Tower bolt looseness judgment method, system and equipment and medium

Abstract

The invention relates to the technical field of electric power systems, and discloses a tower bolt looseness judgment method, system, equipment and medium, the method comprises the following steps: acquiring vibration signals of a plurality of bolts of a target tower and a bolt looseness state, and performing modal decomposition on the vibration signals to obtain a plurality of modal components; the method comprises the following steps: constructing a plurality of modal components and a bolt loosening state, constructing a plurality of training data sets by using the plurality of modal components and the bolt loosening state, training the plurality of training data sets by using a BP-BiLSTM neural network, obtaining a plurality of BP-BiLSTM prediction sub-models, carrying out weighted calculation to obtain a BP-BiLSTM prediction model, and predicting an unidentified bolt loosening state by using the BP-BiLSTM prediction model. Therefore, the efficiency and the accuracy of tower bolt looseness judgment are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems, and in particular to a method, system, equipment and medium for determining loose bolts on a tower. Background Art

[0002] Bolt connections are widely used in pole towers because of their many advantages, such as easy disassembly and assembly, and simple installation. Bolts are also the key to connecting and transmitting force between poles. The quality and tightening condition of bolts not only affect the overall reliability of the pole tower, but also threaten the safe operation of the line.

[0003] During the operation of the tower, wind vibration and the dancing of the conductors can cause the bolts of the tower to loosen or fall off, which can seriously reduce the strength of the tower structure and even cause a tower collapse accident. Therefore, it is indispensable to conduct a safety inspection on the bolt condition of the tower. However, the traditional bolt loosening identification relies on manual operation, which easily leads to low efficiency and poor accuracy in bolt loosening identification. Summary of the invention

[0004] In view of this, the present invention provides a method, system, device and medium for determining loose bolts in a tower, which solves the technical problem that traditional bolt loosening determination relies on manual operation, which easily leads to low efficiency and poor accuracy in bolt loosening determination.

[0005] A first aspect of the present invention provides a method for determining whether a tower bolt is loose, comprising:

[0006] Collect vibration signals and bolt looseness status of multiple bolts of the target tower;

[0007] Using a variational modal decomposition algorithm to perform modal decomposition on the vibration signal to obtain multiple modal components;

[0008] Constructing a plurality of training data sets according to the plurality of modal components and the bolt loosening states corresponding to the modal components;

[0009] Using a BP-BiLSTM neural network to train the plurality of training data sets respectively, to obtain a plurality of BP-BiLSTM prediction sub-models;

[0010] Using a plurality of the BP-BiLSTM prediction sub-models to perform weighted calculation to obtain a BP-BiLSTM prediction model;

[0011] The BP-BiLSTM prediction model is used to predict the unidentified bolt loosening state.

[0012] Preferably, before the step of using a variational modal decomposition algorithm to perform modal decomposition on the vibration signal to obtain a plurality of modal components, the method further includes:

[0013] The vibration signal is normalized.

[0014] Preferably, the method further comprises:

[0015] The network parameters of the variational mode decomposition algorithm are optimized by using a frost particle algorithm. The network parameters of the variational mode decomposition algorithm are optimized by using a frost particle algorithm, including:

[0016] Initialize the parameters of the frost particle algorithm, where the position of the frost particle population represents the network parameters of the variational mode decomposition algorithm;

[0017] The current position of the frost ice particle is updated according to the movement property of the frost ice particle, and the current position of the frost ice particle is:

[0018]

[0019] In the formula, is the current position of the frost particle after updating. For the frost population The best crystal The position of a particle, for A random number, is the movement angle of frost particles, For environmental factors, is the particle adhesion force, for A random number, , They represent the upper and lower bounds of the particle motion space respectively;

[0020] in,

[0021]

[0022] Where h is the current iteration number, is the maximum number of iterations;

[0023] The particle adhesion coefficient is calculated based on the ratio of the current number of iterations to the maximum number of iterations;

[0024] Randomly generate a Random numbers within , when the random number When the random number is less than the adhesion coefficient, the process proceeds to the step of updating the current position of the frost particles according to the motion properties of the frost particles until the random number Until it is not less than the adhesion coefficient, and proceed to the next step;

[0025] Determining a minimum envelope entropy as a fitness value according to the updated current position of the frost ice particle, and normalizing the fitness value;

[0026] Randomly generate a Random numbers within , when the random number When it is less than the normalized fitness value, the process proceeds to the step of updating the current position of the frost ice particle according to the motion properties of the frost ice particle;

[0027] By comparing the fitness value of the current position of the frost ice particle after the update with the fitness value of the current position, the optimal frost ice particle position and the optimal fitness value are updated;

[0028] When the optimal fitness value reaches the convergence condition and when the number of iterations is less than the maximum number of iterations, the current position of the frost ice particle is output, and the optimal network parameters of the variational mode decomposition algorithm are determined according to the current position of the frost ice particle.

[0029] Preferably, the vibration signal is modally decomposed by using a variational modal decomposition algorithm to obtain multiple modal components, including:

[0030] The variational problem is constructed by minimizing the sum of bandwidth estimates of each modal component of the vibration signal;

[0031] Introducing a quadratic penalty term and a Lagrangian operator into the variational problem, and converting the variational problem into an unconstrained problem to obtain an augmented Lagrangian function;

[0032] The augmented Lagrangian function is transformed from the time domain to the frequency domain, and the variational problem is solved using an alternating multiplier algorithm to obtain optimized modal components and center frequencies;

[0033] Iterative calculation is performed according to the optimized modal components and center frequencies until convergence accuracy is reached, and the vibration signal is decomposed into multiple modal components.

[0034] Preferably, the BP-BiLSTM neural network is used to train the plurality of training data sets respectively to obtain a plurality of BP-BiLSTM prediction sub-models, including:

[0035] For each of the training data sets, forward propagation is performed on the training data set using a BP neural network to obtain a first BP predicted output value;

[0036] Comparing the first BP predicted output value with the actual value to determine the total data error;

[0037] Determining whether the total error of the data is greater than a predetermined error threshold;

[0038] When it is determined that the total data error is greater than the predetermined error threshold, the total data error is back-propagated to update the weights and biases of the BP neural network layer by layer;

[0039] Using the updated BP neural network to forward propagate the training data set to each of the training data sets, and obtaining a first BP predicted output value until the total error of the data is no greater than the predetermined error threshold, and outputting a second BP predicted output value as input data of the BiLSTM neural network;

[0040] The input data of the BiLSTM neural network is input into the BiLSTM neural network for forward propagation and backward propagation until the iteration converges, and the iteration is stopped to obtain a trained BP-BiLSTM prediction sub-model.

[0041] In a second aspect, the present invention further provides a tower bolt loosening determination system, comprising:

[0042] A signal acquisition module, used to collect vibration signals of multiple bolts of the target tower and the looseness status of the bolts;

[0043] A modal decomposition module, used to perform modal decomposition on the vibration signal using a variational modal decomposition algorithm to obtain multiple modal components;

[0044] A data set construction module, used to construct a plurality of training data sets according to a plurality of the modal components and the bolt loosening states corresponding to the modal components;

[0045] A model training module, used to train the plurality of training data sets respectively using a BP-BiLSTM neural network to obtain a plurality of BP-BiLSTM prediction sub-models;

[0046] A model weighting module, used to perform weighted calculation using a plurality of the BP-BiLSTM prediction sub-models to obtain a BP-BiLSTM prediction model;

[0047] The bolt loosening prediction module is used to predict the unidentified bolt loosening state by using the BP-BiLSTM prediction model.

[0048] In a third aspect, the present invention further provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for determining looseness of tower bolts as described in the first aspect.

[0049] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed, the steps of the method for determining looseness of tower bolts as described in the first aspect are implemented.

[0050] In a fifth aspect, the present invention further provides a computer program product, comprising a computer program stored on a non-transitory computer-readable storage medium, wherein the computer program comprises program instructions, wherein when the program instructions are executed by a computer, the computer executes the steps of the method for determining looseness of tower bolts as described in the first aspect.

[0051] It can be seen from the above technical scheme that the present invention collects vibration signals and bolt loosening states of multiple bolts of a target tower, and performs modal decomposition on the vibration signals to obtain multiple modal components. Multiple modal components and bolt loosening states are used to construct multiple training data sets, and the BP-BiLSTM neural network is used to train the multiple training data sets respectively to obtain multiple BP-BiLSTM prediction sub-models for weighted calculation to obtain a BP-BiLSTM prediction model. The BP-BiLSTM prediction model is used to predict the unidentified bolt loosening state, thereby improving the efficiency and accuracy of tower bolt loosening judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0053] Figure 1 A flow chart of a method for determining looseness of a tower bolt provided by an embodiment of the present invention;

[0054] Figure 2 This is the BiLSTM neural network structure diagram;

[0055] Figure 3 A schematic diagram of the structure of a tower bolt loosening identification system provided by an embodiment of the present invention;

[0056] Figure 4 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0057] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] An embodiment of the present application provides a method for determining loose tower bolts, which can be applicable to the situation of determining loose tower bolts. The method can be executed by a tower bolt looseness determination device, which can be implemented in the form of hardware and / or software, and the tower bolt looseness determination device can be configured in a computer device.

[0059] like Figure 1 As shown, the embodiment of the present application provides a method for determining whether a tower bolt is loose, comprising:

[0060] Step S1: collecting vibration signals and bolt looseness status of multiple bolts of a target tower.

[0061] Among them, sensors can be installed at multiple bolts on the tower to obtain vibration signals at the multiple bolts, and the looseness status of the multiple bolts can be evaluated to determine the bolt looseness status of the multiple bolts, and the bolt looseness status includes loose and not loose.

[0062] Step S2: using a variational modal decomposition algorithm to perform modal decomposition on the vibration signal to obtain multiple modal components.

[0063] Before step S2, in order to make the training sample data more refined, the vibration signal may be normalized.

[0064] The embodiment of the present application also uses the frost particle algorithm to optimize the network parameters of the variational mode decomposition algorithm, thereby searching for the number of decomposition layers of the variational mode decomposition algorithm through an efficient parameter update mechanism. and penalty factor , determine the optimal parameter combination .

[0065] Specifically, the frost particle algorithm is used to optimize the network parameters of the variational mode decomposition algorithm, including:

[0066] Step S21, initializing the parameters of the frost particle algorithm, wherein the position of the frost particle population represents the network parameters of the variational mode decomposition algorithm.

[0067] The parameters of the frost particle algorithm include the location of the frost population, the maximum number of iterations, and the upper bound of the optimization problem. and lower bound .

[0068] Among them, for the frost ice particle population The location is:

[0069]

[0070] In the formula, is the population of frost particles, Indicates the sequence number of the frost crystal. Indicates the ordinal number of the frost ice particle.

[0071] In addition, all frost particles have a fitness value As shown below:

[0072]

[0073] In the formula, The fitness of each frost particle.

[0074] Step S22: Update the current position of the frost ice particle according to the motion properties of the frost ice particle. The current position of the frost ice particle is:

[0075]

[0076] In the formula, is the current position of the frost particle after updating. For the frost population The best crystal The position of a particle, for A random number, Control the direction of particle movement and follow the direction of particle movement Iterations change, is the movement angle of frost particles, For environmental factors, is the particle adhesion force, for A random number, Used to control the distance between particles. , They represent the upper and lower bounds of the particle motion space respectively;

[0077] Among them, in order to ensure convergence settings:

[0078]

[0079] Where h is the current iteration number, is the maximum number of iterations.

[0080]

[0081] In the formula, Used to control the number of steps in the step function.

[0082] Among them, the motion property of frost ice particles is that each frost ice particle Before condensation, they will move according to a certain rule. Particles in a free state will move to the vicinity of the soft ice and condense, but they will not be able to condense if they exceed a certain range. During the formation of soft ice, the randomness of each particle will cause the condensation area to increase, and it will eventually reach a stable state under the influence of the environment.

[0083] Step S23, calculating the particle adhesion coefficient according to the ratio of the current iteration number to the maximum iteration number.

[0084] in, It will increase with the number of iterations, affecting the probability of particle condensation; for A random number, which is Controls whether particles condense, that is, whether particle positions are updated.

[0085] Among them, the adhesion coefficient E is calculated as:

[0086]

[0087] Step S24: randomly generate a Random numbers within , when the random number If it is less than the adhesion coefficient, it will go to the step of updating the current position of the frost particles according to the motion properties of the frost particles until the random number Until it is not less than the adhesion coefficient, and proceed to the next step;

[0088] Step S25, determining the minimum envelope entropy as the fitness value according to the updated current position of the frost ice particle, and normalizing the fitness value;

[0089] Step S26: Randomly generate a Random numbers within , when the random number When it is less than the normalized fitness value, the process proceeds to the step of updating the current position of the frost ice particle according to the motion properties of the frost ice particle;

[0090] Among them, as the condensation area of ​​soft ice increases, strong randomness and wide coverage are generated, allowing the algorithm to quickly cover the entire space search. Hard ice will condense in the same direction due to external factors. Since the growth direction of hard ice is consistent, it is easy for hard ice to cross, realizing the dimensional exchange of ordinary particles and optimal particles, which is beneficial to improve the accuracy of the solution. Due to the cross-exchange between particles during the iteration process, local optimality is avoided and convergence is improved. As the hard ice grows and punctures, the cross-replacement between particles is shown in the following formula:

[0091]

[0092] Where: is the normalized value of the current fitness value, indicating the The probability of a particle being selected; for A random number.

[0093] Step S27, updating the optimal frost ice particle position and the optimal fitness value by comparing the fitness value of the current position of the updated frost ice particle with the fitness value of the current position;

[0094] Step S28: When the optimal fitness value reaches the convergence condition and when the number of iterations is less than the maximum number of iterations, the current position of the frost ice particle is output, and the optimal network parameters of the variational mode decomposition algorithm are determined according to the current position of the frost ice particle.

[0095] After obtaining the optimal network parameters of the variational modal decomposition algorithm, the optimal network parameters are updated to the variational modal decomposition algorithm, and the variational modal decomposition algorithm is used to perform modal decomposition on the vibration signal to obtain multiple modal components.

[0096] Specifically, in step S2, the vibration signal is modally decomposed using a variational modal decomposition algorithm to obtain multiple modal components, including:

[0097] Step S201: construct a variational problem by minimizing the sum of bandwidth estimation values ​​of each modal component of the vibration signal.

[0098] Among them, VMD has good applicability for processing nonlinear signals. It can decompose the signal into a specified number of intrinsic mode functions (IMFs), each of which has a certain center frequency. and finite bandwidth. The optimization goal is to minimize the sum of the estimated values ​​of each IMF bandwidth. The constructed variational problem formula is as follows:

[0099]

[0100] In the formula, the original signal is decomposed by VMD into have their own center frequencies of The modal components ; Find the time partial derivative of a logarithm; is the Dirac function; is a mixed operator; is the original tower bolt vibration signal.

[0101] Step S202: introducing a quadratic penalty term and a Lagrangian operator into the variational problem, converting the variational problem into an unconstrained problem, and obtaining an augmented Lagrangian function.

[0102] In order to find the optimal solution to the constrained variational problem, the quadratic penalty term and the Lagrangian operator are introduced. The extended Lagrangian expression is:

[0103]

[0104]

[0105] In the formula, is the Lagrangian function, and the following three terms represent the gradient function norm, quadratic penalty term, and Lagrange multiplier term, is the penalty factor, which is the penalty factor searched by the RIME algorithm in step 2.5 , is the Lagrange multiplier. The first term is mainly used to evaluate the bandwidth, and the other two terms are used to ensure the reconstruction constraints.

[0106] Step S203: transform the augmented Lagrangian function from the time domain to the frequency domain, and solve the variational problem using the alternating multiplier algorithm to obtain optimized modal components and center frequencies.

[0107] Using the alternating multiplier algorithm to continuously update , and Find the optimal solution. Each IMF component and its center frequency The iteration formula is:

[0108]

[0109]

[0110] In the formula, is the Fourier transform of the original signal, is the Fourier transform of the Lagrange multiplier, is the frequency value, is the center frequency of each order IMF.

[0111] Step S204: performing iterative calculations according to the optimized modal components and center frequencies until convergence accuracy is reached, and decomposing the vibration signal into multiple modal components.

[0112] The convergence condition is:

[0113]

[0114] In the formula, It is the preset discrimination accuracy.

[0115] Step S3: construct multiple training data sets according to multiple modal components and the bolt loosening states corresponding to the modal components.

[0116] Among them, each modal component corresponds to a loose state of the bolt, so a training data set can be constructed using each modal component and the loose state of the bolt corresponding to the modal component, so that multiple training data sets can be obtained.

[0117] Step S4: Use the BP-BiLSTM neural network to train multiple training data sets respectively to obtain multiple BP-BiLSTM prediction sub-models.

[0118] Among them, BP neural network is good at processing nonlinear mapping relationships, while BiLSTM is particularly suitable for processing long-term dependency problems of time series data because it can consider information from past and future time points at the same time.

[0119] The BP neural network model consists of an input layer, a hidden layer and an output layer.

[0120] The hidden layer The input and output of a neuron are:

[0121]

[0122] In the formula, is the number of neurons in the input layer; The first The first input to the hidden layer The weight of each neuron; The first Input values; For the hidden layer The threshold of a neuron.

[0123]

[0124] In the formula, is the hidden layer activation function.

[0125] The output layer The input of a neuron is:

[0126]

[0127] In the formula, is the number of neurons in the hidden layer; is the hidden layer The first input to the hidden layer The weight of each neuron; The output layer The threshold of a neuron.

[0128] The output layer The output of a neuron for:

[0129]

[0130] In the formula, is the output layer activation function.

[0131] If the system has samples, and compare the result obtained by the output layer with the actual expected result. The quadratic error criterion function of samples for:

[0132]

[0133] In the formula, is the number of output layers; For the The sample in The target value on the output layer; For the The sample in The actual value at the output layer.

[0134] The total error of the system for the sample is:

[0135]

[0136] If the above error calculation result does not meet the conditions, the error will be back-propagated and the weights and biases will be updated layer by layer.

[0137] Output layer weight correction:

[0138]

[0139] Where: is the step length.

[0140] Output layer threshold correction:

[0141]

[0142] Hidden layer weight correction:

[0143]

[0144] Hidden layer threshold correction:

[0145]

[0146] The above formula can be simplified as:

[0147]

[0148]

[0149]

[0150]

[0151] At this point, the BP neural network has completed one forward propagation and one backward propagation process, that is, one iteration.

[0152] The BiLSTM neural network structure is as follows Figure 2 As shown in the figure, BiLSTM consists of two layers of LSTM. One layer of forward propagation LSTM1 mainly obtains historical data features, and one layer of back-propagation LSTM2 mainly obtains future data features, further exploring the time series characteristics of the data, reducing the risk of underfitting, and improving prediction accuracy.

[0153] The network parameter training expression of forward propagation LSTM1 is as follows:

[0154]

[0155] The network parameter training expression of back-propagation LSTM2 is as follows:

[0156]

[0157] The output of LSTM1 constructs the fully connected layer LSTM1, and the output of LSTM2 constructs the fully connected layer LSTM2. Training results of the fully connected layer The expression is as follows:

[0158]

[0159] In the formula, express Moment prediction output; is the weight parameter of the fully connected layer; is the bias parameter of the full connection.

[0160] To this end, in step S4, the BP-BiLSTM neural network is used to train multiple training data sets respectively to obtain multiple BP-BiLSTM prediction sub-models, including:

[0161] Step S401: for each training data set, forward propagation is performed on the training data set using a BP neural network to obtain a first BP prediction output value;

[0162] Step S402: Compare the first BP predicted output value with the actual value to determine the total data error;

[0163] Step S403, determining whether the total data error is greater than a predetermined error threshold;

[0164] Step S404: when it is determined that the total data error is greater than a predetermined error threshold, the total data error is back-propagated to update the weights and biases of the BP neural network layer by layer;

[0165] Step S405: using the updated BP neural network to forward propagate the training data set using the BP neural network to obtain a first BP predicted output value for each training data set, until the total data error is no greater than a predetermined error threshold, and outputting a second BP predicted output value as input data for the BiLSTM neural network;

[0166] Step S406: Input the input data of the BiLSTM neural network into the BiLSTM neural network for forward propagation and backward propagation until the iteration converges, stop the iteration, and obtain the trained BP-BiLSTM prediction sub-model.

[0167] Step S5: Use multiple BP-BiLSTM prediction sub-models to perform weighted calculation to obtain a BP-BiLSTM prediction model.

[0168] Among them, by setting the weights of each BP-BiLSTM prediction sub-model, the sum of the weights of each BP-BiLSTM prediction sub-model is 1. In general, the weights of each BP-BiLSTM prediction sub-model can be equal. By performing weighted calculation on multiple BP-BiLSTM prediction sub-models, a BP-BiLSTM prediction model is obtained, so that the most matching prediction model can be flexibly selected according to the unique characteristics of each decomposed component, thereby obtaining better prediction accuracy than the traditional single prediction model.

[0169] Step S6: Use the BP-BiLSTM prediction model to predict the unidentified bolt looseness state.

[0170] It should be noted that the embodiment of the present application collects vibration signals and bolt looseness status of multiple bolts of the target tower, and performs modal decomposition on the vibration signals to obtain multiple modal components. Multiple modal components and bolt looseness status are used to construct multiple training data sets, and the BP-BiLSTM neural network is used to train the multiple training data sets respectively to obtain multiple BP-BiLSTM prediction sub-models for weighted calculation to obtain a BP-BiLSTM prediction model. The BP-BiLSTM prediction model is used to predict the unidentified bolt looseness status, thereby improving the efficiency and accuracy of tower bolt looseness identification.

[0171] Based on the same inventive concept, an embodiment of the present application further provides a tower bolt loosening identification system for implementing the tower bolt loosening identification method mentioned above.

[0172] The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the tower bolt loosening identification system provided below can refer to the limitations on the tower bolt loosening identification method above, and will not be repeated here.

[0173] like Figure 3 As shown, the embodiment of the present application also provides a tower bolt loosening determination system, comprising:

[0174] The signal acquisition module 100 is used to collect vibration signals and bolt looseness status of multiple bolts of the target tower;

[0175] The modal decomposition module 200 is used to perform modal decomposition on the vibration signal using a variational modal decomposition algorithm to obtain multiple modal components;

[0176] A data set construction module 300 is used to construct multiple training data sets according to multiple modal components and bolt loosening states corresponding to the modal components;

[0177] The model training module 400 is used to train multiple training data sets using a BP-BiLSTM neural network to obtain multiple BP-BiLSTM prediction sub-models;

[0178] A model weighting module 500 is used to perform weighted calculation using multiple BP-BiLSTM prediction sub-models to obtain a BP-BiLSTM prediction model;

[0179] The bolt loosening prediction module 600 is used to predict the unidentified bolt loosening state by using the BP-BiLSTM prediction model.

[0180] In some embodiments, the system further includes: a normalization processing module, configured to perform normalization processing on the vibration signal.

[0181] In some embodiments, the system further includes: a parameter optimization module for optimizing network parameters of the variational mode decomposition algorithm using a frost particle algorithm; optimizing network parameters of the variational mode decomposition algorithm using a frost particle algorithm includes:

[0182] Initialize the parameters of the frost particle algorithm, where the position of the frost particle population represents the network parameters of the variational mode decomposition algorithm;

[0183] Update the current position of the frost particles according to the movement properties of the frost particles. The current position of the frost particles is:

[0184]

[0185] In the formula, is the current position of the frost particle after updating. For the frost population The best crystal The position of a particle, for A random number, is the movement angle of frost particles, For environmental factors, is the particle adhesion force, for A random number, , They represent the upper and lower bounds of the particle motion space respectively;

[0186] in,

[0187]

[0188] Where h is the current iteration number, is the maximum number of iterations;

[0189] The particle adhesion coefficient is calculated based on the ratio of the current number of iterations to the maximum number of iterations;

[0190] Randomly generate a Random numbers within , when the random number If it is less than the adhesion coefficient, it will go to the step of updating the current position of the frost particles according to the motion properties of the frost particles until the random number Until it is not less than the adhesion coefficient, and proceed to the next step;

[0191] Determine the minimum envelope entropy as the fitness value according to the updated current position of the frost ice particle, and normalize the fitness value;

[0192] Randomly generate a Random numbers within , when the random number When it is less than the normalized fitness value, the process proceeds to the step of updating the current position of the frost ice particle according to the motion properties of the frost ice particle;

[0193] By comparing the fitness value of the current position of the frost ice particle after the update with the fitness value of the current position, the optimal frost ice particle position and the optimal fitness value are updated;

[0194] When the optimal fitness value reaches the convergence condition and when the number of iterations is less than the maximum number of iterations, the current position of the frost ice particle is output, and the optimal network parameters of the variational mode decomposition algorithm are determined according to the current position of the frost ice particle.

[0195] In some embodiments, a variational modal decomposition algorithm is used to perform modal decomposition on the vibration signal to obtain multiple modal components, including:

[0196] The variational problem is constructed by minimizing the sum of bandwidth estimates of each modal component of the vibration signal;

[0197] By introducing quadratic penalty terms and Lagrangian operators into the variational problem, we can transform the variational problem into an unconstrained problem and obtain the augmented Lagrangian function.

[0198] The augmented Lagrangian function is transformed from the time domain to the frequency domain, and the variational problem is solved using the alternating multiplier algorithm to obtain the optimized modal components and center frequencies.

[0199] Iterative calculation is performed according to the optimized modal components and center frequencies until convergence accuracy is reached, and the vibration signal is decomposed into multiple modal components.

[0200] In some embodiments, a plurality of training data sets are trained respectively using a BP-BiLSTM neural network to obtain a plurality of BP-BiLSTM prediction sub-models, including:

[0201] For each training data set, the BP neural network is used to perform forward propagation on the training data set to obtain the first BP predicted output value;

[0202] Compare the first BP predicted output value with the actual value to determine the total data error;

[0203] Determine whether the total error of the data is greater than a predetermined error threshold;

[0204] When the total error of the data is judged to be greater than the predetermined error threshold, the total error of the data is back-propagated, and the weights and biases of the BP neural network are updated layer by layer;

[0205] Using the updated BP neural network to forward propagate the training data set to each training data set, and obtaining the first BP prediction output value until the total data error is no greater than a predetermined error threshold, and outputting the second BP prediction output value as the input data of the BiLSTM neural network;

[0206] The input data of the BiLSTM neural network is input into the BiLSTM neural network for forward propagation and back propagation until the iteration converges, and the iteration is stopped to obtain the trained BP-BiLSTM prediction sub-model.

[0207] like Figure 4 As shown, an embodiment of the present application further provides an electronic device, the electronic device 10 includes a memory 20 and a processor 30, the memory 20 stores a computer program, and when the computer program is executed by the processor 30, the processor 30 executes the steps of the method for determining loose tower bolts in the above embodiment.

[0208] The embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed, the steps of the method for determining looseness of tower bolts in the above embodiment are implemented.

[0209] An embodiment of the present application also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer executes the steps of the method for determining looseness of tower bolts in the above-mentioned embodiment.

[0210] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, electronic devices, computer storage media and computer program products can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0211] It should be noted that the terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or apparatus.

[0212] In several embodiments provided by the present invention, it is understood that each box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved.

[0213] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, electronic devices, computer storage media, computer program products and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0214] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0215] 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. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0216] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or 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 executing all or part of the steps of the method described in each embodiment of the present invention through a computer device (which can be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (full name in English: Read-Only Memory, English abbreviation: ROM), random access memory (full name in English: Random Access Memory, English abbreviation: RAM), disk or optical disk and other media that can store program codes.

[0217] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for determining the looseness of a tower bolt, characterized in that: include: Collect vibration signals and bolt looseness status of multiple bolts of the target tower; Using a variational modal decomposition algorithm to perform modal decomposition on the vibration signal to obtain multiple modal components; Constructing a plurality of training data sets according to the plurality of modal components and the bolt loosening states corresponding to the modal components; Using a BP-BiLSTM neural network to train the plurality of training data sets respectively, to obtain a plurality of BP-BiLSTM prediction sub-models; Using a plurality of the BP-BiLSTM prediction sub-models to perform weighted calculation to obtain a BP-BiLSTM prediction model; The BP-BiLSTM prediction model is used to predict the unidentified bolt loosening state.

2. The method for determining the looseness of a tower bolt according to claim 1, characterized in that: Before the step of using a variational modal decomposition algorithm to perform modal decomposition on the vibration signal to obtain a plurality of modal components, the method further includes: The vibration signal is normalized.

3. The method for determining the looseness of a tower bolt according to claim 1, characterized in that: Also includes: The network parameters of the variational mode decomposition algorithm are optimized using a frost particle algorithm; The method of using the frost particle algorithm to optimize the network parameters of the variational mode decomposition algorithm includes: Initialize the parameters of the frost particle algorithm, where the position of the frost particle population represents the network parameters of the variational mode decomposition algorithm; The current position of the frost ice particle is updated according to the movement property of the frost ice particle, and the current position of the frost ice particle is: In the formula, The current position of the frost particle after updating. For the frost population The best crystal The position of a particle, for A random number, is the movement angle of frost particles, For environmental factors, is the particle adhesion, for A random number, , They represent the upper and lower bounds of the particle motion space respectively; in, Where h is the current iteration number, is the maximum number of iterations; The particle adhesion coefficient is calculated based on the ratio of the current number of iterations to the maximum number of iterations; Randomly generate a Random numbers within , when the random number When the random number is less than the adhesion coefficient, the process proceeds to the step of updating the current position of the frost particles according to the motion properties of the frost particles until the random number Until it is not less than the adhesion coefficient, and proceed to the next step; Determining a minimum envelope entropy as a fitness value according to the updated current position of the frost ice particle, and normalizing the fitness value; Randomly generate a Random numbers within , when the random number When it is less than the normalized fitness value, the process proceeds to the step of updating the current position of the frost ice particle according to the motion properties of the frost ice particle; By comparing the fitness value of the current position of the frost ice particle after the update with the fitness value of the current position, the optimal frost ice particle position and the optimal fitness value are updated; When the optimal fitness value reaches the convergence condition and when the number of iterations is less than the maximum number of iterations, the current position of the frost ice particle is output, and the optimal network parameters of the variational mode decomposition algorithm are determined according to the current position of the frost ice particle.

4. The method for determining the looseness of a tower bolt according to claim 1, characterized in that: The vibration signal is modally decomposed by using a variational modal decomposition algorithm to obtain multiple modal components, including: The variational problem is constructed by minimizing the sum of bandwidth estimates of each modal component of the vibration signal; Introducing a quadratic penalty term and a Lagrangian operator into the variational problem, and converting the variational problem into an unconstrained problem to obtain an augmented Lagrangian function; The augmented Lagrangian function is transformed from the time domain to the frequency domain, and the variational problem is solved using an alternating multiplier algorithm to obtain optimized modal components and center frequencies; Iterative calculation is performed according to the optimized modal components and center frequencies until convergence accuracy is reached, and the vibration signal is decomposed into multiple modal components.

5. The method for determining the looseness of a tower bolt according to claim 1, characterized in that: The BP-BiLSTM neural network is used to train the plurality of training data sets respectively to obtain a plurality of BP-BiLSTM prediction sub-models, including: For each of the training data sets, forward propagation is performed on the training data set using a BP neural network to obtain a first BP predicted output value; Comparing the first BP predicted output value with the actual value to determine the total data error; Determining whether the total error of the data is greater than a predetermined error threshold; When it is determined that the total data error is greater than the predetermined error threshold, the total data error is back-propagated to update the weights and biases of the BP neural network layer by layer; Using the updated BP neural network to forward propagate the training data set to each of the training data sets, and obtaining a first BP predicted output value until the total error of the data is no greater than the predetermined error threshold, and outputting a second BP predicted output value as input data of the BiLSTM neural network; The input data of the BiLSTM neural network is input into the BiLSTM neural network for forward propagation and backward propagation until the iteration converges, and the iteration is stopped to obtain a trained BP-BiLSTM prediction sub-model.

6. A tower bolt loosening identification system, characterized in that: include: A signal acquisition module, used to collect vibration signals of multiple bolts of the target tower and the looseness status of the bolts; A modal decomposition module, used to perform modal decomposition on the vibration signal using a variational modal decomposition algorithm to obtain multiple modal components; A data set construction module, used to construct a plurality of training data sets according to a plurality of the modal components and the bolt loosening states corresponding to the modal components; A model training module, used to train the plurality of training data sets respectively using a BP-BiLSTM neural network to obtain a plurality of BP-BiLSTM prediction sub-models; A model weighting module, used to perform weighted calculation using a plurality of the BP-BiLSTM prediction sub-models to obtain a BP-BiLSTM prediction model; The bolt loosening prediction module is used to predict the unidentified bolt loosening state by using the BP-BiLSTM prediction model.

7. An electronic device, characterized in that: The electronic device includes a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor executes the steps of the method for determining looseness of a tower bolt as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the steps of the method for determining the looseness of a tower bolt as described in any one of claims 1 to 5 are implemented.

9. A computer program product, characterized in that The computer program product includes a computer program stored on a non-transitory computer-readable storage medium, wherein the computer program includes program instructions, wherein when the program instructions are executed by a computer, the computer is caused to execute the steps of the method for determining looseness of a tower bolt as described in any one of claims 1 to 5.

Images