Abrasion prediction method and system for suspension clamp

Abstract

The embodiment of the invention discloses a wear prediction method and system for a suspension clamp, and relates to the technical field of electric power fittings. The precision of the wear prediction result of the suspension clamp can be effectively improved. Comprising the following steps: acquiring working state data and real-time meteorological data, which are acquired in real time, of the suspension clamp; acquiring working state data and real-time meteorological data, which are acquired in real time, of the target suspension clamp; performing variational mode decomposition preprocessing on the working state data to obtain a mode function; and inputting the modal function and the real-time meteorological data into a pre-trained wear prediction model based on a Transform model so as to predict the wear degree of the target suspension clamp. The method is suitable for suspension clamp wear prediction scenes.

Description

Technical Field

[0001] This invention relates to the field of power fittings technology, and in particular to a method and system for predicting the wear of suspension clamps. Background Technology

[0002] Suspension clamps are critical components in power transmission lines used to secure conductors. They are subjected to wind vibration, mechanical loads, and environmental corrosion over long periods, making them highly susceptible to wear. This can lead to conductor detachment or breakage, causing major power accidents. Therefore, it is necessary to monitor the wear degree of each suspension clamp in the transmission line to ensure timely maintenance and replacement of any clamps posing a safety hazard, thereby guaranteeing the safe and stable operation of the transmission line.

[0003] Currently, wear monitoring of suspension clamps mainly relies on finite element analysis, which uses mathematical approximations to simulate and analyze the clamps, thereby predicting their wear condition. However, this requires a large amount of computation, and to facilitate calculation, the finite element model is often simplified, which may reduce the accuracy of the analysis results. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a method and system for predicting the wear of suspension clamps, which can effectively improve the accuracy of the wear prediction results of suspension clamps.

[0005] In a first aspect, embodiments of the present invention provide a method for predicting the wear of a suspension clamp, comprising the steps of: acquiring real-time collected working state data of a target suspension clamp and real-time meteorological data; performing variational mode decomposition preprocessing on the working state data to obtain a mode function; inputting the mode function and the real-time meteorological data into a pre-trained wear prediction model based on a Transformer model to predict the wear degree of the target suspension clamp; wherein, the training method of the wear prediction model based on the Transformer model includes: establishing a finite element model of the target suspension clamp; based on the established finite element model, acquiring working state data of the target suspension clamp under different working load conditions, the working state data including: contact pressure and contact slippage; acquiring the wear degree of the target suspension clamp corresponding to the working state data; using the working state data of the target suspension clamp under different working load conditions, the wear degree of the target suspension clamp, and historical meteorological data as a training dataset, and training an initial wear prediction model based on the Transformer model according to the training dataset to obtain the wear prediction model based on the Transformer model.

[0006] Optionally, establishing the finite element model of the target suspension clamp includes: constructing a three-dimensional geometric model of the target suspension clamp based on its material, geometric dimensions, and shape; importing the three-dimensional geometric model into a finite element analysis tool, dividing it into meshes, refining the mesh locally in the contact area, and setting material properties and boundary conditions to obtain the finite element model of the target suspension clamp.

[0007] Optionally, obtaining the working state data of the target suspension clamp under different working loads based on the established finite element model includes: inputting different working loads into the finite element model of the target suspension clamp for solution calculation to obtain the contact pressure and contact slip of the target suspension clamp under different working loads.

[0008] Optionally, obtaining the wear degree of the target suspension clamp corresponding to the working status data includes: calculating the wear degree of the target suspension clamp corresponding to the working status data based on the Archard model.

[0009] Optionally, the Archard model is represented as: In the formula, h represents the degree of wear; k represents the wear coefficient, which is obtained by fitting historical data of similar parts; p represents the contact pressure; s represents the contact slip; H represents the material hardness of the target suspension clamp; and D represents the thickness of the target suspension clamp.

[0010] Optionally, the step of performing variational mode decomposition preprocessing on the working state data to obtain mode functions includes: adaptively decomposing the working state data using variational mode decomposition, and decomposing the working state data into k mode functions by solving a constrained variational problem. This allows for the extraction of stable features;

[0011] The constrained variational problem is expressed as follows:

[0012] minutes { u k } ， { oh k } ∑ k=1 k || ∂ t [( d (t)+ j π t ) * u k (t)] e - j oh k t || 2 2 ;

[0013] ;

[0014] in, Let be the center frequency of the k-th modal function. Let k be the k-th mode function obtained from the decomposition, where k is the number of modes. Let j be the Dirac delta function, and j denote the complex unit, satisfying j 2 =-1, This represents the partial derivative with respect to time t. This represents the convolution calculation, where x(t) is the original non-stationary time series to be decomposed, and t represents the time variable.

[0015] Optionally, the working state data is decomposed into k modal functions by solving a constrained variational problem. This includes: introducing a quadratic penalty factor and Lagrange multipliers to transform the constrained variational problem into an unconstrained variational problem, expressed as:

[0016] ;

[0017] Where a is the quadratic penalty factor, λ is the Lagrange multiplier, λ(t) is a time-dependent Lagrange multiplier function, and x(t) is the original non-stationary time series to be decomposed. To reconstruct the time series;

[0018] Update the variable { using the alternating direction multiplier algorithm. }、{ } and λ, solving an unconstrained variational problem, are expressed as:

[0019] u ̂ k n+1 ( oh )= x ̂ ( oh )- ∑ i ≠ k u ̂ i ( oh )+ l ̂ ( oh ) 2 1+2 α [ oh - oh k ] 2 ;

[0020] ;

[0021] l ̂ n+1 ( oh )= l ̂ n ( oh )+ t [x( oh )- ∑ k=1 k u k n+1 ( oh ) ] ;

[0022] , , u x(t) and Fourier transform, Let n be the center frequency and n be the number of iterations. As a secondary penalty factor, This is expressed as a noise tolerance parameter.

[0023] Optionally, the modal functions and the real-time meteorological data are input into a pre-trained wear prediction model based on the Transformer model to predict the wear degree of the target suspension clamp. This includes: aligning the modal functions obtained after variational mode decomposition with the real-time meteorological data according to timestamps to form multivariate input features, and inputting them into the input layer of the wear prediction model based on the Transformer model; dynamically calculating the interaction weights between each modal function through the multi-head self-attention mechanism in the wear prediction model based on the Transformer model to capture the nonlinear coupling relationship between the working state data and the meteorological data, and generating wear prediction values ​​corresponding to each modal function; and obtaining the predicted wear degree of the target suspension clamp based on weighted fusion according to the wear prediction values ​​corresponding to each modal function.

[0024] Optionally, after inputting the modal function and the real-time meteorological data into a pre-trained Transformer-based wear prediction model to predict the wear degree of the target suspension clamp, the method further includes: calculating the wear rate per unit time of the target suspension clamp based on the predicted wear degree. : In the formula, Indicates the predicted degree of wear. This represents the predicted time value; based on the current wear level of the target suspension clamp and its wear rate per unit time, the remaining lifespan of the target suspension clamp is calculated. In the formula, RL represents the remaining lifespan of the target suspension clamp. The threshold value representing the degree of wear, where h represents the degree of wear of the target suspension clamp at the current moment. This indicates the wear rate of the target suspension clamp per unit time.

[0025] Optionally, the method further includes: during the model training process of the wear prediction model based on the Transformer model, employing a pollination algorithm to optimize the model's learning rate, thereby improving the convergence speed and prediction accuracy of the wear prediction model based on the Transformer model; the optimization of the model's learning rate using the pollination algorithm includes:

[0026] Step 1: Initialize a set of candidate learning rates, denoted as X = {x1, x2, ..., x...} i ,...x n}; where x i It is the candidate learning rate corresponding to the i-th pollen;

[0027] Step 2: Randomly generate a random number ε. When ε < P, iteratively update the learning rate using a global pollination strategy. The specific formula is as follows:

[0028] Where P represents the switching probability P∈[0,1], used to weigh the relative importance of global and local pollination strategies. This represents the learning rate of the i-th pollen after the iteration update is completed; This represents the learning rate for the i-th pollen at the t-th iteration; The scaling factor that controls the step size, where t is the number of iterations; This represents the current optimal learning rate; The random step size of the Levy distribution is represented by the following formula:

[0029] ;

[0030] ;

[0031] s 2 = t (1+ l ) lt [(1+ l ) / 2] × sin ( pl / 2) 2 ( l -1) / 2 1 / l ;

[0032] Where U is a symmetric expression with a mean of 0 and a variance of 0. Random numbers that are normally distributed. It is a random number that follows a normal distribution with a mean of 0 and a variance of 1. For standard gamma functions, Take 1.5;

[0033] Step 3: When ε > P, iteratively update the learning rate using a local pollination strategy, as shown in the following formula:

[0034] ;in, is a random number that follows a uniform distribution on the interval [0,1]. and This represents a random selection in the current t-th iteration that is different from... The learning rate;

[0035] Step 4: Repeat steps 2-3 until the maximum number of iterations is reached to obtain the final optimal learning rate.

[0036] Optionally, the method further includes: comparing the wear degree predicted by the wear prediction model based on the Transformer model with a preset wear warning threshold; if the wear degree exceeds the preset wear warning threshold, generating a warning signal.

[0037] Secondly, embodiments of the present invention also provide a wear prediction system for a suspension clamp, the system comprising: a data acquisition module for acquiring real-time collected working status data of the target suspension clamp and real-time meteorological data; a data processing module for performing variational mode decomposition preprocessing on the working status data to obtain a mode function; and a wear prediction module for inputting the mode function and the real-time meteorological data into a pre-trained wear prediction model based on a Transformer model to predict the wear degree of the target suspension clamp; wherein, the training method of the wear prediction model based on the Transformer model includes: establishing A finite element model of the target suspension clamp is established. Based on the established finite element model, the working state data of the target suspension clamp under different working load conditions are obtained. The working state data includes contact pressure and contact slippage. The wear degree of the target suspension clamp corresponding to the working state data is obtained. The working state data of the target suspension clamp under different working load conditions, the wear degree of the target suspension clamp, and historical meteorological data are used as training datasets. The initial wear prediction model based on the Transformer model is trained according to the training datasets to obtain the wear prediction model based on the Transformer model.

[0038] Thirdly, embodiments of the present invention also provide an electronic device, the electronic device comprising: a housing, a processor, a memory, a circuit board, and a power supply circuit, wherein the circuit board is disposed within the space enclosed by the housing, and the processor and the memory are disposed on the circuit board; the power supply circuit is used to supply power to various circuits or devices of the above-mentioned electronic device; the memory is used to store executable program code; the processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, for executing the wear prediction method for the suspension clamp described in any of the first aspects above.

[0039] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing one or more programs, which can be executed by one or more processors to implement the wear prediction method for suspension clamps as described in any of the first aspects.

[0040] This invention provides a method and system for predicting the wear of suspension clamps. It uses working state data characterizing the stress response of the suspension clamp structure and meteorological data characterizing multi-source environmental data as input features to the model. Simultaneously, it utilizes variational mode decomposition to extract stable features, constructing a wear prediction model based on the Transformer model. The wear prediction model is then trained using finite element method combined with deep learning, effectively achieving accurate prediction of the wear degree of the suspension clamp and improving the accuracy of wear prediction. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a schematic flowchart of a method for predicting the wear of a suspension clamp according to an embodiment of the present invention;

[0043] Figure 2 A schematic flowchart of a method for predicting the wear of a suspension clamp according to another embodiment of the present invention;

[0044] Figure 3 This is a flowchart illustrating step S210 of the wear prediction method for suspension clamps provided in an embodiment of the present invention.

[0045] Figure 4 This is a schematic diagram of the wear prediction system architecture for a suspension clamp according to an embodiment of the present invention;

[0046] Figure 5 This is a schematic block diagram illustrating the architecture of an embodiment of the electronic device of the present invention. Detailed Implementation

[0047] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0048] It should be understood that the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0049] Example 1

[0050] See Figure 1 As shown, this embodiment of the invention provides a method for predicting the wear of a suspension clamp, including the following steps:

[0051] S110. Acquire real-time working status data of the target suspension clamp and real-time meteorological data.

[0052] Specifically, the working status data includes the contact pressure and contact slip of the target suspension clamp; the meteorological data includes wind speed, temperature and humidity.

[0053] In this step, sensors are pre-deployed near the target suspension clamp to monitor its operational status in real time. This prevents sudden damage to the suspension clamp from posing a risk to the power grid. When the wear condition of the suspension clamp cannot be observed, real-time operational status data is collected to predict its wear, facilitating timely maintenance or replacement. For example, a thin-film pressure sensor deployed at the contact point between the suspension clamp and the conductor measures the contact pressure; an eddy current displacement sensor fixed to the clamp's support non-contactly measures the displacement of the conductor relative to the clamp; and meteorological data is collected from an anemometer and temperature / humidity sensors at a nearby weather station. It should be noted that the operational status data of the suspension clamp may also include stress and strain, contact friction, and contact condition data, which are not limited in this application.

[0054] S120. Perform variational mode decomposition preprocessing on the working state data to obtain mode functions;

[0055] S130. Input the modal function and the real-time meteorological data into a pre-trained wear prediction model based on the Transformer model to predict the wear degree of the target suspension clamp.

[0056] In this step, after acquiring the working status data of the target suspension clamp and real-time meteorological data through various sensors, variational mode decomposition preprocessing is performed on the working status data to decompose the non-stationary working status data into stationary mode functions. Each mode function is a stationary time series, representing the state characteristics of different frequency bands. Then, the decomposed mode functions and meteorological time series data are time-aligned and input into a pre-trained wear prediction model based on the Transformer model to output the wear value of the target suspension clamp for future time periods, so as to predict the wear degree of the target suspension clamp.

[0057] The training method for the wear prediction model based on the Transformer model includes:

[0058] S210. Establish the finite element model of the target suspension clamp;

[0059] S220. Based on the established finite element model, obtain the working state data of the target suspension clamp under different working conditions and loads. The working state data includes: contact pressure and contact slippage.

[0060] S230, obtained Take The wear degree of the target suspension clamp corresponding to the working status data;

[0061] S240. Using the working status data of the target suspension clamp under different working conditions and loads, the wear degree of the target suspension clamp, and historical meteorological data as training datasets, the initial wear prediction model based on the Transformer model is trained based on the training datasets to obtain the wear prediction model based on the Transformer model.

[0062] The wear prediction method for suspension clamps provided in this embodiment uses working state data characterizing the stress response of the target suspension clamp structure and meteorological data characterizing multi-source environmental data as input features of the model. At the same time, variational mode decomposition is used to extract stable features to construct a wear prediction model based on the Transformer model. The wear prediction model is trained by combining finite element technology with deep learning, which effectively realizes the accurate prediction of the wear degree of suspension clamps and can improve the accuracy of wear prediction.

[0063] Optionally, in some embodiments, step S210, establishing the finite element model of the target suspension clamp, includes:

[0064] S211. Construct a three-dimensional geometric model of the target suspension clamp based on its material, geometric dimensions, and shape.

[0065] S212. Import the three-dimensional geometric model into the finite element analysis tool, divide it into meshes, refine the mesh locally in the contact area, set material properties and boundary conditions, and thus obtain the finite element model of the target suspension clamp.

[0066] In this embodiment, based on the actual dimensions of the target suspension clamp, such as its material, geometric dimensions, and shape parameters, a three-dimensional geometric model of the target suspension clamp is constructed using CAD software (e.g., Solidworks software). Complex parts, such as the contact surface, are locally refined. The three-dimensional model is imported into finite element software and meshed. At the same time, the contact area is locally meshed. According to the material, geometric dimensions, and shape of the target suspension clamp, corresponding material properties, such as elastic modulus, Poisson's ratio, density, and hardness, and boundary conditions are set to obtain the finite element model of the target suspension clamp.

[0067] Optionally, in some embodiments, step S220, obtaining the working state data of the target suspension clamp under different working loads based on the established finite element model, includes:

[0068] Different working loads are input into the finite element model of the target suspension clamp for calculation, and the contact pressure and contact slip of the target suspension clamp under different working loads are obtained.

[0069] In this embodiment, finite element analysis is performed on different working conditions (such as strong winds, icing, normal operation, etc.) to calculate the contact pressure and contact slip of the target suspension clamp under different working conditions, which are used as inputs for wear analysis.

[0070] Optionally, in some embodiments, step S230, obtaining the wear degree of the target suspension clamp corresponding to the working status data, includes: calculating the wear degree of the target suspension clamp corresponding to the working status data based on the Archard model.

[0071] The Archard model is represented as follows:

[0072] ;

[0073] In the formula, h represents the degree of wear; k represents the wear coefficient, which is obtained by fitting historical data of similar parts; p represents the contact pressure; s represents the contact slip; H represents the material hardness of the target suspension clamp; and D represents the thickness of the target suspension clamp.

[0074] In this embodiment, the wear degree of the target suspension clamp corresponding to the working state data of different working loads is calculated based on the Archard model; the working state data corresponding to different working loads, the wear degree of the target suspension clamp, and historical meteorological data are used as training datasets, and the initial wear prediction model based on the Transformer model is trained based on the training datasets to obtain the trained wear prediction model based on the Transformer model.

[0075] Optionally, in some embodiments, step S120, which involves performing variational mode decomposition preprocessing on the operating state data to obtain mode functions, includes:

[0076] The operating state data is adaptively decomposed using variational mode decomposition, and then decomposed into k mode functions by solving a constrained variational problem. This allows for the extraction of stable features;

[0077] The constrained variational problem is expressed as follows:

[0078] minutes { u k } ， { oh k } ∑ k=1 k || ∂ t [( d (t)+ j π t ) * u k (t)] e - j oh k t || 2 2 ;

[0079] ;

[0080] in, Let be the center frequency of the k-th modal function. Let k be the k-th mode function obtained from the decomposition, where k is the number of modes. Let j be the Dirac delta function, and j denote the complex unit, satisfying j 2 =-1, This represents the partial derivative with respect to time t. This represents the convolution calculation, where x(t) is the original non-stationary time series to be decomposed, and t represents the time variable.

[0081] The working state data is decomposed into k modal functions by solving a constrained variational problem. This includes: introducing a quadratic penalty factor and Lagrange multipliers to transform the constrained variational problem into an unconstrained variational problem, expressed as:

[0082] ;

[0083] Where a is the quadratic penalty factor, λ is the Lagrange multiplier, λ(t) is a time-dependent Lagrange multiplier function, and x(t) is the original non-stationary time series to be decomposed. To reconstruct the time series;

[0084] Update the variable { using the alternating direction multiplier algorithm. }、{ } and λ, solving an unconstrained variational problem, are expressed as:

[0085] u ̂ k n+1 ( oh )= x ̂ ( oh )- ∑ i ≠ k u ̂ i ( oh )+ l ̂ ( oh ) 2 1+2 α [ oh - oh k ] 2 ;

[0086] ;

[0087] l ̂ n+1 ( oh )= l ̂ n ( oh )+ t [x( oh )- ∑ k=1 k u k n+1 ( oh ) ] ;

[0088] , , u x(t) and Fourier transform, Let n be the center frequency and n be the number of iterations. As a secondary penalty factor, This is expressed as a noise tolerance parameter.

[0089] In this embodiment, variational mode decomposition (VMD) decomposes the original non-stationary operating state time series data into stationary mode functions with specific stable frequencies. This effectively eliminates noise and non-stationary components in the original operating state data. Furthermore, the decomposed mode functions can be aligned and fused with meteorological data, avoiding interference from high-frequency noise. Moreover, VMD also processes the operating state data in the training dataset during model training, providing more comprehensive and richer input features and improving the robustness of model predictions.

[0090] Optionally, in some embodiments, in step S120, the modal function and the real-time meteorological data are input into a pre-trained wear prediction model based on the Transformer model to predict the wear degree of the target suspension clamp, including:

[0091] The modal functions obtained after variational mode decomposition are aligned with the real-time meteorological data according to the timestamp to form multivariate input features, which are then input into the input layer of the wear prediction model based on the Transformer model.

[0092] By using the multi-head self-attention mechanism in the wear prediction model based on the Transformer model, the interaction weights between each modal function are dynamically calculated, the nonlinear coupling relationship between working status data and meteorological data is captured, and the wear prediction value corresponding to each modal function is generated.

[0093] Based on the wear prediction values ​​corresponding to each modal function, the predicted wear degree of the target suspension clamp is obtained through weighted fusion.

[0094] In this embodiment, before predicting the wear degree of the target suspension clamp using the wear prediction model based on the Transformer model, the data in the input model is preprocessed and features are constructed: variational mode decomposition is performed on the real-time collected working state data of the target suspension clamp to obtain mode functions; meteorological data and mode functions are aligned according to timestamps to form multivariate input features, and each input feature is standardized, for example, by Z-score, to eliminate dimensional differences; the multivariate input features are input to the input layer of the pre-trained wear prediction model based on the Transformer model; each mode function is mapped to a unified dimension through an independent fully connected layer; meteorological data, wind speed, temperature, and humidity are concatenated and mapped to the same latitude through a fully connected layer; the two types of features are concatenated in the feature dimension to form a fused input matrix; a multi-head self-attention mechanism is used to calculate the cross-feature attention weight between each mode function and the meteorological data; a prediction head (fully connected layer) is set separately for each mode component; the prediction results corresponding to each mode function are output; the meteorological data is extracted for trend through a separate temporal convolutional neural network and fused with the mode prediction results in a weighted manner.

[0095] Specifically, the wear prediction values ​​corresponding to each modal function are added together according to the time step to obtain the wear prediction value of the target suspension clamp:

[0096] In the formula, Indicates the predicted degree of wear. Let represent the predicted value of the k-th mode function at time t, where k is the total number of mode functions. Indicates the weather bias term. ,in, This represents a convolutional neural network used to capture the complex interactions of meteorological factors. This represents the temperature value at time t. This represents the humidity value at time t. This represents the wind speed value at time t.

[0097] Optionally, in some embodiments, after inputting the modal function and the real-time meteorological data into a pre-trained Transformer-based wear prediction model to predict the wear degree of the target suspension clamp, the method further includes:

[0098] Based on the predicted degree of wear, calculate the wear rate per unit time of the target suspension clamp. :

[0099] In the formula, Indicates the predicted degree of wear. Indicates the predicted time value;

[0100] Based on the current wear level of the target suspension clamp and its wear rate per unit time, the remaining lifespan of the target suspension clamp is calculated:

[0101] In the formula, RL represents the remaining lifespan of the target suspension clamp. The threshold value representing the degree of wear, where h represents the degree of wear of the target suspension clamp at the current moment. This indicates the wear rate of the target suspension clamp per unit time.

[0102] Optionally, in some embodiments, the method further includes:

[0103] During the training process of the wear prediction model based on the Transformer model, the flower pollination algorithm is used to optimize the learning rate of the model in order to improve the convergence speed and prediction accuracy of the wear prediction model based on the Transformer model.

[0104] The optimization of the learning rate of the model using the flower pollination algorithm includes the following steps:

[0105] Step 1: Initialize a set of candidate learning rates, denoted as X = {x1, x2, ..., x...} i ,...x n}; where x i It is the candidate learning rate corresponding to the i-th pollen;

[0106] Step 2: Randomly generate a random number ε. When ε < P, iteratively update the learning rate using a global pollination strategy. The specific formula is as follows:

[0107] Where P represents the switching probability P∈[0,1], used to weigh the relative importance of global and local pollination strategies. This represents the learning rate of the i-th pollen after the iteration update is completed; This represents the learning rate for the i-th pollen at the t-th iteration; The scaling factor that controls the step size, where t is the number of iterations; This represents the current optimal learning rate; The random step size of the Levy distribution is represented by the following formula:

[0108] ;

[0109] ;

[0110] s 2 = t (1+ l ) lt [(1+ l ) / 2] × sin ( pl / 2) 2 ( l -1) / 2 1 / l ;

[0111] Where U is a symmetric expression with a mean of 0 and a variance of 0. Random numbers that are normally distributed. It is a random number that follows a normal distribution with a mean of 0 and a variance of 1. For standard gamma functions, Take 1.5;

[0112] Step 3: When ε > P, iteratively update the learning rate using a local pollination strategy, as shown in the following formula:

[0113] ;in, is a random number that follows a uniform distribution on the interval [0,1]. and This represents a random selection in the current t-th iteration that is different from... The learning rate;

[0114] Step 4: Repeat steps 2 to 3 until the maximum number of iterations is reached to obtain the final optimal learning rate.

[0115] In this embodiment, the flower pollination algorithm optimizes key parameters of the model, such as the learning rate. During model training, the learning rate can be adjusted based on loss feedback. Compared to a fixed learning rate, it can quickly locate the high-performance range in the early stages and achieve stable convergence in the later stages. Furthermore, the optimal learning rate balances the stability and efficiency of gradient updates, improving the model's convergence speed and prediction accuracy. It should be noted that the flower pollination algorithm in this embodiment is only one type of optimization algorithm. Other optimization algorithms, such as genetic algorithms, simulated annealing algorithms, and intelligent optimization algorithms, can also be used to optimize the wear prediction model, such as the learning rate, the number of hidden layers, and the number of hidden layer nodes.

[0116] Optionally, in some embodiments, the method further includes: comparing the wear degree predicted by the wear prediction model based on the Transformer model with a preset wear warning threshold; if the wear degree exceeds the preset wear warning threshold, generating a warning signal.

[0117] In this embodiment, based on the wear prediction model of the present invention, a suspension clamp wear early warning system can also be established, a wear early warning threshold can be set, the wear degree predicted by the wear prediction model based on the Transformer model is compared with the wear early warning threshold, and when the wear early warning threshold is exceeded, an early warning signal is generated to remind maintenance personnel to perform timely inspection and maintenance. At the same time, based on the prediction model and the early warning system, an intelligent maintenance system can be developed to realize the automatic detection, prediction and maintenance of the target suspension clamp, thereby improving maintenance efficiency.

[0118] The wear prediction model in this invention is based on the Transformer model. The Transformer model's self-attention and multi-layer attention mechanisms effectively capture long-distance dependencies, overcoming the bottleneck of the LSTM model. This allows the model to better learn the complex relationship between various factors affecting the wear rate and lifespan of suspension clamps. Furthermore, the wear prediction model in this invention uses a pollination algorithm to optimize the learning rate, automatically searching for the optimal learning rate based on the model's loss function, thus improving the model's convergence speed and prediction accuracy. Taking a UHV transmission line as an example, a comparative experiment was conducted between traditional methods based on historical data and the method of this invention. Experimental results show that the method of this invention significantly outperforms traditional methods in terms of prediction accuracy, reducing the average prediction error by more than 44.2%. It should be noted that the wear prediction model in this invention can also employ other neural network models, such as the LSTM model and the GRU model, to predict the wear and lifespan of suspension clamps.

[0119] Furthermore, the finite element model in the embodiments of the present invention is not limited to the model in the above embodiments. A more simplified finite element model can also be used, such as ignoring the detailed structure of the suspension clamp or using a simplified material model to reduce the amount of calculation and improve the calculation efficiency; or an experimental model: by experimentally testing the wear of the suspension clamp under different working conditions, and using regression analysis and other methods to establish a wear and life prediction model to collect training datasets.

[0120] Example 2

[0121] This invention also provides a wear prediction system for suspension clamps, see below. Figure 4 As shown, the system includes:

[0122] Data acquisition module 41 is used to acquire real-time working status data of the target suspension clamp and real-time meteorological data;

[0123] Data processing module 42 is used to perform variational mode decomposition preprocessing on the working state data to obtain mode functions;

[0124] Wear prediction module 43 is used to input the modal function and the real-time meteorological data into a pre-trained wear prediction model based on the Transformer model to predict the wear degree of the target suspension clamp.

[0125] It also includes: a model training module, specifically used for:

[0126] Establish a finite element model of the target suspension clamp;

[0127] Based on the established finite element model, the working state data of the target suspension clamp under different working conditions and loads are obtained. The working state data includes: contact pressure and contact slip.

[0128] Obtain the wear degree of the target suspension clamp corresponding to the working status data;

[0129] The working status data of the target suspension clamp under different working conditions and loads, the wear degree of the target suspension clamp, and historical meteorological data are used as training datasets. The initial wear prediction model based on the Transformer model is trained based on the training datasets to obtain the wear prediction model based on the Transformer model.

[0130] Optionally, in some embodiments, the model training module is specifically used to construct a three-dimensional geometric model of the target suspension clamp based on its material, geometric dimensions, and shape; import the three-dimensional geometric model into a finite element analysis tool, mesh it, refine the mesh locally in the contact area, and set material properties and boundary conditions to obtain the finite element model of the target suspension clamp.

[0131] Optionally, in some embodiments, the model training module is further used to input different working loads into the finite element model of the target suspension clamp for solution calculation, so as to obtain the contact pressure and contact slip of the target suspension clamp under different working loads.

[0132] Optionally, in some embodiments, the model training module is further used to calculate the wear degree of the target suspension clamp corresponding to the working state data based on the Archard model.

[0133] Optionally, in some embodiments, the Archard model is represented as:

[0134] ;

[0135] In the formula, h represents the wear depth; k represents the wear coefficient, which is obtained by fitting historical data of similar parts; p represents the contact pressure; s represents the contact slip; H represents the material hardness of the target suspension clamp; and D represents the thickness of the target suspension clamp.

[0136] Optionally, in some embodiments, the data processing module is used to adaptively decompose the working state data using variational mode decomposition, decomposing the working state data into k mode functions by solving a constrained variational problem. This allows for the extraction of stable features;

[0137] The constrained variational problem is expressed as follows:

[0138] minutes { u k } ， { oh k } ∑ k=1 k || ∂ t [( d (t)+ j π t ) * u k (t)] e - j oh k t || 2 2 ;

[0139] ;

[0140] in, Let be the center frequency of the k-th modal function. Let k be the k-th mode function obtained from the decomposition, where k is the number of modes. Let j be the Dirac delta function, and j denote the complex unit, satisfying j 2 =-1, This represents the partial derivative with respect to time t. This represents the convolution calculation, where x(t) is the original non-stationary time series to be decomposed, and t represents the time variable.

[0141] Optionally, in some embodiments, the data processing module is further configured to introduce a quadratic penalty factor and Lagrange multipliers to transform the constrained variational problem into an unconstrained variational problem, expressed as:

[0142] ;

[0143] Where a is the quadratic penalty factor, λ is the Lagrange multiplier, λ(t) is a time-dependent Lagrange multiplier function, and x(t) is the original non-stationary time series to be decomposed. To reconstruct the time series;

[0144] Update the variable { using the alternating direction multiplier algorithm. }、{ } and λ, solving an unconstrained variational problem, are expressed as:

[0145] u ̂ k n+1 ( oh )= x ̂ ( oh )- ∑ i ≠ k u ̂ i ( oh )+ l ̂ ( oh ) 2 1+2 α [ oh - oh k ] 2 ;

[0146] ;

[0147] l ̂ n+1 ( oh )= l ̂ n ( oh )+ t [x( oh )- ∑ k=1 k u k n+1 ( oh ) ] ;

[0148] , , u x(t) and Fourier transform, Let n be the center frequency and n be the number of iterations. As a secondary penalty factor, This is expressed as a noise tolerance parameter.

[0149] The model prediction module is specifically used to align the modal functions obtained after variational mode decomposition with the real-time meteorological data according to timestamps to form multivariate input features, which are then input into the input layer of the pre-trained Transformer-based wear prediction model. Through the multi-head self-attention mechanism in the Transformer-based wear prediction model, the interaction weights between each modal function are dynamically calculated to capture the nonlinear coupling relationship between the working state data and the meteorological data, generating wear prediction values ​​corresponding to each modal function. Based on the wear prediction values ​​corresponding to each modal function, the predicted wear degree of the target suspension clamp is obtained through weighted fusion.

[0150] The system further includes a life prediction module, used to calculate the wear rate per unit time of the target suspension clamp based on the predicted wear degree after inputting the modal function and the real-time meteorological data into a pre-trained wear prediction model based on the Transformer model. :

[0151] In the formula, Indicates the predicted degree of wear. Indicates the predicted time value;

[0152] Based on the current wear level of the target suspension clamp and its wear rate per unit time, the remaining lifespan of the target suspension clamp is calculated:

[0153] In the formula, RL represents the remaining lifespan of the target suspension clamp. The threshold value representing the degree of wear, where h represents the degree of wear of the target suspension clamp at the current moment. This indicates the wear rate of the target suspension clamp per unit time.

[0154] The system further includes a wear warning module, which compares the wear degree predicted by the wear prediction model based on the Transformer model with a preset wear warning threshold, and generates a warning signal if the wear degree exceeds the preset wear warning threshold.

[0155] Example 3

[0156] Figure 5 This is a schematic block diagram of the architecture of an embodiment of the electronic device of the present invention; based on the same technical concept as the foregoing Embodiment 1, the electronic device provided by the embodiments of the present invention, such as... Figure 5 As shown, the steps and flow of any of the embodiments described in Embodiment 1 of the present invention can be implemented.

[0157] The aforementioned electronic device may include a processor 51 and a memory 52, wherein the memory 52 is used to store executable program code; the processor 51 runs a program corresponding to the executable program code by reading the executable program code stored in the memory 52, for executing the wear prediction method of the suspension clamp described in any of the preceding embodiments.

[0158] For details on the specific execution process of the above steps by the processor 51 and the steps further executed by the processor 51 by running executable program code, please refer to the description of Embodiment 1 of the present invention, which will not be repeated here.

[0159] This invention also provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the wear prediction method for suspension clamps described in any of the preceding embodiments.

[0160] The electronic device exists in various forms, including but not limited to: (1) Mobile communication devices: These devices are characterized by their mobile communication capabilities and are primarily designed to provide voice and data communication. These terminals include smartphones (such as iPhones), multimedia phones, feature phones, and low-end phones.

[0161] (2) Ultra-mobile personal computer devices: These devices fall under the category of personal computers, possessing computing and processing capabilities, and generally also have mobile internet access features. These terminals include PDAs, MIDs, and UMPCs, such as the iPad.

[0162] (3) Portable entertainment devices: These devices can display and play multimedia content. This category includes audio and video players (such as iPods), handheld game consoles, e-book readers, as well as smart toys and portable car navigation devices.

[0163] (4) Server: A device that provides computing services. The components of a server include a processor, hard disk, memory, system bus, etc. Servers are similar to general computer architectures, but because they need to provide highly reliable services, they have higher requirements in terms of processing power, stability, reliability, security, scalability, and manageability.

[0164] (5) Other electronic devices with data interaction functions.

[0165] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0166] The various embodiments in this specification are described in a related manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0167] For ease of description, if systems, servers, etc. are involved, they may be described separately as various units / modules based on their functions. Of course, in implementing this invention, the functions of each unit / module can be implemented in one or more software and / or hardware.

[0168] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0169] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for predicting the wear of a suspension clamp, characterized in that, include: Acquire real-time data on the working status of the target suspension clamp and real-time meteorological data; The working state data is preprocessed by variational mode decomposition to obtain the mode functions; The modal function and the real-time meteorological data are input into a pre-trained wear prediction model based on the Transformer model to predict the wear degree of the target suspension clamp. The training method for the wear prediction model based on the Transformer model includes: Establish a finite element model of the target suspension clamp; Based on the established finite element model, the working state data of the target suspension clamp under different working conditions and loads are obtained. The working state data includes: contact pressure and contact slippage. Obtain the wear degree of the target suspension clamp corresponding to the working status data; The working status data of the target suspension clamp under different working conditions and loads, the wear degree of the suspension clamp, and historical meteorological data are used as training datasets. The initial wear prediction model based on the Transformer model is trained based on the training datasets to obtain the wear prediction model based on the Transformer model.

2. The wear prediction method for suspension clamps according to claim 1, characterized in that, The establishment of the finite element model of the target suspension clamp includes: Construct a three-dimensional geometric model of the target suspension clamp based on its material, geometric dimensions, and shape; The three-dimensional geometric model is imported into a finite element analysis tool, meshed, and the mesh is locally refined in the contact area. Material properties and boundary conditions are set to obtain the finite element model of the target suspension clamp.

3. The wear prediction method for suspension clamps according to claim 2, characterized in that, Based on the established finite element model, the working state data of the target suspension clamp under different working load conditions are obtained, including: Different working loads are input into the finite element model of the target suspension clamp for calculation, and the contact pressure and contact slip of the target suspension clamp under different working loads are obtained.

4. The wear prediction method for suspension clamps according to claim 3, characterized in that, The degree of wear of the target suspension clamp corresponding to the obtained working status data includes: The wear degree of the target suspension clamp corresponding to the working status data is calculated based on the Arcard model.

5. The wear prediction method for suspension clamps according to claim 4, characterized in that, The Archard model is represented as follows: ； In the formula, h represents the degree of wear; k represents the wear coefficient, which is obtained by fitting historical data of similar parts; p represents the contact pressure; s represents the contact slip; H represents the material hardness of the target suspension clamp; and D represents the thickness of the target suspension clamp.

6. The wear prediction method for suspension clamps according to claim 1, characterized in that, The step of performing variational mode decomposition preprocessing on the working state data to obtain mode functions includes: The working state data is adaptively decomposed using variational mode decomposition, and the working state data is decomposed into... by solving a constrained variational problem. k Modal functions This allows for the extraction of stable features; The constrained variational problem is expressed as follows: ； ； in, For the first k The center frequency of each modal function For the k-th mode function obtained from the decomposition, k For the number of modes, Let j be the Dirac delta function, and j denote the complex unit, satisfying... j 2 =-1, This represents the partial derivative with respect to time t. This represents convolution calculation. x ( t ) represents the original non-stationary time series to be decomposed, and t represents the time variable.

7. The wear prediction method for suspension clamps according to claim 6, characterized in that, The working state data is decomposed into... by solving a constrained variational problem. k Modal functions include: By introducing a quadratic penalty factor and Lagrange multipliers, the constrained variational problem is transformed into an unconstrained variational problem, expressed as: ； in, a Let λ be the quadratic penalty factor, λ be the Lagrange multiplier, and λ(t) be a time-dependent Lagrange multiplier function. x ( t ( ) represents the original non-stationary time series to be decomposed. To reconstruct the time series; Update the variable { using the alternating direction multiplier algorithm. }、{ } and λ, solving an unconstrained variational problem, are expressed as: ； ； ； , , They are respectively u , x ( t )and Fourier transform, Let n be the center frequency and n be the number of iterations. As a secondary penalty factor, This is expressed as a noise tolerance parameter.

8. The wear prediction method for suspension clamps according to claim 1, characterized in that, The modal function and the real-time meteorological data are input into a pre-trained wear prediction model based on the Transformer model to predict the wear degree of the target suspension clamp, including: The modal functions obtained after variational mode decomposition are aligned with the real-time meteorological data according to the timestamp to form multivariate input features, which are then input into the input layer of the prediction model based on the Transformer model. By using the multi-head self-attention mechanism in the wear prediction model based on the Transformer model, the interaction weights between each modal function are dynamically calculated, the nonlinear coupling relationship between working status data and meteorological data is captured, and the wear prediction value corresponding to each modal function is generated. Based on the wear prediction values ​​corresponding to each modal function, the predicted wear degree of the target suspension clamp is obtained through weighted fusion.

9. The wear prediction method for suspension clamps according to claim 1, characterized in that, After inputting the modal function and the real-time meteorological data into a pre-trained Transformer-based wear prediction model to predict the wear degree of the target suspension clamp, the method further includes: Based on the predicted degree of wear, calculate the wear rate per unit time of the target suspension clamp. : In the formula, Indicates the predicted degree of wear. Indicates the predicted time value; Based on the current wear level of the target suspension clamp and the wear rate per unit time of the target suspension clamp, the remaining life of the target suspension clamp is calculated: In the formula, RL represents the remaining lifespan of the target suspension clamp. The threshold value representing the degree of wear, where h represents the degree of wear of the target suspension clamp at the current moment. This indicates the wear rate of the target suspension clamp per unit time.

10. The wear prediction method for suspension clamps according to claim 1, characterized in that, The method further includes: During the training process of the wear prediction model based on the Transformer model, the flower pollination algorithm is used to optimize the learning rate of the model in order to improve the convergence speed and prediction accuracy of the wear prediction model based on the Transformer model. The learning rate of the model optimized using the flower pollination algorithm includes: Step 1: Initialize a set of candidate learning rates, denoted as X = { x 1, x 2,..., x i ,... x n };in, x i It is the candidate learning rate corresponding to the i-th pollen; Step 2: Randomly generate a random number ε. When ε < P, iteratively update the learning rate using a global pollination strategy. The specific formula is as follows: Where P represents the switching probability P∈[0,1], used to weigh the relative importance of global and local pollination strategies. This represents the learning rate of the i-th pollen after the iteration update is completed; This represents the learning rate for the i-th pollen at the t-th iteration; The scaling factor that controls the step size, where t is the number of iterations; This represents the current optimal learning rate; The random step size of the Levy distribution is represented by the following formula: ； ； ； Where U is a symmetric expression with a mean of 0 and a variance of 0. Random numbers that are normally distributed. It is a random number that follows a normal distribution with a mean of 0 and a variance of 1. For standard gamma functions, Take 1.5; Step 3: When ε > P, iteratively update the learning rate using a local pollination strategy, as shown in the following formula: ;in, is a random number that follows a uniform distribution on the interval [0,1]. and This represents a random selection in the current t-th iteration that is different from... The learning rate; Step 4: Repeat steps 2 to 3 until the maximum number of iterations is reached to obtain the final optimal learning rate.

11. The wear prediction method for suspension clamps according to claim 1, characterized in that, The method further includes: The wear level predicted by the wear prediction model based on the Transformer model is compared with a preset wear warning threshold. If the wear level exceeds the preset wear warning threshold, a warning signal is generated.

12. A wear prediction system for suspension clamps, characterized in that, The system includes: The data acquisition module is used to acquire real-time working status data of the target suspension clamp and real-time meteorological data; The data processing module is used to perform variational mode decomposition preprocessing on the working state data to obtain mode functions; The wear prediction module is used to input the modal function and the real-time meteorological data into a pre-trained wear prediction model based on the Transformer model to predict the wear degree of the target suspension clamp. The training method for the wear prediction model based on the Transformer model includes: Establish a finite element model of the target suspension clamp; Based on the established finite element model, the working state data of the target suspension clamp under different working conditions and loads are obtained. The working state data includes: contact pressure and contact slippage. Obtain the wear degree of the target suspension clamp corresponding to the working status data; The working status data of the target suspension clamp under different working conditions and loads, the wear degree of the target suspension clamp, and historical meteorological data are used as training datasets. The initial wear prediction model based on the Transformer model is trained based on the training datasets to obtain the wear prediction model based on the Transformer model.

13. An electronic device, characterized in that, include: Processor and memory; wherein, memory is used to store executable program code; The processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, for executing the wear prediction method for the suspension clamp as described in any one of claims 1 to 11.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the wear prediction method for suspension clamps according to any one of claims 1 to 11.

Images