A method, device and equipment for diagnosing loose core fault of a reactor

The gradient enhancement decision tree algorithm handles vibration signals, which solves the problem of loose faults of high-voltage parallel reactors being difficult to detect, achieves fast and accurate fault diagnosis, and improves the safe and stable operation of the equipment.

CN113723496BActive Publication Date: 2024-11-26POWER RES INST OF STATE GRID SHAANXI ELECTRIC POWER CO LTD +2
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202110990077.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-08-26
Publication Date
2024-11-26
Estimated Expiration
2041-08-26

AI Technical Summary

Technical Problem

The faults caused by loose fastening structure during long-term operation of the high-voltage parallel reactor are difficult to detect, affecting the safe and stable operation of the equipment.

Method used

The reactor core loosening fault diagnosis method is used based on the gradient enhancement decision tree algorithm. By collecting and processing vibration signals, an initial prediction model is constructed to determine whether the reactor is in a nut loose fault and predict the degree of looseness.

Benefits of technology

It realizes rapid and accurate diagnosis of loose faults of high-voltage parallel reactors, reduces operating and maintenance costs, extends equipment life, and prevents serious accidents.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113723496B_ABST
    Figure CN113723496B_ABST
Patent Text Reader

Abstract

The present invention discloses a method, device and equipment for diagnosing the loose core fault of a reactor. The vibration measuring device is used to collect vibration signals of a high-voltage shunt reactor, the surface of the reactor oil tank is selected as the measurement position, the collected signals are preprocessed, and the preprocessed vibration signals are model predicted to detect whether the reactor core has a loose fault. The GBDT classification algorithm is used to effectively classify the collected vibration signals into appropriate labels, and the classification speed is fast and the accuracy is high. When collecting signals, the requirements for the vibration position and the load size are not high, so that it is possible to make a preliminary judgment on the occurrence of a loose nut fault of a high-voltage shunt reactor and predict the degree of looseness.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of circuit equipment fault diagnosis, and in particular relates to a reactor core fault diagnosis method, device and equipment. Background Art

[0002] In the current environment of increasing voltage levels and rapid development of the market economy, power companies have increased their requirements for the safe and economical operation of power transmission and transformation equipment. High-voltage shunt reactors are a key link in the power system. If their reliability fluctuates, it will directly affect the safety and stability of the power system. According to statistics, due to the high voltage level and large capacity of high-voltage shunt reactors, the vibration caused by long-term operation is prone to loosening of structural parts, among which the fastening clamps that constrain the iron core are the parts that are prone to loosening. Due to its special structure such as the multi-air gap iron core structure, the high-voltage shunt reactor generates stronger vibration in the alternating magnetic field. During the vibration, the connecting structural parts such as the iron core cake and the fastening clamps, the winding and the pressure plate will produce axial and radial relative displacement and a certain degree of deformation, causing bolt loosening, insulation material aging and other faults. Therefore, the vibration of each component of the reactor to a certain extent represents its mechanical state. Actively carrying out the diagnosis of reactor loosening faults in order to timely discover and take corresponding measures can not only reduce the operation and maintenance costs and extend the life of the equipment, but also prevent the occurrence of serious accidents, bringing good social and economic benefits. By analyzing the vibration signal, the operating status of the reactor can be monitored and evaluated, so that potential faults can be discovered in advance and checked in time, which is of great significance to the safe and stable operation of the power system.

[0003] At present, the research on reactors by scholars at home and abroad is mainly focused on the development of noise reduction and vibration reduction devices, chromatographic analysis of insulating oil and electromagnetic field analysis, etc., while there are few studies on reactor vibration. The vibration signal analysis method can not only diagnose the winding and core conditions at the same time, but also has no electrical connection with the entire power system, and can conveniently, safely and reliably realize live detection. At present, some studies on reactor vibration include three-dimensional modeling of the reactor core based on the finite element analysis software COMSOL, simulation calculation of the electromagnetic force of the core and the force of steel structures such as clamps; the establishment of a three-dimensional temperature rise calculation model of the reactor based on electromagnetic, fluid and temperature, and the study of the temperature distribution characteristics of the reactor as a whole and the internal windings; the analysis of the vibration mechanism of the reactor, the introduction of two vibration signal testing methods, and the spectrum analysis of the measured vibration signal and the analysis of the vibration signal based on the measured UHV shunt reactor, clarifying the vibration spectrum and the distribution characteristics of the oil tank surface. It can be seen that the vibration method has been a hot topic of research at home and abroad in recent years. The vibration signal collected on the surface of the high-voltage shunt reactor will reflect the current health status of the equipment after processing. The vibration signal generated when a fault occurs will be different from that under normal circumstances. If the equipment can be judged in what state it is, it can be reported to the researcher in time for maintenance to avoid more serious accidents and economic losses. Summary of the invention

[0004] The present invention provides a method, device and equipment for diagnosing the loose core fault of a reactor, which solves the technical problem that it is difficult to detect the abnormality after the fastening structure of the high-voltage shunt reactor inevitably loosens during long-term operation.

[0005] To achieve the above object, the present invention provides a method for diagnosing a loose core fault of a reactor based on a gradient boosting decision tree algorithm, comprising the following steps:

[0006] S1. Collect the vibration source signal of the high-voltage shunt reactor in working state, extract the key data features in the vibration source signal; construct the initial prediction model using GBDT algorithm;

[0007] S2. Set the classification labels for the extracted data features as training sets to train the initial prediction model. During the training process, the prediction model continuously learns the features contained in the data. Each iteration is trained based on the residual value of the previous round of training model. After multiple iterations of residual calculation and weighting, the strong learner learns the features contained in the vibration signal of the reactor under different states, and obtains the target prediction model.

[0008] S3. Collect the vibration signal of the reactor to be diagnosed, input the collected vibration signal into the target prediction model, determine whether the current reactor is in a state of nut loosening fault, and predict the degree of looseness of the reactor.

[0009] Furthermore, in S1, the initial prediction model constructed using the GBDT algorithm includes the following steps:

[0010] S A1, initialize the threshold and weight of the first weak learner;

[0011] S A2, establish M classification and regression trees;

[0012] SA3. Calculate the negative gradient of the loss function corresponding to the regression tree, and sort the regression trees in descending order of negative gradient to obtain a strong learner.

[0013] Furthermore, in SA2, the CART regression tree is used to fit the data, and the negative gradient r of the loss function is m,i Fit a regression tree and get its corresponding leaf node region R m,j , j = 1, 2, ..., J, J is the number of leaf nodes in the regression tree; then calculate the leaf area j of each leaf node in the generated decision tree, and calculate the best residual fitting value c m,j , using the best residual fit value c m,j Get a strong learner;

[0014]

[0015] Among them, f m-1 (x i ) is the model obtained by the m-1th calculation.

[0016] Furthermore, in S1, a piezoelectric acceleration vibration sensor is used to collect the vibration source signal of the reactor.

[0017] Furthermore, in S1, the collected vibration source signal is subjected to fast Fourier transform and wavelet denoising to extract key data features.

[0018] Furthermore, in S1, the key data feature is the fundamental frequency of the vibration signal.

[0019] Furthermore, in S2, the extracted data features are divided into a test set and a training set in a ratio of 1:4 using a cross-validation method.

[0020] Furthermore, in S2, the labels include normal state, 11N.m, 7N.m and 3N.m.

[0021] A reactor core loose fault diagnosis device comprises a vibration sensor, an acquisition card and a terminal connected in sequence, wherein the vibration sensor is used to measure a vibration source signal of the reactor, the acquisition card is used to acquire the vibration source signal and transmit it to the terminal, and the terminal is used to determine whether the reactor is in a nut loose fault condition according to the vibration source signal, and to predict the degree of looseness of the reactor.

[0022] A computer device comprises an electrically connected memory and a processor, wherein the memory stores a calculation program that can be run on the processor, and when the processor executes the calculation program, the steps of the above-mentioned reactor core loosening fault diagnosis method are implemented.

[0023] Compared with the prior art, the present invention has at least the following beneficial technical effects:

[0024] The vibration detection device is used to collect vibration signals of the high-voltage shunt reactor to determine the current working condition of the reactor. The gradient boosting decision tree (GBDT) is selected for the training of the prediction model. The additive model and the forward step-by-step algorithm are used to optimize the learning. Multiple weak classifiers are superimposed in series, the weights are increased layer by layer, and the residual under the current model is calculated. Each round of GBDT training is based on the negative gradient value of the previous round of training model. Finally, each classifier is weighted to give a prediction value to achieve the classification effect. Multiple weak learners constitute a strong learner, and the classification results are analyzed to determine the fault state of the reactor, which is of great significance for the fault diagnosis of high-voltage shunt reactors.

[0025] The GBDT algorithm is used to train features with multiple decision trees as weak learners. The resulting strong learner effectively classifies the collected vibration signals into appropriate labels. Since the trees can be calculated in parallel and have good generalization and expression capabilities, the classification is fast and accurate. When collecting signals, the requirements for vibration position and load size are not high, making it possible to make a preliminary judgment on the occurrence of loose nut failure in high-voltage shunt reactors and predict the degree of looseness.

[0026] Furthermore, in S2, the extracted data features are divided into a test set and a training set in a ratio of 1:4 using a cross-validation method, so as to ensure that the trained model has good accuracy when the amount of collected data is not sufficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a flow chart of the fault diagnosis method for the loose core of the reactor;

[0028] Figure 2a is the fundamental frequency proportion of channel 1 at a certain measurement point;

[0029] Figure 2b is the fundamental frequency proportion of channel 2 at a certain measurement point;

[0030] Figure 2c is the fundamental frequency proportion of the three channels at a certain measurement point;

[0031] Figure 2dis the fundamental frequency proportion of the 4 channels at a certain measurement point;

[0032] Figure 3 is the fundamental frequency proportion of different measuring points under the same state;

[0033] Figure 4 This is a schematic diagram of a reactor core loose fault diagnosis device;

[0034] Figure 5 A schematic diagram of the structure of a computer device provided by the present invention. DETAILED DESCRIPTION

[0035] In order to make the purpose and technical solution of the present invention clearer and easier to understand, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. The specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0036] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first" and "second" are used only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more. In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0037] In practical applications, looseness caused by vibration is not easy to detect, and long-term operation in a loose state may cause more serious faults. Therefore, a vibration detection device is used to collect vibration signals from high-voltage parallel reactors to determine the current working condition of the reactor. GBDT is a method of Boosting learning, which uses an additive model and a forward step-by-step algorithm to optimize learning. The basic classification idea is to use multiple weak classifiers to stack in series, increase weights layer by layer, and calculate the residual under the current model (the error value between the predicted value and the actual value). Each round of GBDT training is based on the negative gradient value of the previous round of training model. Finally, each classifier is weighted to give a predicted value to achieve the classification effect. Multiple weak learners constitute a strong learner, and a software and hardware combined online detection device is built to classify vibration signals. The classification results are analyzed to determine the fault state of the reactor. This

[0038] The present invention uses a vibration measuring device to collect vibration signals from a high-voltage shunt reactor, selects the surface of the reactor oil tank as a measuring position, preprocesses the collected signals, and analyzes the preprocessed vibration signals to detect whether a loose fault occurs to the reactor core.

[0039] Example 1

[0040] Reference Figure 1 , a reactor core loose fault diagnosis method based on gradient boosting decision tree algorithm, comprising the following steps:

[0041] 1. Use vibration sensors to collect source signals of high-voltage shunt reactors in working state. After transmitting the collected signals to the terminal, fast Fourier transform (FFT) and wavelet denoising are performed to make the vibration signals achieve smooth filtering and denoising preprocessing. The main purpose is to suppress unnecessary components, extract key data features based on spectrum analysis, and select features with obvious differences for analysis.

[0042] 2. Use the processed data as a training set to train the prediction model constructed by the GBDT algorithm. The model constructed by the GBDT algorithm is a strong learner composed of multiple weak learners serially superimposed and weighted. The steps are as follows:

[0043] (a) Initialize the threshold and weight of a weak learner;

[0044] (b) Establish M classification and regression trees m = 1, 2, ..., M, where M is related to the number of iterations and also determines the accuracy of strong learning;

[0045] (c) Calculate the negative gradient (i.e., residual) of the loss function corresponding to each classification and regression tree, and find the appropriate node for classifying the collected vibration signal. Traverse each possible value of the feature on the regression tree and sort them in descending order according to the residual value to update the strong learner.

[0046] That is, serial superposition is performed until the strong learner f with the highest accuracy is obtained. M (x),

[0047]

[0048] Among them, m>5.

[0049] The constructed strong learner has the characteristics of fast computing speed, high accuracy and good robustness, which facilitates the rapid classification of data.

[0050] 3. Input the vibration signal of the reactor after processing into the model for training, set the classification label, and find that the loose state of fasteners is divided into three cases: 11N.m, 7N.m and 3N.m, so the label is set as normal state, 11N.m, 7N.m and 3N.m. During the training process, set the appropriate number of iterations and learning rate so that the model can continuously learn the features contained in the data. Each iteration is trained on the basis of the residual value of the previous round of training model. After multiple iterations of residual calculation and weighted weighting, the strong learner learns the features contained in the vibration signal of the reactor under different states, the model reaches a suitable error value, and the prediction model training is completed.

[0051] 4. Combine the software and hardware to form a complete vibration detection device. Set the vibration sensor on the surface of the reactor and connect it to the terminal device through a signal transmission line. The terminal device contains a processing program and a trained prediction model to classify the collected vibration signals, determine whether the current high-voltage shunt reactor is in a nut loosening fault, and predict the degree of looseness of the reactor, so as to check and deal with it in time to prevent more serious faults and economic losses in the future.

[0052] Example 2

[0053] Reference Figure 4 A reactor core loose fault diagnosis device comprises a vibration sensor, an acquisition card and a terminal connected in sequence, wherein the vibration sensor is used to measure a vibration source signal of the reactor, the acquisition card is used to acquire the vibration source signal and transmit it to the terminal, and the terminal is used to determine whether the reactor is in a nut loose fault condition according to the vibration source signal, and predict the degree of looseness of the reactor.

[0054] Example 3

[0055] The present invention provides a computer device, such as Figure 5 As shown, it includes an electrically connected memory and a processor, wherein the memory stores a calculation program that can be run on the processor, and when the processor executes the calculation program, the steps of the above-mentioned fault diagnosis method are implemented.

[0056] The computer program may be divided into one or more modules / units, and the one or more modules / units are stored in the memory and executed by the processor to accomplish the present invention.

[0057] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0058] The processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0059] The memory may be used to store the computer programs and / or modules, and the processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory.

[0060] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0061] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0062] Specific Examples

[0063] A vibration measuring device is used to collect source signals of a BKS-76 / 10 three-phase oil-immersed shunt reactor. The vibration signals collected are mainly under normal conditions, 11N.m, 7N.m and 3N.m nut loose conditions, and the collected source signals are processed, and finally fault analysis and prediction are performed. The specific implementation steps are as follows:

[0064] S1: First, place the piezoelectric acceleration vibration sensor on the surface of the three-phase oil-immersed shunt reactor tank, then transmit the collected vibration signal through the signal transmission line, use the acquisition card to collect the signal, and finally transmit it to the terminal system. The piezoelectric acceleration vibration sensor used in this experiment has a more suitable range and sensitivity, and is light in weight, which can reduce the impact of errors and has no additional impact on the entire vibration system.

[0065] The measurement positions selected are 72 measuring points in 9 rows and 8 columns on the surface of the vibration signal reactor. Each measuring point is denoted as w according to row i and column j. ij After selecting the measuring points, the source signal collection work was started. The continuous vibration signals with time series were collected under normal conditions, 11N.m, 7N.m and 3N.m nut loosening conditions. The collected source signal is S(k).

[0066] S(k)=f(k)+ε*e(k)k=0,1,...,n-1

[0067] Where f(k) is the vibration signal, S(k) is the noisy signal, e(k) is the noise, and ε is the standard deviation of the noise coefficient.

[0068] The source signal S(k) is processed by FFT transformation and wavelet packet processing to filter out the noise signal and obtain the vibration signal f(k), thereby achieving the purpose of wavelet noise reduction and smoothing filtering. Finally, spectrum analysis is performed to obtain various characteristics of the signal.

[0069] Under different conditions, the fundamental frequency ratio of the vibration signal measured at the same measuring point is different after comparison. Figure 2a to Figure 2d As shown in Figure 2, under the same condition, the fundamental frequency proportion of the vibration signal measured at different measuring points is also different, such as Figure 3 As shown, we choose to use the fundamental frequency ratio as a feature to classify and predict the processed data.

[0070] S2: Since the amount of collected data is not sufficient, but in order to ensure that the trained model has a good accuracy, the processed data is divided into a test set and a training set at a ratio of 1:4 using the cross-validation method. The training set T is input into the classification model composed of the GBDT algorithm for training, where

[0071] T={(x1,y1),(x2,y2),...(x i ,y i ),...,(x N ,y N )}

[0072]

[0073] Loss function:

[0074] L(y,f(x))=ylog(1+exp(-f(x)))+(1-y)[f(x)+log(1+exp(-f(x)))]

[0075] Among them, χ, y are training samples, (x i ,y i ) is a single sample of logistic regression, R is a real number set, R n is the n-th Cartesian set of the real number set R.

[0076] S3: The main steps of model construction are:

[0077] (a) Initialize the first weak learner and obtain the initial learner:

[0078]

[0079] Note: Estimate the constant value that minimizes the loss function. The initial learner f0(x) is a classification regression tree with only one root node.

[0080] Among them, f0(x) is the initial learner, L(y i,c) is the loss function, c is the constant value that minimizes the loss function;

[0081] (b) Establish M classification and regression trees m = 1, 2, ..., M: For i = 1, 2, ..., N, calculate the negative gradient r of the loss function corresponding to the mth tree m,i , that is, the pseudo residual

[0082] Note: Pseudo residual is the value of the calculated loss function under the current model, which is used to estimate the residual.

[0083] Among them, f(x) is the current learner, L(y i ,f(x i )) is the loss function of the current model, f m-1 (x) is the m-1th learner;

[0084] In order to prevent overfitting and underfitting, the CART regression tree is used to fit the data. m,i Fit a regression tree and get its corresponding leaf node region R m,j , j = 1, 2, ..., J, J is the number of leaf nodes of the regression tree; then for the generated decision tree, calculate the leaf area j of each leaf node and calculate the best residual fitting value c m,j :

[0085]

[0086] Note: The best residual fit value c m,j Used to estimate the value of the corresponding leaf node area under the condition of minimizing the loss function.

[0087] Among them, f m-1 (x i ) is the m-1th learner, and C is the parameter of each leaf node, that is, the mean of the label value.

[0088] (c) Finally, a strong learner is obtained by updating:

[0089]

[0090] Among them, I(x∈R m,j ) is the basis function set;

[0091] S4: Use the strong learner built in the prediction model to learn the data features, and set four labels: normal, 11N.m, 7N.m and 3N.m nut looseness. First, input the training set to train the prediction model. The learner learns the fundamental frequency ratio contained in the data, and then calculates the residual. The residual is used as the label of the sample to train the weak learner f0(x), find the best partition node of the regression tree, and traverse each possible value of each feature. Parameter setting requires setting a suitable learning rate (learning_rate), number of iterations (n_trees), tree depth (max_depth), and setting a suitable threshold and weight so that the model can iterate continuously and generate enough trees to make the strong learner have higher accuracy. In addition, each iteration is trained on the basis of the negative gradient value of the previous round of training model. After repeated residual calculations and weighting, a suitable error value is reached to extract the features in different labels.

[0092] S7: Test the trained model by inputting the test set and classifying the data in the test set into appropriate feature labels. The purpose is to test the accuracy of the prediction model. Finally, error analysis is performed to ensure that the prediction accuracy of the learner reaches the ideal value, indicating that the model training effect meets the requirements.

[0093] S8: The prediction model built using the GBDT algorithm compares the fundamental frequency ratios of different states and different measuring points, and the trained model is directly put into use. In the future, the vibration signals of 72 measuring points collected at any time can be labeled and classified to achieve the prediction effect. For oil-immersed high-voltage shunt reactors, if it can be discovered in time whether the current reactor is in a normal working state or a fault state with different degrees of nut looseness, it can attract the attention of operation and maintenance personnel in time.

[0094] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A method for diagnosing a loose core fault of a reactor, characterized in that: The following steps are involved: S1. Collect the vibration source signal of the high-voltage shunt reactor in the working state and extract the key data features in the vibration source signal; The initial prediction model built using the GBDT algorithm; S2, setting classification labels for the extracted data features as training sets to train the initial prediction model; During the training process, the prediction model continuously learns the features contained in the data. Each iteration is trained based on the residual value of the previous round of training model. After multiple iterations of residual calculation and weighting, the strong learner learns the features contained in the vibration signal of the reactor in different states and obtains the target prediction model. S3, collecting vibration signals of the reactor to be diagnosed, inputting the collected vibration signals into the target prediction model, determining whether the current reactor is in a state of nut loosening fault, and predicting the degree of looseness of the reactor; In S1, the initial prediction model constructed using the GBDT algorithm includes the following steps: S A1, initialize the threshold and weight of the first weak learner; S A2, establish M classification and regression trees; SA3. Calculate the negative gradient of the loss function corresponding to the regression tree, and sort the regression trees in descending order of negative gradient to obtain a strong learner. In SA2, the CART regression tree is used to fit the data, and the negative gradient of the loss function is Fit a regression tree and get its corresponding leaf node area , =1,2,...,J, J is the number of leaf nodes in the regression tree; then calculate the leaf area of ​​each leaf node in the generated decision tree , calculate the best residual fit value , using the best residual fit value Get a strong learner; in, is the model obtained by the m-1th calculation, and c is the constant value that minimizes the loss function.

2. A method for diagnosing a loose core fault of a reactor according to claim 1, characterized in that: In S1, a piezoelectric acceleration vibration sensor is used to collect the vibration source signal of the reactor.

3. A method for diagnosing a loose core fault of a reactor according to claim 1, characterized in that: In S1, the collected vibration source signal is subjected to fast Fourier transform and wavelet denoising to extract key data features.

4. A method for diagnosing a loose core fault of a reactor according to claim 1, characterized in that: In S1, the key data feature is the fundamental frequency of the vibration signal.

5. The method for diagnosing a loose core fault of a reactor according to claim 1, characterized in that: In S2, the extracted data features are divided into a test set and a training set in a ratio of 1:4 using a cross-validation method.

6. A method for diagnosing a loose core fault of a reactor according to claim 1, characterized in that: In S2, the labels include normal state, 11N.m, 7N.m and 3N.m.

7. A reactor core loose fault diagnosis device, characterized in that: The method for implementing the method of claim 1 comprises a vibration sensor, an acquisition card and a terminal connected in sequence, wherein the vibration sensor is used to measure the vibration source signal of the reactor, the acquisition card is used to acquire the vibration source signal and transmit it to the terminal, and the terminal is used to determine whether the reactor is in a nut loosening fault according to the vibration source signal, and predict the degree of looseness of the reactor.

8. A computer device, characterized in that: The method comprises an electrically connected memory and a processor, wherein the memory stores a computing program executable on the processor, and when the processor executes the computing program, the steps of the method described in any one of claims 1 to 6 are implemented.

Citation Information

Patent Citations

  • Electric reactor mechanical fault joint detection model, method and device based on vibration and sound

    CN110929769A

  • Fault diagnosis apparatus, recording medium recording fault diagnosis program and fault diagnosis method

    US20060132958A1

  • Fault diagnosis and preliminary location system and method for transformer core looseness

    US20140118003A1