatigue curve of the spring, the spring of the circuit breaker needs to be tested after being disassembled, and the stress and deformation states of the spring of the operating mechanism cannot be rapidly analyzed in the operation process of the circuit breaker.
Based on the method, the intelligent evaluation method for the fatigue state of the spring of the circuit breaker based on image recognition is provided, and computer vision, pattern recognition, neural network and other methods are applied to the fatigue state monitoring of the spring. The method is a novel non-contact type circuit breaker mechanical state testing method which detects the deformation of the spring of the circuit breaker operating mechanism by using a computer vision technology and dynamically evaluates the performance of the spring in the switching-on and switching-off process. The method is applied to detecting the motion parameters of the breaker spring, particularly, a high-speed camera is used for replacing a traditional displacement and deformation sensor to obtain a motion image sequence of the breaker spring, the motion process of the spring is tracked by an image matching method, finally, relevant working parameters of the motion process of the spring are obtained through curve fitting, characteristic vectors are established according to the deformation characteristics, and the fatigue state and the stress relaxation degree of the spring are judged through intelligent comparison with the spring deformation in a normal state.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides an intelligent evaluation method for the fatigue state of a breaker spring based on image recognition.
The invention solves the technical problems through the following technical scheme: an intelligent evaluation method for a fatigue state of a breaker spring based on image recognition comprises the following steps:
(1) establishing a circuit breaker spring state evaluation basic database;
(2) setting parameters of a testing device for actually measuring the deformation process of the operating spring of the circuit breaker, capturing an image in the deformation process of the spring, and processing the image to obtain a high-speed image sequence of the spring;
(3) taking the pixel coordinate of the fixed end of the spring as an initial point, wherein the spring between the initial point and the top end of the movable end of the spring is a moving target to be identified; selecting a certain area on the moving target as a matching template, and adopting an image matching algorithm to identify each frame of image in the high-speed image sequence in the step (2) to obtain the coordinate of the central pixel position of the corresponding matching template, wherein the pixel position represents the actual working position of the spring of the operating mechanism;
(4) calculating the distance from the central pixel coordinate of the matching template to the pixel coordinate of the initial point in each frame of image, and drawing a displacement-time curve of the moving target in the spring deformation process of the operating mechanism and a speed-time curve of the central pixel coordinate of the matching template;
(5) acquiring spring working parameters through a displacement-time curve and a speed-time curve;
(6) extracting a test sample with obvious representative characteristics from the basic database in the step (1), obtaining working parameters and fatigue state evaluation values of the test sample, and establishing a fatigue characteristic parameter vector representing the fatigue state of the spring of the operating mechanism of the high-voltage circuit breaker; simultaneously acquiring actually measured spring fatigue characteristic parameter vectors;
(7) the fatigue state of the spring was analyzed using genetic algorithm-adaptive learning rate back propagation (GA-SALBP): extracting a typical sample in a basic database, solving a fatigue characteristic parameter vector of the typical sample, taking the fatigue characteristic parameter vector of the typical sample as a training sample, and training a GA-SALBP model; and substituting the spring fatigue characteristic parameter vector of the unknown fatigue state into the trained GA-SALBP model to obtain the fatigue state value and the stress relaxation condition of the spring of the breaker, and finally setting a fatigue accident early warning threshold value to realize the purpose of monitoring the state of the spring of the high-voltage breaker.
Further, the inherent size of the spring is recorded in the basic database in the step (1) according to the model of the breaker spring, and meanwhile, spring fatigue test basic data and the relation between the telescopic deformation process of the breaker spring and the pressure and the fatigue degree are recorded.
Further, the testing device in the step (2) is a high-speed camera, the parameters required to be set by the high-speed camera include a shooting focal length, a trigger speed, a camera frame rate, a resolution and exposure time, the shooting duration of the high-speed camera is set according to the model of the circuit breaker and the opening and closing time, and the actual size of each pixel of the shot image of the high-speed camera in the actual shooting plane is determined and calibrated.
Further, the method for determining the actual size of each pixel of the shot image in the actual shooting plane is as follows: a scale with known specific size is placed in the same plane with a shot breaker spring, the pixel length of the scale in an image is obtained, and the quotient of the actual size of the scale and the pixel length is used as the actual size of each pixel in the actual shooting plane of the image.
Further, the working parameters of the spring in the step (5) comprise the initial working height of the spring, the final working height of the spring, the expansion and contraction time, the maximum movement stroke of the spring, the radial reciprocating times of a middle ring of the spring and the maximum acceleration of the working of the spring;
taking the distance from the central pixel coordinate of the initial frame matching template to the initial point pixel coordinate as the initial working height of the spring, taking the distance from the central pixel coordinate of the ending frame matching template to the initial point pixel coordinate as the final working height of the spring, and taking the difference value between the upper bound and the lower bound of the displacement-time curve as the maximum movement stroke h of the springm。
Furthermore, an initial frame and an end frame are determined by adopting a gray level difference method of corresponding positions of adjacent frames of the image.
Further, the spring initial working height Ht0Is calculated by the expression oft0＝k|s0S, spring final working height HtIs calculated byHas the expression of Ht＝k|st-s, spring expansion time t is y ═ stMaximum of the intersection of ± epsilon and the displacement-time curve;
wherein s is0Matching the central pixel coordinates of the template for the initial frame, s being the initial point pixel coordinates, stMatching the template center pixel coordinates for the end frame, k being the actual size of each pixel of the image in the actual capture plane, and ε being the maximum value of the assumed spring expansion and contraction amplitude.
Further, the specific process of analyzing the spring fatigue state by adopting the GA-SALBP model in the step (7) comprises the following steps:
(7.1) setting an N-dimensional vector x which is input as a spring fatigue state characteristic parameter, setting M samples in a basic database, calculating the mean value of each characteristic parameter, comparing the mean value with the characteristic value of each sample, eliminating samples with obvious abnormity of partial characteristic parameters caused by the influence of environmental factors or recognition and recording errors, and leaving proper M ' samples to form a characteristic parameter matrix of the N multiplied by M ' circuit breaker operating mechanism spring, M ' fatigue state degree evaluation values and stress relaxation degrees;
(7.2) establishing a BP (back propagation) neural network, determining the number of neurons in the middle layer according to the number of characteristic parameters of the fatigue state of the spring, wherein the weight number of the BP neural network is the product of the number N of the characteristic parameters of the spring plus the number of the evaluation indexes and the number of neurons in the middle layer; obtaining a group of complete weights with small errors of the BP neural network by using a GA genetic algorithm, taking the group of complete weights as initial weights of the BP neural network, substituting sample values, namely a characteristic parameter matrix of a spring, and learning and training the BP neural network;
(7.3) in the training process, the weight is adjusted according to the spring performance evaluation error energy and the learning rate, and the learning rate is adjusted by using the change of the error energy after the weight is adjusted twice; the change speed of the learning rate is adjusted according to the requirement, and the method has the advantages that the increase and decrease judgment of errors is avoided, the increase/decrease factor is not needed to be set, the learning rate is reduced when the errors are increased, and the learning rate is properly improved in the error decrease direction;
(7.4) when the error energy is smaller than a set threshold value, namely the result obtained by inputting the fatigue state characteristic parameter of any one spring sample is very close to the fatigue state estimated value of the sample actually, completing the training of the BP neural network; at the moment, the movement process of the spring of the operating mechanism to be tested is tracked through image recognition, the fatigue state characteristic parameter vector is calculated and substituted into the trained BP neural network, the fatigue state of the spring is analyzed, and the purpose of intelligently evaluating the fatigue state of the spring of the operating mechanism of the circuit breaker is achieved.
Further, the adjustment of the learning rate in the step (7.3) is expressed by
Figure BDA0001658299830000051
The value of k is a constant greater than 1.
Compared with the prior art, the intelligent evaluation method for the fatigue state of the spring of the circuit breaker based on image recognition, provided by the invention, comprises the steps of establishing a basic database for evaluating the spring state of the circuit breaker by statistical analysis and classification of a plurality of off-line test samples, detecting and analyzing the position of a key motion target in the spring deformation process by adopting a high-speed image sequence and an NCC algorithm, realizing the detection of the spring deformation in the normal opening and closing process of the high-voltage circuit breaker, obtaining a curve representing the fatigue state of the spring, and further obtaining a characteristic parameter vector of the fatigue state of the spring; analyzing the obtained spring fatigue characteristic parameter vector by adopting a GA-SALBP model to obtain a fatigue state value and a stress relaxation condition of the spring of the circuit breaker, and realizing the purpose of monitoring the state of the spring of the high-voltage circuit breaker; when the GA-SALBP model is adopted to analyze the obtained spring fatigue characteristic parameter vector, the evaluation on the performance of the test spring is realized by combining with the evaluation error energy index of the circuit breaker spring performance, so that the whole evaluation method is more effective and the calculation speed is higher.
Drawings
In order to more clearly illustrate the technical solution of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only one embodiment of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on the drawings without creative efforts.
FIG. 1 is a schematic flow chart of an intelligent evaluation method for fatigue state of a breaker spring based on image recognition, which is disclosed by the invention;
FIG. 2 is a schematic diagram depicting the spring parameters of the present invention;
FIG. 3 is a schematic flow chart of the GA-SALBP neural network for determining the fatigue state and stress relaxation degree of the spring of the operating mechanism according to the present invention.
Detailed Description
The technical solutions in the present invention are clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, the intelligent evaluation method for the fatigue state of the spring of the circuit breaker based on image recognition provided by the invention comprises the following steps:
(1) and establishing a basic database for evaluating the state of the spring of the circuit breaker.
The inherent sizes of the spring are recorded in the basic database according to the model of the breaker spring, and the inherent sizes comprise the diameter D of the spring, the middle diameter D of the spring and the inner diameter D of the spring1Outer diameter D2Free height H0Helix angle α, original pitch b, number of turns n of the spring (some parameters can be derived from each other); meanwhile, the spring fatigue test basic data and the relation between the telescopic deformation process of the spring of the circuit breaker and the pressure and the fatigue degree are recorded, wherein the relation comprises key parameters for describing the performance of the spring, such as the working height of the spring, the movement speed, the stress telescopic deformation curve, the oscillation frequency and the like. The data samples in the basic database are obtained by simulating the on-off state of the breaker on site through a spring fatigue tester, shooting and analyzing and identifying the spring stretching process characteristics through a high-speed camera, combining the test results of a mechanical sensor and carrying out repeated tests on the similar springs in a fatigue tester, and the data samples are stored in the basic database after being subjected to statistical classification, namely the breaker spring state evaluation basic database.
(2) The method comprises the steps of setting parameters of a testing device for actually measuring the deformation process of the circuit breaker operating spring, capturing images of the spring in the deformation process, and processing to obtain a high-speed image sequence of the spring.
The testing device is a high-speed camera, the high-speed camera is installed in a proper position near the spring of the circuit breaker to be tested, and the parameter of the high-speed camera is adjusted to enable the spring image of the operating mechanism to be clear; the parameters required to be set by the high-speed camera comprise a shooting focal length, a trigger speed, a camera frame rate, a resolution and exposure time, the shooting duration of the high-speed camera is set according to the model of the circuit breaker and the opening and closing time, and the actual size k (mm) of each pixel of a shooting image of the high-speed camera in an actual shooting plane is determined and calibrated.
The method for determining the actual size k comprises the following steps: the method comprises the steps of placing a scale with known specific size in the same plane as a shot breaker spring, obtaining the pixel length of the scale in an image, taking the quotient of the actual size of the scale and the pixel length as the actual size k of each pixel in the actual shooting plane of the image, facilitating calculation of final spring working parameters, and comparing the final spring working parameters with spring free state parameters to enable results to be more visual.
When the current trigger of the breaker opening and closing coil exceeds a threshold value, the moment when the breaker is opened and closed is judged, the high-speed camera is controlled to start capturing images, and a high-speed image sequence of the high-voltage breaker spring is obtained through image acquisition, preprocessing and storage.
(3) Searching an initial frame of spring movement in a high-speed image sequence, taking the pixel coordinate of the fixed end of the spring as an initial point A for calculating the actual working length and the movement speed change of the spring in the deformation process, taking the adjacent area (namely a matching template) of the top end of the movable end of the spring as a target point B, and taking the spring between the initial point and the target point as a movement target to be identified (as shown in FIG. 2); and distinguishing the motion of the breaker in the opening and closing process through the target point B, wherein the motion is the obvious characteristic internode of the spring of the breaker operating mechanism.
And (3) identifying the central pixel position (namely the pixel position of the spring characteristic internode in each frame image) of the spring top end adjacent region (namely the matching template) of each frame image in the high-speed image sequence in the step (2) by adopting an image matching algorithm (NCC algorithm), wherein the pixel position represents the actual working position of the spring of the operating mechanism.
(4) And calculating the distance from the pixel coordinate of the target point B to the pixel coordinate of the initial point A (namely the working height of the spring) in each frame of image, drawing a displacement-time curve of the moving target in the deformation process of the spring of the operating mechanism and a speed-time curve of the target point B, wherein the speed value is the point-by-point slope value of the displacement-time curve of the spring target point B in the moving direction.
(5) Acquiring spring working parameters through a displacement-time curve and a speed-time curve, wherein the spring working parameters comprise initial spring working height, final spring working height, stretching time, maximum spring movement stroke, radial reciprocating times of a middle ring of a spring and maximum acceleration of spring working; with the pixel coordinates s of the target point in the initial frame0The distance from the pixel coordinate s of the initial point is taken as the initial working height H of the springt0To end frame target point pixel coordinate stThe distance from the pixel coordinate s of the initial point is taken as the final working height H of the springtObtaining the expansion time t of the spring through a displacement-time curve, and taking the difference value of the upper bound and the lower bound of the displacement-time curve as the maximum movement stroke h of the springmCalculating the radial reciprocating times f of the middle ring of the spring in the expansion timec(i.e. the direction change times of the target point B on the spring in the spring expansion and contraction time), and the maximum acceleration a of the spring work is obtainedm。
Initial working height H of springt0Is calculated by the expression oft0＝k|s0S, spring final working height HtIs calculated by the expression oft＝k|st-s, spring expansion time t is y ═ stMaximum of the intersection of ± epsilon and the displacement-time curve; epsilon is the maximum value of the determined spring expansion and contraction amplitude, epsilon defaults to 0-2 pixels, action caused by the fact that the spring expansion and contraction amplitude is too small and possibly caused by recognition errors is eliminated, and k is a coefficient and reflects the calculation relation between the actual size of the spring and the pixel length.
The initial frame and the end frame are determined by adopting a gray level difference method of corresponding positions of adjacent frames of the image, a large number of irrelevant static frame images are filtered, and the influence of a large number of redundant images on the processing speed before and after the operation of the breaker is avoided.
(6) Extracting a test sample with obvious representative characteristics from the basic database in the step (1) to obtain working parameters and fatigue state evaluation values of the test sample; through main component extraction, the correlation of data among test samples is analyzed to obtain spring working parameters which are possibly changed due to the change of the fatigue state of the spring, a fatigue characteristic parameter vector representing the fatigue state of the spring of the operating mechanism of the high-voltage circuit breaker is established, and the complexity of the parameters is reduced; simultaneously obtaining actually measured spring fatigue characteristic parameter vector hm,am,fc]；
(7) The fatigue state of the spring was analyzed using genetic algorithm-adaptive learning rate back propagation (GA-SALBP): extracting a typical sample in a basic database, solving a fatigue characteristic parameter vector of the typical sample, taking the fatigue characteristic parameter vector of the typical sample as a training sample, and training a GA-SALBP model; and substituting the spring fatigue characteristic parameter vector of the unknown fatigue state into the trained GA-SALBP model to obtain a fatigue state value (a value between 0 and 1) and a stress relaxation condition of the breaker spring, and finally setting a fatigue accident early warning threshold value (which can be adjusted according to actual working requirements) as an early warning value of the occurrence of the spring fatigue fracture accident, thereby realizing the purpose of monitoring the state of the high-voltage breaker spring.
As shown in FIG. 3, the specific process of analyzing the fatigue state of the spring by using the GA-SALBP model comprises the following steps:
(7.1) setting N-dimensional vector x which is input as characteristic parameters of the fatigue state of the spring, setting M samples in a basic database, calculating the mean value of each characteristic parameter, comparing the mean value with the characteristic values of the samples, eliminating samples with obvious abnormity of partial characteristic parameters caused by the influence of environmental factors or recognition and recording errors, and leaving proper M ' samples to form a characteristic parameter matrix of the N multiplied by M ' circuit breaker operating mechanism spring, M ' fatigue state degree evaluation values and stress relaxation degrees.
(7.2) establishing a three-layer BP neural network, determining the number of neurons in the middle layer according to the number of characteristic parameters of the fatigue state of the spring, wherein the weight number of the BP neural network is the product of the number N of the characteristic parameters of the spring plus the number of evaluation indexes and the number of neurons in the middle layer; and obtaining a group of complete weights with smaller errors of the BP neural network by using a GA genetic algorithm to serve as initial weights of the BP neural network, and substituting the sample values, namely the characteristic parameter matrix of the spring, to learn and train the BP neural network.
(7.3) evaluating error energy according to spring performance during training
Figure BDA0001658299830000091
(i.e. sum of squares of error values of estimated value of degree of fatigue of actual spring and stress relaxation
Figure BDA0001658299830000092
ej(n) is the estimated value of the degree of fatigue state of the spring and the error of stress relaxation) for the jth time and the learning rate eta, the learning rate eta is adjusted by the change of the error energy after the weight is adjusted for the previous and subsequent times, and the adjustment expression of the learning rate is
Figure BDA0001658299830000093
The value of p is a constant larger than 1, eta (n) is the learning rate in the nth training process, the learning rate is the adjusted learning rate in the (n +1) th training process, and e (n) and e (n +1) are recursive error energy for evaluating the performance of the spring for two times of training; the change speed of the learning rate is adjusted according to the requirement, and the method has the advantages that the increase and decrease judgment of errors is avoided, the increase/decrease factor is not needed to be set, the learning rate is reduced when the errors are increased, and the learning rate is properly improved in the error decrease direction.
(7.4) when the error energy is smaller than a set threshold value, namely the result obtained by inputting the fatigue state characteristic parameter of any one spring sample is very close to the fatigue state estimated value of the sample actually, completing the training of the BP neural network; at the moment, the movement process of the spring of the operating mechanism to be tested is tracked through image recognition, the fatigue state characteristic parameter vector is calculated and substituted into the trained BP neural network, the fatigue state of the spring is analyzed, and the purpose of intelligently evaluating the fatigue state of the spring of the operating mechanism of the circuit breaker is achieved.
The initial weight and the threshold value of the BP neural network are obtained by utilizing the genetic algorithm, so that the convergence speed of the BP neural network is accelerated on the premise of ensuring the calculation precision of the BP neural network, and the BP neural network has better precision and relatively higher convergence speed by the training method of the self-adaptive learning rate.
The above disclosure is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of changes or modifications within the technical scope of the present invention, and shall be covered by the scope of the present invention.

Claims (8)

1. An intelligent evaluation method for a fatigue state of a breaker spring based on image recognition is characterized by comprising the following steps:
(1) establishing a circuit breaker spring state evaluation basic database;
(2) setting parameters of a testing device for actually measuring the deformation process of the operating spring of the circuit breaker, capturing an image in the deformation process of the spring, and processing the image to obtain a high-speed image sequence of the spring;
(3) taking the pixel coordinate of the fixed end of the spring as an initial point, wherein the spring between the initial point and the top end of the movable end of the spring is a moving target to be identified; selecting a certain area on the moving target as a matching template, and adopting an image matching algorithm to identify each frame of image in the high-speed image sequence in the step (2) to obtain the position coordinates of the central pixel of the corresponding matching template;
(4) calculating the distance from the central pixel coordinate of the matching template to the pixel coordinate of the initial point in each frame of image, and drawing a displacement-time curve of the moving target in the spring deformation process of the operating mechanism and a speed-time curve of the central pixel coordinate of the matching template;
(5) acquiring spring working parameters through a displacement-time curve and a speed-time curve;
(6) extracting a test sample with obvious representative characteristics from the basic database in the step (1), obtaining working parameters and fatigue state evaluation values of the test sample, and establishing a fatigue characteristic parameter vector representing the fatigue state of the spring of the operating mechanism of the high-voltage circuit breaker; simultaneously acquiring actually measured spring fatigue characteristic parameter vectors;
(7) the fatigue state of the spring was analyzed using the GA-SALBP model: extracting a typical sample in a basic database, solving a fatigue characteristic parameter vector of the typical sample, taking the fatigue characteristic parameter vector of the typical sample as a training sample, and training a GA-SALBP model; then substituting the spring fatigue characteristic parameter vector of the unknown fatigue state into the trained GA-SALBP model to obtain the fatigue state value and the stress relaxation condition of the breaker spring, and finally setting a fatigue accident early warning threshold value to realize the purpose of monitoring the state of the high-voltage breaker spring;
the specific process of analyzing the spring fatigue state by adopting the GA-SALBP model in the step (7) is as follows:
(7.1) setting an N-dimensional vector x which is input as a spring fatigue state characteristic parameter, setting M samples in a basic database, calculating the mean value of each characteristic parameter, comparing the mean value with the characteristic value of each sample, eliminating samples with obvious abnormity of partial characteristic parameters caused by the influence of environmental factors or recognition and recording errors, and leaving proper M ' samples to form a characteristic parameter matrix of the N multiplied by M ' circuit breaker operating mechanism spring, M ' fatigue state degree evaluation values and stress relaxation degrees;
(7.2) establishing a BP (back propagation) neural network, determining the number of neurons in the middle layer according to the number of characteristic parameters of the fatigue state of the spring, wherein the weight number of the BP neural network is the product of the number N of the characteristic parameters of the spring plus the number of the evaluation indexes and the number of neurons in the middle layer; obtaining a group of complete weights with small errors of the BP neural network by using a GA genetic algorithm, taking the group of complete weights as initial weights of the BP neural network, substituting sample values, namely a characteristic parameter matrix of a spring, and learning and training the BP neural network;
(7.3) in the training process, the weight is adjusted according to the spring performance evaluation error energy and the learning rate, and the learning rate is adjusted by using the change of the error energy after the weight is adjusted twice;
(7.4) when the error energy is smaller than a set threshold value, namely the result obtained by inputting the fatigue state characteristic parameter of any one spring sample is very close to the fatigue state estimated value of the sample actually, completing the training of the BP neural network; at the moment, the movement process of the spring of the operating mechanism to be tested is tracked through image recognition, the fatigue state characteristic parameter vector is calculated and substituted into the trained BP neural network, the fatigue state of the spring is analyzed, and the purpose of intelligently evaluating the fatigue state of the spring of the operating mechanism of the circuit breaker is achieved.
2. The intelligent evaluation method for the fatigue state of the spring of the circuit breaker based on the image recognition as claimed in claim 1, wherein the basic database of the step (1) records the inherent size of the spring according to the model of the spring of the circuit breaker, and records the basic data of the fatigue test of the spring, and the relation between the expansion and contraction deformation process of the spring of the circuit breaker and the pressure and fatigue degree.
3. The intelligent evaluation method for the fatigue state of the spring of the circuit breaker based on the image recognition as claimed in claim 1, wherein the testing device in the step (2) is a high-speed camera, the parameters to be set by the high-speed camera include a shooting focal length, a trigger speed, a camera frame rate, a resolution and an exposure time, the shooting duration of the high-speed camera is set according to the model of the circuit breaker and the switching-on and switching-off time, and the actual size of each pixel of the shot image of the high-speed camera in an actual shooting plane is determined and calibrated.
4. The intelligent evaluation method for the fatigue state of the spring of the circuit breaker based on the image recognition as claimed in claim 3, wherein the method for determining the actual size of each pixel of the shot image in the actual shooting plane is as follows: a scale with known specific size is placed in the same plane with a shot breaker spring, the pixel length of the scale in an image is obtained, and the quotient of the actual size of the scale and the pixel length is used as the actual size of each pixel in the actual shooting plane of the image.
5. The intelligent evaluation method for the fatigue state of the spring of the circuit breaker based on the image recognition as claimed in claim 1, wherein the spring working parameters in the step (5) comprise the initial working height of the spring, the final working height of the spring, the expansion and contraction time, the maximum movement stroke of the spring, the radial reciprocating times of the middle ring of the spring and the maximum acceleration of the working of the spring;
taking the distance from the central pixel coordinate of the initial frame matching template to the initial point pixel coordinate as the initial working height of the spring, taking the distance from the central pixel coordinate of the ending frame matching template to the initial point pixel coordinate as the final working height of the spring, and taking the difference value between the upper bound and the lower bound of the displacement-time curve as the maximum movement stroke h of the springm。
6. The intelligent evaluation method for the fatigue state of the spring of the circuit breaker based on the image recognition as claimed in claim 5, wherein the initial frame and the end frame are determined by a gray level difference method of corresponding positions of adjacent frames of the image.
7. The intelligent evaluation method for the fatigue state of the spring of the circuit breaker based on the image recognition as claimed in claim 5, wherein the initial working height H of the spring ist0Is calculated by the expression oft0＝k|s0S, spring final working height HtIs calculated by the expression oft＝k|st-s, spring expansion time t is y ═ stMaximum of the intersection of ± epsilon and the displacement-time curve;
wherein s is0Matching the central pixel coordinates of the template for the initial frame, s being the initial point pixel coordinates, stMatching the template center pixel coordinates for the end frame, k being the actual size of each pixel of the image in the actual capture plane, and ε being the maximum value of the assumed spring expansion and contraction amplitude.
8. The intelligent evaluation method for the fatigue state of the spring of the circuit breaker based on the image recognition as claimed in claim 1, wherein the adjustment expression of the learning rate in the step (7.3) is
Figure FDA0003305612880000041
p is a constant greater than 1, e (n) and e (n +1) are recursive two training roundsAnd evaluating error energy of the spring performance.
CN201810450389.8A 2018-05-11 2018-05-11 Intelligent evaluation method for fatigue state of breaker spring based on image recognition Active CN108921812B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810450389.8A CN108921812B (en) 2018-05-11 2018-05-11 Intelligent evaluation method for fatigue state of breaker spring based on image recognition

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810450389.8A CN108921812B (en) 2018-05-11 2018-05-11 Intelligent evaluation method for fatigue state of breaker spring based on image recognition

Publications (2)

Publication Number Publication Date
CN108921812A CN108921812A (en) 2018-11-30
CN108921812B true CN108921812B (en) 2022-04-22

Family

ID=64403753

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810450389.8A Active CN108921812B (en) 2018-05-11 2018-05-11 Intelligent evaluation method for fatigue state of breaker spring based on image recognition

Country Status (1)

Country Link
CN (1) CN108921812B (en)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109655248A (en) * 2018-12-18 2019-04-19 安徽江淮汽车集团股份有限公司 Solenoid valve response time test method
CN112097673B (en) * 2019-06-18 2022-11-15 上汽通用汽车有限公司 Virtual matching method and system for vehicle body parts
CN111067531A (en) * 2019-12-11 2020-04-28 中南大学湘雅医院 Wound measuring method and device and storage medium
CN111649883B (en) * 2020-05-25 2022-04-05 河北金力新能源科技股份有限公司 Evaluation method of circular knife spring of splitting machine and circular knife spring elasticity testing device used in evaluation method
CN112798242A (en) * 2020-12-25 2021-05-14 合肥开关厂有限公司 Method for calculating spring fatigue degree of vacuum circuit breaker
CN113791339B (en) * 2021-07-19 2024-07-19 国网浙江省电力有限公司乐清市供电公司 Circuit breaker performance state detection method based on R-NCC image recognition algorithm
CN115356095B (en) * 2022-08-16 2024-09-24 广西电网有限责任公司电力科学研究院 Method for evaluating state performance of opening and closing spring of circuit breaker based on variance of motion equation

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101102004B1 (en) * 2005-12-30 2012-01-04 삼성전자주식회사 A method and system for quantitating fatigue resulting from a three dimensional display
US9390517B2 (en) * 2011-01-25 2016-07-12 Samsung Sdi Co., Ltd. System for image analysis and method thereof
CN102693450B (en) * 2012-05-16 2014-11-12 北京理工大学 A prediction method for crankshaft fatigue life based on genetic nerve network
CN103839412B (en) * 2014-03-27 2015-12-02 北京建筑大学 A kind of crossing dynamic steering ratio combination method of estimation based on Bayes's weighting
CN104062111A (en) * 2014-06-13 2014-09-24 云南电力试验研究院（集团）有限公司电力研究院 Method for collecting mechanical characteristic parameters of breaker based on high-speed camera
CN106651906A (en) * 2015-10-30 2017-05-10 国网山西省电力公司电力科学研究院 Test method for motion characteristic of high-voltage breaker based on improved Gaussian mixture model
CN107860564B (en) * 2017-09-25 2019-10-18 广西电网有限责任公司电力科学研究院 A kind of flaccid state online testing device for high-voltage circuitbreaker operation spring
CN107860562B (en) * 2017-09-25 2019-06-14 广西电网有限责任公司电力科学研究院 A kind of high-voltage circuitbreaker operation spring weakness test method
CN107884164B (en) * 2017-09-25 2019-07-26 广西电网有限责任公司电力科学研究院 A kind of breaker spring method for testing performance of NCC-SAR algorithm

Also Published As

Publication number Publication date
CN108921812A (en) 2018-11-30

Similar Documents

Publication Publication Date Title
CN108921812B (en) Intelligent evaluation method for fatigue state of breaker spring based on image recognition
Pan et al. A new image recognition and classification method combining transfer learning algorithm and mobilenet model for welding defects
CN111259930B (en) General target detection method of self-adaptive attention guidance mechanism
CN110598736B (en) Power equipment infrared image fault positioning, identifying and predicting method
CN112766103B (en) Machine room inspection method and device
CN107860562A (en) A kind of primary cut-out operates spring weakness method of testing
CN108984893A (en) A kind of trend forecasting method based on gradient method for improving
CN108287303A (en) A kind of breaker mechanic property scene charged test method based on NCC-P-S optimization algorithms
CN106651906A (en) Test method for motion characteristic of high-voltage breaker based on improved Gaussian mixture model
CN117451744B (en) Method, device, equipment and storage medium for detecting defect of infrared lens
CN118154997B (en) Insulator quality detection method
CN117056814B (en) Transformer voiceprint vibration fault diagnosis method
CN103310191B (en) The human motion recognition method of movable information image conversion
CN117152735A (en) Tomato maturity grading method based on improved yolov5s
CN117764980A (en) Automatic identification and measurement method for defects of composite material based on infrared multi-feature fusion
CN111340748A (en) Battery defect identification method and device, computer equipment and storage medium
CN117516939A (en) Bearing cross-working condition fault detection method and system based on improved EfficientNetV2
CN113762131A (en) Method and terminal for detecting deformation position of oil casing double-layer pipe column of high-sulfur-content gas field
CN113343550A (en) Partial discharge fault diagnosis method based on local image characteristics
CN109166122A (en) Circuit breaker operation mechanism telescopic spring characteristic test method based on image procossing
CN106803080B (en) Complementary pedestrian detection method based on shape Boltzmann machine
CN117790353B (en) EL detection system and EL detection method
CN118469102B (en) Dynamic equipment residual life prediction method based on data driving
Shi et al. Indoor Fall Detection Based on Yolov5 And Openpose
CN118116061B (en) Image processing system based on personnel identification

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant