A method and system for intelligent diagnosis of slope damage for construction of photovoltaics

Abstract

The application provides a slope damage intelligent diagnosis method and system for building photovoltaics. It belongs to the field of slope damage diagnosis. The method comprises the following steps: according to the spatial and temporal variability of the multi-source parameters of the slope photovoltaic anchoring system, a three-dimensional photovoltaic structure-anchor-rock-soil coupled stress wave propagation numerical simulation model is established, and simulation data sets under different damage positions and damage degrees of the anchoring system are obtained; the simulation data sets are input into a CNN-LSTM deep learning model for training; the optimal excitation frequency is calibrated; stress wave excitation is applied according to the calibrated optimal excitation frequency, and wave response data, environmental compensation data and structure state data are synchronously collected; the data are input into the trained diagnosis model, and the damage position and damage degree are inversely output; the damage position is mapped to a specific anchor rod number according to the topological relationship of the photovoltaic array, and a graded early warning is triggered according to the damage degree. The method can accurately locate the damage position of the anchoring structure and quantify the damage degree.

Description

Technical Field

[0001] This application relates to the field of slope damage diagnosis, and more specifically, to an intelligent method and system for diagnosing slope damage for photovoltaic construction. Background Technology

[0002] In recent years, slope photovoltaic (PV) power stations have been widely used in mountainous and hilly areas. Slope PV systems typically use micropiles or anchor foundations to anchor the PV supports to the slope's soil and rock. The anchoring structure, as a key load-bearing component connecting the upper PV structure to the lower slope, directly affects the structural safety and long-term stable operation of the entire PV system. During long-term service, under the coupled effects of continuous wind loads, temperature cycles, rainfall infiltration, and slope creep, the anchoring structure is prone to damage such as prestress loss, anchor loosening, anchor corrosion, and even anchor failure. If these damages are not detected and addressed in time, they may lead to overall instability of the PV array or even slope collapse.

[0003] Patent application CN120509762B proposes a method for evaluating the safety of photovoltaic slopes that considers the spatiotemporal variability of multi-source parameters. This method can assess the safety status of the slope as a whole and output a safety factor and failure probability. However, this method cannot diagnose the specific root causes of slope instability, such as whether there is damage to the anchoring structure, where the damage is located, and the extent of the damage.

[0004] Therefore, there is an urgent need for an intelligent diagnostic method that can accurately locate the damage position of the anchoring structure and quantify the degree of damage. Summary of the Invention

[0005] The purpose of this application is to provide an intelligent diagnostic method and system for slope damage in photovoltaic construction, which can accurately locate the damage location of the anchoring structure and quantify the degree of damage.

[0006] This application is implemented as follows: In a first aspect, this application provides an intelligent diagnosis method for slope damage in photovoltaic construction, comprising the following steps: S1: Based on the spatiotemporal variability of multi-source parameters of the slope photovoltaic anchoring system, a three-dimensional photovoltaic structure-anchor-soil coupled stress wave propagation numerical simulation model is established to obtain simulation datasets of the anchoring system under different damage locations and damage degrees. S2: Input the simulation dataset into the CNN-LSTM deep learning model for training. During training, define a loss function that includes physical constraints: ; ; in, For the total loss function, For data fitting terms, This is a physical consistency penalty term based on the one-dimensional wave equation. For balance coefficient, The total number of training samples, Let i be the displacement field of the i-th training sample. Using time as the coordinate, Let be the spatial coordinates along the anchor bolt axis. Let be the propagation velocity of the stress wave in the anchor bolt. The damping ratio of the soil and rock mass. L2 norm operator; S3: Deploy active excitation devices and sensor arrays in the slope photovoltaic anchoring system, perform frequency sweep excitation on the anchor rod under test, and calibrate the optimal excitation frequency; apply stress wave excitation according to the calibrated optimal excitation frequency, and simultaneously collect wave response data, environmental compensation data, and structural status data; S4: Input the collected fluctuation response data, environmental compensation data, and structural state data into the trained diagnostic model, and output the damage location and damage degree; map the damage location to the specific anchor number according to the photovoltaic array topology, and trigger graded early warning according to the damage degree.

[0007] Based on the first aspect, step S1 includes: S11: Establish a three-dimensional numerical simulation model of stress wave propagation coupled with photovoltaic structure-anchor bolt-soil, and determine the input parameter vector of the simulation model. Its expression is: ; in, The length of the anchor bolt. The diameter of the anchor rod. The elastic modulus of the anchor bolt. For anchor bolt density, The longitudinal wave velocity of the soil and rock mass. The transverse wave velocity of the rock and soil mass. The damping ratio of the soil and rock mass. For the photovoltaic structure location parameters, For the excitation center frequency, For temperature, The water content of the soil and rock mass. Location of injury The degree of damage; S12: Apply transient excitation to the end of the model anchor bolt, calculate the propagation response of the stress wave in the structure, and extract stress wave characteristics, including wave velocity. Energy attenuation coefficient and main frequency offset Its expression is: ; ; ; in, Let j be the axial coordinate of the j-th measuring point. Let j be the time when the first wave of the stress wave arrives at the j-th measuring point. The peak acceleration at the j-th measuring point is... Main frequency, The center frequency of the excitation signal; S13: Combine the stress wave characteristics obtained in step S12 with the environmental parameters in the simulation input to obtain the stress wave feature vector. Its expression is: ; S14: Location of damage and degree of damage As the output label, it is compared with the stress wave feature vector obtained in step S13. Pairing samples together forms a training sample, and all simulation samples constitute the simulation dataset. Its expression is: .

[0008] Based on the first aspect, the CNN-LSTM deep learning model in step S2 is used to establish a nonlinear mapping relationship from stress wave response signal features to damage parameters, and its expression is: ; in, For the input feature vector, The mapping function is parameterized by the CNN-LSTM hybrid neural network. These are the learnable parameters of the network. Location of injury The degree of damage.

[0009] Based on the first aspect, step S3 includes: S31: Install an active excitation device and a broadband response sensor array at the exposed end of the anchor bolt to be tested in the photovoltaic array on the slope; S32: Perform frequency sweep excitation on the anchor rod under test, collect the response signal at each frequency, and calculate the coherence coefficient. and signal-to-noise ratio Select the optimal frequency The diagnostic excitation frequency of this anchor bolt is expressed as follows: ; in, , Weighting coefficient; coherence coefficient The expression is: ; in, To determine the power spectrum of the excitation signal, In response to the power spectrum of the signal, The cross-power spectrum of the excitation signal and the response signal; S33: According to the calibrated optimal frequency Stress wave excitation is applied, and wave response data, environmental compensation data, and structural state data are collected simultaneously. S34: Extract features from the collected fluctuation response data to construct the model input vector. Its expression is: ; in, For the number of measurement points, For the wave velocity of each measurement segment, The energy attenuation coefficient for each measurement segment is... Main frequency offset For temperature, This represents the water content of the soil and rock mass.

[0010] Based on the first aspect, step S4 includes: S41: The on-site feature vector constructed in step S3 Input the CNN-LSTM diagnostic model trained in step S2 to obtain the damage parameters, the expression of which is: ; in, For the predicted location of the damage, To predict the extent of damage, For the trained CNN-LSTM diagnostic model; S42: The damage location output in step S41 Mapping to a specific anchor bolt number yields spatial location information containing the anchor bolt number and damage depth; S43: Based on the degree of damage Triggering a tiered warning: When A green alert is triggered and logged; when A yellow alert is triggered, prompting on-site verification and handling to be completed within one month; when A red alert is triggered, prompting an immediate shutdown and reinforcement.

[0011] Based on the first aspect, in step S11, considering the spatial variability of soil and rock parameters, the soil and rock parameters are regarded as spatial random fields, and Gaussian autocorrelation function is used to describe their spatial correlation. Latin hypercube sampling is used to randomly select N sets of random parameter combinations as the parameter input of the simulation model.

[0012] Based on the first aspect, the CNN-LSTM deep learning model in step S2 consists of two one-dimensional convolutional layers, one LSTM layer, one dropout layer, and two fully connected layers.

[0013] Based on the first aspect, the training parameters in step S2 are as follows: the Adam optimizer is used, the initial learning rate is set to 0.001, and the batch size is 32; training is terminated when the validation set loss no longer decreases for 20 consecutive rounds, and the model parameters with the minimum validation set loss are saved.

[0014] Secondly, this application provides an intelligent slope damage diagnosis system for photovoltaic construction, comprising: The data acquisition module is used to establish a three-dimensional numerical simulation model of the propagation of coupled stress waves between the photovoltaic structure, anchor rod, and soil and rock based on the spatiotemporal variability of the multi-source parameters of the slope photovoltaic anchoring system, and to obtain simulation datasets of the anchoring system under different damage locations and damage degrees. The model training module is used to input simulation datasets into the CNN-LSTM deep learning model for training. During training, a loss function with physical constraints is defined. ; ; in, For the total loss function, For data fitting terms, This is a physical consistency penalty term based on the one-dimensional wave equation. For balance coefficient, The total number of training samples, Let i be the displacement field of the i-th training sample. Using time as the coordinate, Let be the spatial coordinates along the anchor bolt axis. Let be the propagation velocity of the stress wave in the anchor bolt. The damping ratio of the soil and rock mass. L2 norm operator; The data acquisition module is calibrated and used to deploy active excitation devices and sensor arrays in the slope photovoltaic anchoring system, perform frequency sweep excitation on the anchor rod under test, and calibrate the optimal excitation frequency; apply stress wave excitation according to the calibrated optimal excitation frequency, and simultaneously acquire wave response data, environmental compensation data, and structural status data. The output early warning module is used to input the collected fluctuation response data, environmental compensation data, and structural state data into the trained diagnostic model, and inversely output the damage location and damage degree; according to the photovoltaic array topology, the damage location is mapped to the specific anchor number, and a graded early warning is triggered according to the damage degree.

[0015] Thirdly, this application provides an electronic device, comprising: Memory, used to store one or more programs; processor; When one or more programs are executed by the processor, the method described above is implemented.

[0016] Compared with the prior art, this application has at least the following advantages or beneficial effects: 1. Capable of precise localization of damage. Existing technology (CN120509762B) can only assess the overall safety status (safety factor, failure probability) of a slope photovoltaic system, but cannot pinpoint the exact location of the problem. This invention, through the propagation response of stress waves in the anchor bolts, inversely determines the location and extent of damage, achieving precise localization from the overall to the local level, providing an actionable decision-making basis for accurate maintenance.

[0017] 2. This invention addresses the poor interpretability of purely data-driven models through physical constraint mechanisms. Traditional deep learning methods directly learn mapping relationships from data, potentially learning statistically reasonable but physically erroneous patterns (such as wave velocity increasing with damage). This invention introduces a physical consistency penalty term based on a one-dimensional wave equation into the loss function, giving the diagnostic results clear physical interpretability.

[0018] 3. Improving diagnostic robustness through multi-dimensional stress wave feature fusion. This invention simultaneously extracts stress wave features from three dimensions: wave velocity, energy attenuation coefficient, and dominant frequency shift, and fuses them with environmental compensation data. This enables the model to distinguish between response changes caused by environmental variations and response changes caused by damage, thereby improving the robustness and accuracy of diagnosis. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of an intelligent slope damage diagnosis method for photovoltaic construction according to this application; Figure 2 This is a schematic diagram of the CNN-LSTM deep learning model in an intelligent diagnosis method for slope damage in photovoltaic construction according to this application. Figure 3 This is a schematic diagram of the intelligent slope damage diagnosis system for photovoltaic construction according to this application; Figure 4 This is a schematic diagram of the structure of an electronic device according to this application.

[0021] icon: 1. Data acquisition module; 2. Model training module; 3. Calibration data acquisition module; 4. Output warning module; 5. Processor; 6. Memory; 7. Communication interface. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0023] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the various embodiments and features described below can be combined with each other.

[0024] Example This application provides a method and system for intelligent diagnosis of slope damage for photovoltaic construction, which can accurately locate the location of damage to the anchoring structure and quantify the degree of damage.

[0025] Please refer to Figure 1 The intelligent diagnosis method for slope damage used in photovoltaic construction includes the following steps: S1: Based on the spatiotemporal variability of multi-source parameters of the slope photovoltaic anchoring system, a three-dimensional photovoltaic structure-anchor-soil coupled stress wave propagation numerical simulation model is established to obtain simulation datasets of the anchoring system under different damage locations and damage degrees. This step establishes a three-dimensional numerical simulation model of stress wave propagation coupled with a photovoltaic structure and anchor bolts, generating a multi-condition simulation dataset covering different damage locations and degrees, providing sufficient labeled sample data for deep learning model training. As one implementation method, step S1 includes: S11: Establish a three-dimensional numerical simulation model of stress wave propagation coupled with photovoltaic structure-anchor bolt-soil, and determine the input parameter vector of the simulation model. Its expression is: ; in, The length of the anchor bolt. The diameter of the anchor rod. The elastic modulus of the anchor bolt. For anchor bolt density, The longitudinal wave velocity of the soil and rock mass. The transverse wave velocity of the rock and soil mass. The damping ratio of the soil and rock mass. For the photovoltaic structure location parameters, For the excitation center frequency, For temperature, The water content of the soil and rock mass. Location of injury To determine the degree of damage; in this embodiment, the preferred anchor bolt length is... =2.5m, anchor bolt diameter =28mm, anchor bolt elastic modulus =200GPa, anchor bolt density =7850kg / m³. Geotechnical parameters. , , The model is treated as a spatial random field, and its spatial correlation is described by a Gaussian autocorrelation function with a correlation length of 1.0 m. N=2000 random parameter combinations are randomly selected using Latin hypercube sampling as the parameter input for the simulation model. Through Latin hypercube sampling and spatial random field modeling, the spatial variability of soil and rock parameters is fully considered, generating a statistically representative simulation dataset that provides diverse damage condition samples for model training.

[0026] S12: Apply transient excitation to the end of the model anchor bolt, calculate the propagation response of the stress wave in the structure, and extract stress wave characteristics, including wave velocity. Energy attenuation coefficient and main frequency offset Its expression is: ; ; ; in, For the first Axial coordinates of each measuring point For the first wave of stress wave to reach the first The time at each measuring point For the first Peak acceleration at each measuring point Main frequency, The center frequency of the excitation signal is used; the wave velocity in this step reflects the change in the elastic modulus of the anchor bolt, the energy attenuation coefficient reflects the change in the damping characteristics of the anchor bolt, and the dominant frequency offset reflects the dispersion effect caused by damage. These three features characterize the damage state of the anchor bolt from different dimensions, providing multi-dimensional feature inputs for subsequent diagnosis.

[0027] S13: Combine the stress wave characteristics obtained in step S12 with the environmental parameters in the simulation input to obtain the stress wave feature vector. Its expression is: ; This step integrates environmental compensation data (temperature, moisture content) with stress wave characteristics, enabling the model to distinguish between response changes caused by environmental changes and response changes caused by damage, thereby improving the robustness of the diagnosis.

[0028] S14: Location of damage and degree of damage As the output label, it is compared with the stress wave feature vector obtained in step S13. Pairing samples together forms a training sample, and all simulation samples constitute the simulation dataset. Its expression is: .

[0029] In this step, each simulation sample contains both an input feature vector and an output label, forming a complete supervised learning sample, which provides a data foundation for training the CNN-LSTM model.

[0030] S2: Input the simulation dataset into the CNN-LSTM deep learning model for training. During training, define a loss function that includes physical constraints: ; ; in, For the total loss function, For data fitting terms, This is a physical consistency penalty term based on the one-dimensional wave equation. For balance coefficient, The total number of training samples, Let i be the displacement field of the i-th training sample. Using time as the coordinate, Let be the spatial coordinates along the anchor bolt axis. Let be the propagation velocity of the stress wave in the anchor bolt. The damping ratio of the soil and rock mass. L2 norm operator; This step trains a CNN-LSTM deep learning model using a simulation dataset and embeds a one-dimensional wave equation as a physical constraint into the loss function, enabling the model to learn a damage diagnosis mapping relationship that conforms to the physical laws of stress wave propagation. As one implementation method, the CNN-LSTM deep learning model in step S2 is used to establish a nonlinear mapping relationship from stress wave response signal features to damage parameters, expressed as: ; in, For the input feature vector, The mapping function is parameterized by the CNN-LSTM hybrid neural network. These are the learnable parameters of the network. Location of injury The degree of damage.

[0031] By introducing The penalty term forces the stress wave response of the model to satisfy the wave equation, making the diagnostic results physically interpretable.

[0032] S3: Deploy active excitation devices and sensor arrays in the slope photovoltaic anchoring system, perform frequency sweep excitation on the anchor rod under test, and calibrate the optimal excitation frequency; apply stress wave excitation according to the calibrated optimal excitation frequency, and simultaneously collect wave response data, environmental compensation data, and structural status data; This step involves deploying a sensor array on the actual slope photovoltaic anchoring system, determining the optimal excitation frequency for each anchor rod through frequency sweep calibration, and collecting on-site fluctuation response data and environmental compensation data. As one implementation method, step S3 includes: S31: Install an active excitation device and a broadband response sensor array at the exposed end of the anchor bolts to be tested in the slope photovoltaic array; during the deployment phase of the slope photovoltaic system, pre-install intelligent detection nodes at the exposed end of each anchor bolt. Each node includes: a piezoelectric ceramic (PZT) excitation plate (15mm in diameter, 1mm thick, bonded to the center of the anchor bolt end face with epoxy resin), and a broadband accelerometer (sensitivity ≥100mV / g, frequency response 1). The system consists of a 100kHz sensor array (attached to the anchor bolt end face), a temperature sensor, a moisture content sensor, and a wireless communication module (LoRa). The fixed sensor array avoids installation errors and inconsistencies caused by reinstalling sensors for each inspection; the wireless communication module supports remote wake-up and batch inspection.

[0033] S32: Perform frequency sweep excitation on the anchor rod under test, collect the response signal at each frequency, and calculate the coherence coefficient. and signal-to-noise ratio Select the optimal frequency The diagnostic excitation frequency of this anchor bolt is expressed as follows: ; in, , Weighting coefficient; coherence coefficient The expression is: ; in, To determine the power spectrum of the excitation signal, In response to the power spectrum of the signal, The cross-power spectrum of the excitation signal and the response signal; The anchoring conditions and boundary conditions of each anchor rod differ in this step. Frequency sweep calibration ensures that the excitation frequency matches the dynamic characteristics of the anchor rod, maximizes the signal-to-noise ratio and coherence, and improves measurement accuracy.

[0034] S33: According to the calibrated optimal frequency Stress wave excitation is applied, and wave response data, environmental compensation data, and structural state data are collected simultaneously. S34: Extract features from the collected fluctuation response data to construct the model input vector. Its expression is: ; in, For the number of measurement points, For the wave velocity of each measurement segment, The energy attenuation coefficient for each measurement segment is... Main frequency offset For temperature, This represents the water content of the soil and rock mass.

[0035] In this step, the field feature vectors are in the same format as the feature vectors in the simulation dataset, ensuring that the trained model can be directly applied to the field data.

[0036] S4: Input the collected fluctuation response data, environmental compensation data, and structural state data into the trained diagnostic model, and output the damage location and damage degree; map the damage location to the specific anchor number according to the photovoltaic array topology, and trigger graded early warning according to the damage degree.

[0037] This step inputs the feature vectors collected on-site into the trained diagnostic model, inverts and outputs the damage location and damage degree, and triggers a graded early warning based on the damage degree. As one implementation method, step S4 includes: S41: The on-site feature vector constructed in step S3 Input the CNN-LSTM diagnostic model trained in step S2 to obtain the damage parameters, the expression of which is: ; in, For the predicted location of the damage, To predict the extent of damage, For the trained CNN-LSTM diagnostic model; In this step, the model automatically inverts damage parameters from multidimensional stress wave characteristics without the need for manual interpretation; the end-to-end mapping relationship enables automated processing from raw signals to diagnostic results.

[0038] S42: The damage location output in step S41 Mapping to specific anchor bolt numbers yields spatial positioning information containing both the anchor bolt number and the damage depth. This step maps one-dimensional damage depth coordinates to anchor bolt numbers in the actual project, providing maintenance personnel with intuitive and operable positioning information.

[0039] S43: Based on the degree of damage Triggering a tiered warning: When A green alert is triggered and logged; when A yellow alert is triggered, prompting on-site verification and handling to be completed within one month; when A red alert is triggered, prompting an immediate shutdown and reinforcement.

[0040] This step-by-step tiered early warning mechanism directly links the degree of damage with the response measures, avoiding a one-size-fits-all alarm strategy; and when linked with the tracking and control system, it can achieve automatic shutdown protection.

[0041] Furthermore, in step S11, considering the spatial variability of soil and rock parameters, the soil and rock parameters are regarded as spatial random fields, and Gaussian autocorrelation function is used to describe their spatial correlation. Latin hypercube sampling is used to randomly select N sets of random parameter combinations as the parameter input of the simulation model.

[0042] Please refer to Figure 2 The CNN-LSTM deep learning model in step S2 consists of two one-dimensional convolutional layers, one LSTM layer, one dropout layer, and two fully connected layers. CNN excels at extracting local spatial features (wave velocity changes between adjacent measurement points), while LSTM excels at extracting temporally dependent features (time series of stress wave propagation). The combination of the two can comprehensively characterize the propagation pattern of stress waves in the anchor bolt. Traditional pure data-driven models only minimize... It is possible to learn statistically reasonable but physically incorrect mappings (such as wave speed increasing as damage decreases).

[0043] Preferably, the training parameters in step S2 are as follows: using the Adam optimizer, with an initial learning rate of 0.001 and a batch size of 32; training terminates when the validation set loss no longer decreases after 20 consecutive epochs, and the model parameters with the minimum validation set loss are saved. The Adam optimizer is used with an initial learning rate of 0.001 and a batch size of 32 to minimize the loss function. The network is trained to target specific objectives. The dataset is divided into training, validation, and test sets in a 7:1.5:1.5 ratio. Training is terminated when the validation set loss no longer decreases after 20 consecutive epochs, and the model parameters with the minimum validation set loss are saved. The Adam optimizer adaptively adjusts the learning rate to accelerate convergence; an early stopping mechanism prevents overfitting and ensures the model has good generalization ability.

[0044] Please refer to Figure 3The present invention also provides an intelligent slope damage diagnosis system for photovoltaic construction, comprising: Data acquisition module 1 is used to establish a three-dimensional numerical simulation model of the propagation of coupled stress waves between photovoltaic structure, anchor rod and soil and rock based on the spatiotemporal variability of multi-source parameters of the slope photovoltaic anchoring system, and to obtain simulation datasets of the anchoring system under different damage locations and damage degrees. Model training module 2 is used to input the simulation dataset into the CNN-LSTM deep learning model for training. During the training process, a loss function with physical constraints is defined: ; ; in, For the total loss function, For data fitting terms, This is a physical consistency penalty term based on the one-dimensional wave equation. For balance coefficient, The total number of training samples, Let i be the displacement field of the i-th training sample. Using time as the coordinate, Let be the spatial coordinates along the anchor bolt axis. Let be the propagation velocity of the stress wave in the anchor bolt. The damping ratio of the soil and rock mass. L2 norm operator; According to the calibration data acquisition module 3, it is used to deploy active excitation devices and sensor arrays in the slope photovoltaic anchoring system, perform frequency sweep excitation on the anchor rod under test, and calibrate the optimal excitation frequency; apply stress wave excitation according to the calibrated optimal excitation frequency, and simultaneously acquire wave response data, environmental compensation data and structural status data; Output early warning module 4 is used to input the collected fluctuation response data, environmental compensation data and structural state data into the trained diagnostic model, and inversely output the damage location and damage degree; according to the photovoltaic array topology, the damage location is mapped to the specific anchor number, and a graded early warning is triggered according to the damage degree.

[0045] For a detailed implementation of the intelligent slope damage diagnosis system for photovoltaic construction, please refer to the above-mentioned intelligent slope damage diagnosis method for photovoltaic construction, which will not be elaborated further here.

[0046] Please refer to Figure 4 The present invention also provides an electronic device, comprising: Memory 6 is used to store one or more programs; Processor 5; Processor 5 and memory 6 are connected via communication interface 7; When one or more programs are executed by processor 5, all or some of the methods described above are implemented.

[0047] Among them, processor 5 can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0048] The memory 6 may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0049] It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from the spirit or essential characteristics of this application. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of this application is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within this application. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for intelligent diagnosis of slope damage used in photovoltaic construction, characterized in that, Includes the following steps: S1: Based on the spatiotemporal variability of multi-source parameters of the slope photovoltaic anchoring system, a three-dimensional photovoltaic structure-anchor-soil coupled stress wave propagation numerical simulation model is established to obtain simulation datasets of the anchoring system under different damage locations and damage degrees. S2: Input the simulation dataset into the CNN-LSTM deep learning model for training. During training, define a loss function that includes physical constraints: ； ； in, For the total loss function, For data fitting terms, This is a physical consistency penalty term based on the one-dimensional wave equation. For balance coefficient, The total number of training samples, Let i be the displacement field of the i-th training sample. Using time as the coordinate, Let be the spatial coordinates along the anchor bolt axis. Let be the propagation velocity of the stress wave in the anchor bolt. The damping ratio of the soil and rock mass. L2 norm operator; S3: Deploy active excitation devices and sensor arrays in the slope photovoltaic anchoring system, perform frequency sweep excitation on the anchor rod under test, and calibrate the optimal excitation frequency; apply stress wave excitation according to the calibrated optimal excitation frequency, and simultaneously collect wave response data, environmental compensation data, and structural status data; S4: Input the collected fluctuation response data, environmental compensation data, and structural state data into the trained diagnostic model, and output the damage location and damage degree; map the damage location to the specific anchor number according to the photovoltaic array topology, and trigger graded early warning according to the damage degree.

2. The intelligent diagnosis method for slope damage in photovoltaic construction according to claim 1, characterized in that, Step S1 includes: S11: Establish a three-dimensional numerical simulation model of stress wave propagation coupled with photovoltaic structure-anchor bolt-soil, and determine the input parameter vector of the simulation model. Its expression is: ； in, The length of the anchor bolt. The diameter of the anchor rod. The elastic modulus of the anchor bolt. For anchor bolt density, The longitudinal wave velocity of the soil and rock mass. The transverse wave velocity of the rock and soil mass. The damping ratio of the soil and rock mass. For the photovoltaic structure location parameters, For the excitation center frequency, For temperature, The water content of the soil and rock mass. Location of injury The degree of damage; S12: Apply transient excitation to the end of the model anchor bolt, calculate the propagation response of the stress wave in the structure, and extract stress wave characteristics, including wave velocity. Energy attenuation coefficient and main frequency offset Its expression is: ； ； ； in, Let j be the axial coordinate of the j-th measuring point. Let j be the time when the first wave of the stress wave arrives at the j-th measuring point. The peak acceleration at the j-th measuring point is... Main frequency, The center frequency of the excitation signal; S13: Combine the stress wave characteristics obtained in step S12 with the environmental parameters in the simulation input to obtain the stress wave feature vector. Its expression is: ； S14: Location of damage and degree of damage As the output label, it is compared with the stress wave feature vector obtained in step S13. Pairing samples together forms a training sample, and all simulation samples constitute the simulation dataset. Its expression is: 。 3. The intelligent diagnosis method for slope damage in photovoltaic construction according to claim 1, characterized in that, The CNN-LSTM deep learning model in step S2 is used to establish a nonlinear mapping relationship from stress wave response signal features to damage parameters, and its expression is: ； in, For the input feature vector, The mapping function is parameterized by the CNN-LSTM hybrid neural network. These are the learnable parameters of the network. Location of injury The degree of damage.

4. The intelligent diagnosis method for slope damage in photovoltaic construction according to claim 1, characterized in that, Step S3 includes: S31: Install an active excitation device and a broadband response sensor array at the exposed end of the anchor bolt to be tested in the photovoltaic array on the slope; S32: Perform frequency sweep excitation on the anchor rod under test, collect the response signal at each frequency, and calculate the coherence coefficient. and signal-to-noise ratio Select the optimal frequency The diagnostic excitation frequency of this anchor bolt is expressed as follows: ； in, , Weighting coefficient; coherence coefficient The expression is: ； in, To determine the power spectrum of the excitation signal, In response to the power spectrum of the signal, The cross-power spectrum of the excitation signal and the response signal; S33: According to the calibrated optimal frequency Stress wave excitation is applied, and wave response data, environmental compensation data, and structural state data are collected simultaneously. S34: Extract features from the collected fluctuation response data to construct the model input vector. Its expression is: ； in, For the number of measurement points, For the wave velocity of each measurement segment, The energy attenuation coefficient for each measurement segment is... Main frequency offset For temperature, This represents the water content of the soil and rock mass.

5. The intelligent diagnosis method for slope damage in photovoltaic construction according to claim 1, characterized in that, Step S4 includes: S41: The on-site feature vector constructed in step S3 Input the CNN-LSTM diagnostic model trained in step S2 to obtain the damage parameters, the expression of which is: ； in, For the predicted location of the damage, To predict the extent of damage, For the trained CNN-LSTM diagnostic model; S42: The damage location output in step S41 Mapping to a specific anchor bolt number yields spatial location information containing the anchor bolt number and damage depth; S43: Based on the degree of damage Triggering a tiered warning: When A green alert is triggered and logged; when A yellow alert is triggered, prompting on-site verification and handling to be completed within one month; when A red alert is triggered, prompting an immediate shutdown and reinforcement.

6. The intelligent diagnosis method for slope damage in photovoltaic construction according to claim 2, characterized in that, In step S11, considering the spatial variability of soil and rock parameters, the relevant parameters of soil and rock are regarded as a spatial random field. The Gaussian autocorrelation function is used to describe its spatial correlation, and N sets of random parameter combinations are randomly selected by Latin hypercube sampling as the parameter input of the simulation model.

7. The intelligent diagnosis method for slope damage used in photovoltaic construction according to claim 1, characterized in that, The CNN-LSTM deep learning model in step S2 consists of two one-dimensional convolutional layers, one LSTM layer, one dropout layer, and two fully connected layers.

8. The intelligent diagnosis method for slope damage used in photovoltaic construction according to claim 1, characterized in that, The training parameters in step S2 are as follows: the Adam optimizer is used, the initial learning rate is set to 0.001, and the batch size is 32; training is terminated when the validation set loss no longer decreases for 20 consecutive rounds, and the model parameters with the minimum validation set loss are saved.

9. A smart slope damage diagnosis system for photovoltaic construction, characterized in that, include: The data acquisition module is used to establish a three-dimensional numerical simulation model of the propagation of coupled stress waves between the photovoltaic structure, anchor rod, and soil and rock based on the spatiotemporal variability of the multi-source parameters of the slope photovoltaic anchoring system, and to obtain simulation datasets of the anchoring system under different damage locations and damage degrees. The model training module is used to input simulation datasets into the CNN-LSTM deep learning model for training. During training, a loss function with physical constraints is defined. ； ； in, For the total loss function, For data fitting terms, This is a physical consistency penalty term based on the one-dimensional wave equation. Here, N is the balance coefficient, and N is the total number of training samples. Let i be the displacement field of the i-th training sample. Using time as the coordinate, Let be the spatial coordinates along the anchor bolt axis. Let be the propagation velocity of the stress wave in the anchor bolt. The damping ratio of the soil and rock mass. L2 norm operator; The data acquisition module is calibrated and used to deploy active excitation devices and sensor arrays in the slope photovoltaic anchoring system, perform frequency sweep excitation on the anchor rod under test, and calibrate the optimal excitation frequency; apply stress wave excitation according to the calibrated optimal excitation frequency, and simultaneously acquire wave response data, environmental compensation data, and structural status data. The output early warning module is used to input the collected fluctuation response data, environmental compensation data, and structural state data into the trained diagnostic model, and inversely output the damage location and damage degree; according to the photovoltaic array topology, the damage location is mapped to the specific anchor number, and a graded early warning is triggered according to the damage degree.

10. An electronic device, characterized in that, include: Memory, used to store one or more programs; processor; When the one or more programs are executed by the processor, the method as described in any one of claims 1-8 is implemented.

Images