Three-dimensional visualization briefing technology method based on power transmission and transformation engineering construction

Abstract

The embodiment of the application provides a three-dimensional visualization disclosure technical method based on power transmission and transformation engineering construction, and relates to the technical field of power engineering construction. The method comprises the following steps: a digital twin containing material reference waveguide characteristic data is established based on design drawings; defect diagnosis is performed at different stages of the whole construction process, and tower material micro-cracks, insulator string delamination defects and fitting internal damage are identified respectively; multi-source diagnosis results are fused and processed to generate a four-dimensional space-time defect probability distribution; material disposal instructions, construction guidance instructions and maintenance warning instructions are generated based on the distribution through a decision engine; a defect evolution model is established to predict defect expansion trends; feedback data is used to optimize operation safety thresholds and update material reference waveguide characteristic data, forming a self-adaptive optimization closed loop. The application realizes accurate diagnosis, forward-looking prediction and adaptive control of the whole construction process quality, and effectively improves the construction quality and safety level.

Description

Technical Field

[0001] This application relates to the field of power engineering construction technology, and in particular to a three-dimensional visualization disclosure technology method based on power transmission and transformation engineering construction. Background Technology

[0002] As a core component of national energy infrastructure construction, the construction quality of power transmission and transformation projects directly affects the long-term safe and stable operation of the power grid. Traditional construction technical briefings rely heavily on two-dimensional drawings, written descriptions, and on-site verbal guidance, which suffers from problems such as abstract information, large misunderstandings, and unintuitive quality control points. With the development of digital technology, three-dimensional visualization technology has been introduced into the construction briefing process, which improves the intuitiveness and accuracy of the briefing to a certain extent by constructing three-dimensional models.

[0003] Existing technologies primarily focus on the visualization of geometric models, which are static digital models that cannot be correlated with the real-time physical state of materials and components during construction. For hidden defects, they lack the ability to diagnose and warn at each stage of construction. Defects are often discovered late, and may not be exposed until the final acceptance or operation phase, resulting in high rework costs and safety hazards. Existing detection methods are usually carried out independently in a single stage, and the detection data are fragmented and fail to be deeply integrated with the quality control chain of the entire construction process. They cannot grasp the distribution, evolution, and impact on subsequent construction stages from a spatiotemporal perspective.

[0004] Therefore, there is a need for a three-dimensional visualization disclosure technology that can deeply integrate perception, diagnosis, prediction and decision-making functions and realize the self-evolution of digital twins, so as to achieve accurate, forward-looking and adaptive control of the quality of the entire process of power transmission and transformation engineering construction. Summary of the Invention

[0005] The embodiments of this application provide a three-dimensional visualization-based technical method for construction of power transmission and transformation projects, enabling accurate diagnosis, forward-looking prediction, and adaptive control of construction quality. To achieve the above objectives, this application adopts the following technical solution:

[0006] A three-dimensional visualization-based technical method for handover of power transmission and transformation engineering construction, the method comprising:

[0007] Based on the tower material, insulator model, and hardware design drawings, a digital twin including material reference waveguide characteristic data is established;

[0008] Based on the digital twin, defect diagnosis is performed at different stages of the entire construction process;

[0009] During the tower erection phase, micro-cracks in the tower material are identified by micro-impact excitation and comparison with the reference waveguide characteristic data of the material, generating the first diagnostic result.

[0010] During the conductor deployment stage, by sweeping frequency vibration and comparing the material reference waveguide characteristic data, the delamination defects of the insulator string are identified, and a second diagnostic result is generated.

[0011] During the equipment installation phase, internal damage to the fittings is identified by vibration monitoring and comparison with the material reference waveguide characteristic data, generating a third diagnostic result.

[0012] The first, second, and third diagnostic results are fused together to generate a four-dimensional spatiotemporal defect probability distribution that characterizes the spatial location, occurrence probability, and evolution over time of the defect.

[0013] The four-dimensional spatiotemporal defect probability distribution is used as the input of the decision engine. The spatial location, occurrence probability and evolution over time are mapped and matched to the corresponding links in the construction process to generate instructions that are adapted to the quality control requirements of different construction stages. The instructions include material disposal instructions for guiding material replacement or repair, construction guidance instructions for adjusting construction technology or sequence, and maintenance early warning instructions for risk warning.

[0014] A defect evolution model is established based on the four-dimensional spatiotemporal defect probability distribution to predict the defect expansion trend;

[0015] Execute the material handling instructions, construction guidance instructions, and maintenance early warning instructions, collect the generated execution feedback data, and use the feedback data to optimize the operational safety thresholds in the digital twin;

[0016] Based on the optimized operational safety threshold and defect expansion trend, a pre-maintenance plan is generated, the pre-maintenance plan is executed and the actual execution effect data is collected, and the results of the strategy adjustment taken in response to the maintenance warning are recorded;

[0017] The actual implementation data of the pre-maintenance scheme and the adjustment results of the early warning strategy are fed back to the digital twin to update the material reference waveguide characteristic data.

[0018] Specifically, the four-dimensional spatiotemporal defect probability distribution is used as input to the decision engine. Spatial location, occurrence probability, and evolution over time are mapped and matched to corresponding stages of the construction process, generating instructions adapted to the quality control needs of different construction stages. These instructions include material handling instructions for guiding material replacement or repair, construction guidance instructions for adjusting construction techniques or sequences, and maintenance early warning instructions for risk alerts.

[0019] During the component arrival stage, the defect probability in the four-dimensional spatiotemporal defect probability distribution is evaluated and compared with the defect tolerance threshold of the incoming material. When the defect probability value exceeds the tolerance threshold, a material handling instruction to guide material replacement or repair is generated.

[0020] During the installation and construction phase, the spatial coordinate information in the four-dimensional spatiotemporal defect probability distribution and the risk level represented by the defect probability value are integrated with the physical entity status data. When a high-risk area is identified, construction guidance instructions for adjusting the construction process or sequence are generated.

[0021] During the final acceptance phase, the time evolution data and defect probability values ​​in the four-dimensional spatiotemporal defect probability distribution are comprehensively analyzed and compared with the operational safety threshold. When the predicted defect development trend exceeds the safety threshold, a maintenance warning instruction for risk alert is generated.

[0022] Specifically, based on the digital twin, defect diagnosis is performed at different stages of the entire construction process, including:

[0023] During the tower erection phase, the stress wave propagation characteristics in the tower material are analyzed and compared with the propagation characteristics in the reference waveguide characteristic data of the material. When the difference between the stress wave propagation characteristics and the corresponding reference data in the reference waveguide characteristic data of the material exceeds a preset threshold, it is determined that there is a microcrack defect and a first diagnostic result is generated.

[0024] During the conductor deployment stage, the vibration response spectrum characteristics of the insulator string are analyzed and the matching degree is calculated with the vibration spectrum characteristics in the material reference waveguide characteristic data. When the matching degree is lower than the preset matching threshold, it is determined that there is a delamination defect and a second diagnostic result is generated.

[0025] During the equipment installation phase, the vibration signal characteristics of the fittings are analyzed, and the deviation from the normal operation characteristics in the material reference waveguide characteristic data is analyzed. When the deviation exceeds the normal range, it is determined that there is internal damage, and a third diagnostic result is generated.

[0026] Specifically, the process of fusing the first, second, and third diagnostic results to generate a four-dimensional spatiotemporal defect probability distribution characterizing the spatial location, occurrence probability, and evolution over time includes:

[0027] Defect diagnosis results from different construction stages are spatiotemporally aligned, and a multi-scale signal decomposition algorithm is used to extract the energy distribution, spectral characteristics and temporal statistics of various types of defect signals after alignment, forming a feature vector group;

[0028] The feature vector group is input into a deep belief network for feature fusion and dimensionality reduction to obtain a comprehensive defect feature vector.

[0029] Based on the comprehensive defect feature vector, the probability of defect occurrence at spatial points is calculated using a probability density estimation model to form a three-dimensional defect probability distribution.

[0030] By introducing the time dimension and combining it with the state transition model of defect evolution, the trend of defect probability change over time is predicted, and a four-dimensional spatiotemporal defect probability distribution including spatial coordinates, defect probability, and time evolution data is constructed.

[0031] Specifically, the generation of construction guidance instructions for adjusting construction techniques or sequences includes:

[0032] Based on the spatial coordinates and defect probabilities in the four-dimensional spatiotemporal defect probability distribution, a density clustering algorithm is used to identify the boundaries of high-risk areas and generate a three-dimensional risk restricted area model.

[0033] The three-dimensional risk restricted area model is compared with the device model in the digital twin using Boolean operations to obtain the passable space;

[0034] A path optimization algorithm is used to generate a safety installation path in the passable space;

[0035] The safety installation path is discretized into a sequence of path points, and pose adjustment commands and motion control parameters are generated for each path point.

[0036] The posture adjustment command and motion control parameters are transmitted to the installation equipment control system through an industrial network to form construction guidance commands.

[0037] Specifically, establishing a defect evolution model based on the four-dimensional spatiotemporal defect probability distribution to predict defect expansion trends includes:

[0038] Based on the four-dimensional spatiotemporal defect probability distribution, the defect probability value of each spatial point at consecutive time points is obtained to form a defect probability time series.

[0039] A multivariate defect evolution model is established using a time-series predictive neural network. The various input features of the time-series predictive neural network include defect probability time series, environmental parameters, and equipment operating status parameters.

[0040] By weighted fusing various input features of the temporal prediction neural network through an attention mechanism, a comprehensive feature representation is obtained;

[0041] Based on the comprehensive feature representation, the time-series prediction neural network trained end-to-end is used to predict the trend of defect probability changes in future stages.

[0042] The predicted results of the defect probability change trend are compared with the material aging model data in the digital twin;

[0043] Based on the physical constraints provided by the material aging model, the prediction results are corrected to obtain the defect probability prediction after physical model correction.

[0044] Based on the defect probability prediction corrected by the physical model, the remaining service life of the component is calculated as the final defect propagation trend prediction.

[0045] Specifically, the process of predicting defect propagation trends, executing material handling instructions, construction guidance instructions, and maintenance early warning instructions, collecting generated execution feedback data, and using the feedback data to optimize operational safety thresholds in the digital twin includes:

[0046] Based on the prediction results of the defect evolution model, a stochastic simulation method is used to simulate the defect development path under different combinations of environmental parameters and equipment operating state parameters to obtain the probability distribution of defects exceeding the critical state.

[0047] Based on the probability distribution, a predefined risk rule base is queried, which defines the risk level and adjustment strategy corresponding to different probability intervals;

[0048] Execute adjustment strategies matched from the risk rule base, including raising or lowering the operational security threshold;

[0049] The adjusted operational safety thresholds are associated with equipment identification, component location, and service time and stored in the security rule base of the digital twin, and then updated to the alarm configuration of the field monitoring system.

[0050] Specifically, the process of feeding back the actual implementation effect data of the pre-maintenance scheme and the adjustment results of the early warning strategy to the digital twin to update the material reference waveguide characteristic data includes:

[0051] A material property update framework based on a generative model is established, and the monitored defect evolution data is encoded as latent variables;

[0052] The latent variables are fused with the material reference waveguide property data stored in the digital twin, and the updated material property parameters are obtained through decoding and reconstruction.

[0053] The confidence intervals of the updated material property parameters are calculated using statistical inference methods, and the update is accepted when the confidence intervals are stable.

[0054] The updated material reference waveguide property data is stored in a digital twin for subsequent defect diagnosis and analysis.

[0055] Specifically, the generation of maintenance warning instructions for risk alerts includes:

[0056] Based on the time evolution data of the four-dimensional spatiotemporal defect probability distribution, the urgency index of the defect location is calculated.

[0057] Assess the complexity of maintenance operations based on the spatial location and type of the defect;

[0058] The maintenance priority score is calculated by combining the defect probability value, the urgency index, and the maintenance operation complexity.

[0059] Maintenance levels are divided based on the maintenance priority score, and a corresponding response time window and resource configuration scheme are defined for each level;

[0060] Maintenance warning instructions are generated based on the maintenance level and its corresponding response time window and resource configuration scheme. The maintenance warning instructions include defect coordinates, type, maintenance level, suggested maintenance period and required resource information.

[0061] Specifically, after executing the material handling instructions, construction guidance instructions, and maintenance early warning instructions, the process also includes:

[0062] Collect and maintain data;

[0063] The maintenance data is compared with the predicted data of the defect evolution model to calculate the model deviation;

[0064] The defect evolution model is updated based on the incremental learning algorithm and the model bias.

[0065] Analyze the correlation between maintenance effectiveness and quality control during the construction phase, and establish a quality traceability chain;

[0066] The diagnostic thresholds and control strategies for the construction phase are optimized based on the aforementioned quality traceability chain.

[0067] As can be seen from the above technical solution, this application has the following beneficial effects:

[0068] 1. This method constructs a digital twin that integrates the physical properties of materials, and then implements targeted defect diagnosis at key stages of the entire construction process. This enables accurate identification and early detection of hidden defects such as microcracks in tower materials, delamination of insulators, and internal damage to hardware. By fusing multi-source and heterogeneous diagnostic results to generate a four-dimensional spatiotemporal defect probability distribution, the originally isolated and static detection data is transformed into a comprehensive model that dynamically reflects the spatial distribution, probability of occurrence, and evolution of defects over time. Based on this probability distribution-driven intelligent decision engine, material handling, construction guidance, and maintenance early warning instructions are generated that are highly matched with the quality control needs of each stage of construction. This improves the foresight and accuracy of construction quality control and effectively avoids rework costs and safety risks caused by delayed defect detection.

[0069] 2. This method constructs an adaptive optimization closed loop that spans both the digital twin and the physical entity. By utilizing feedback data after instruction execution, it continuously optimizes the operational safety threshold and material baseline characteristic data in the digital twin, enabling the system to possess self-learning and continuous evolution capabilities. The defect evolution model couples data-driven prediction with physical law constraints, predicting defect expansion trends and component remaining lifespan, providing a scientific basis for generating forward-looking pre-maintenance schemes. Ultimately, it achieves adaptive closed-loop control and continuous improvement of the quality throughout the entire construction process of power transmission and transformation projects, providing a solid technical guarantee for building safe and reliable power grid infrastructure. Attached Figure Description

[0070] The invention will now be further described with reference to the accompanying drawings.

[0071] Figure 1 A first flowchart provided for an embodiment of this application;

[0072] Figure 2 A second flowchart provided for embodiments of this application;

[0073] Figure 3 The third flowchart provided for an embodiment of this application. Detailed Implementation

[0074] The terms "first," "second," and "third," etc., used in this application specification, claims, and drawings are for distinguishing different objects, not for specifying a particular order.

[0075] In the embodiments of this application, the words "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0076] Research has revealed that existing technologies lack the ability to diagnose and provide early warnings at each stage of construction. Defects are often discovered late, and may not be exposed until the final acceptance or operation phase, resulting in high rework costs and safety hazards. Existing detection methods are usually carried out independently in a single stage, and the detection data are fragmented and fail to be deeply integrated with the quality control chain of the entire construction process. It is impossible to grasp the distribution, evolution, and impact on subsequent construction stages of defects from a spatiotemporal perspective.

[0077] To address the aforementioned issues, this application provides a three-dimensional visualization-based technical method for the disclosure of information during power transmission and transformation engineering construction.

[0078] Example 1: To solve the above problems, such as Figure 1As shown, this embodiment describes the complete technical process from the creation of a digital twin to the execution of instructions.

[0079] Step 1: Create a digital twin with physical characteristics

[0080] The establishment of a digital twin is the foundation of the entire technology system, and it is constructed from three levels: geometric modeling, physical property attribution, and data association.

[0081] Geometric modeling implementation process:

[0082] Using 3D modeling software, a parametric 3D model with millimeter-level precision was created based on the tower structure drawings, insulator assembly drawings, and hardware connection details provided by the design institute. The tower model adopted a hierarchical structural modeling method, building it step by step from the tower base and tower body to the tower head, ensuring that each component has an independent geometric identifier; the insulator string model was built according to the actual string structure, including components such as ball joint hangers, insulator discs, and cup-head hangers; the hardware model focused on constructing the fine geometric features of the contact surfaces and connection parts.

[0083] Material reference waveguide property data acquisition:

[0084] Tower material reference data acquisition: Sample specimens were extracted from the same batch of tower materials, and full waveform acquisition was performed using a digital ultrasonic flaw detector; by changing the probe angle and excitation frequency, the velocity matrix, attenuation coefficient tensor and dispersion curve clusters of stress waves under different propagation paths were obtained; full matrix capture technology was used to record all possible probe combination data to establish a complete waveguide characteristic reference database.

[0085] Insulator reference data acquisition: Electromagnetic vibration tables were used to perform sweep-frequency vibration tests on insulator specimens, covering their operating frequency band. Multiple vibration response data were collected using an array of accelerometers placed at various locations on the insulator. A modal parameter identification algorithm was employed to extract parameters such as natural frequency, damping ratio, mode shape, and modal stiffness from the transfer function. Reference vibration characteristics under different operating conditions were obtained by varying the excitation amplitude and ambient temperature.

[0086] Hardware reference data acquisition: Cyclic loading tests are performed on hardware specimens on a fatigue testing machine to simulate actual operating conditions; vibration signals are collected by an embedded micro vibration sensor, and acoustic emission sensors are used to monitor the evolution of internal damage; a signal decomposition algorithm is used to decompose the vibration signal into components with different physical meanings to establish a vibration characteristic template under normal operating conditions.

[0087] Digital twin data integration:

[0088] Develop data management middleware to associate 3D models with physical property data. Adopt an object-oriented database design, establishing an attribute table for each model component to store corresponding reference waveguide property data; establish a spatial indexing mechanism to support fast location-based queries of material properties in specific regions; and develop a data interface module to provide unified data access services for subsequent defect diagnosis.

[0089] Step 2: Multi-stage defect diagnosis throughout the construction process

[0090] Deploy corresponding testing systems at each stage of construction to achieve full-process quality monitoring from material arrival to installation completion.

[0091] Microcrack diagnosis during tower assembly:

[0092] Portable automatic impact devices and sensor networks are deployed at the tower erection site. The impact device uses electromagnetic excitation to generate controllable micro-impact force, and the force value is adaptively adjusted according to the thickness and material of the tower. The sensor network consists of multiple piezoelectric accelerometers, which are installed at key locations on the surface of the tower according to an optimized layout.

[0093] Diagnostic algorithm implementation: The acquired stress wave signals are first preprocessed, including noise reduction filtering, trend term elimination, and signal alignment. Then, the signal's characteristic parameters are extracted, including first wave arrival time, first wave amplitude, dominant frequency offset, waveform similarity index, and energy attenuation coefficient. An anomaly detection model based on Mahalanobis distance is developed to calculate the statistical distance between the measured feature vector and the reference feature vector. A dynamic threshold mechanism is set; when the distance value exceeds an adaptive threshold, a microcrack defect is determined to exist. The defect location is accurately calculated using the time difference localization method, and the severity of the defect is assessed based on the distance value, generating a first diagnostic result containing three-dimensional coordinates, defect type, and severity level.

[0094] Diagnosis of insulator delamination defects during conductor laying stage:

[0095] A hydraulic vibration excitation system is integrated into the traction machine and tensioner, which can apply controllable frequency sweep vibration to the insulator string during conductor deployment; the vibration frequency changes according to a logarithmic law, covering the main modal frequency range of the insulator; a high-precision inertial measurement unit is used to collect the vibration response of the insulator string, and the measurement unit transmits the data to the analysis terminal wirelessly.

[0096] Diagnostic algorithm implementation: Spectral analysis is performed on the acquired vibration signals to calculate the power spectral density function and frequency response function; the correlation between the measured vibration mode and the reference vibration mode is calculated using the modal confidence criterion, and the frequency consistency is evaluated using the modal assurance criterion; an anomaly detection algorithm based on support vector data description is developed to construct a hypersphere enclosing normal samples in the feature space; when a new vibration feature is located outside the hypersphere, a layering defect is determined; the layering location is determined through mode shape change analysis, the layering area is evaluated through the degree of frequency shift, and a second diagnostic result is generated.

[0097] Diagnosis of internal damage to fittings during equipment installation:

[0098] Miniature vibration sensors are pre-embedded inside the fittings. The sensors are powered by energy harvesting technology and can work for a long time without the need to replace batteries. Vibration signals are collected in real time during installation and transmitted to the monitoring center via a wireless sensor network.

[0099] Diagnostic algorithm implementation: An adaptive signal decomposition algorithm is used to decompose the vibration signal into multiple intrinsic mode functions, and time-domain and frequency-domain features are extracted from each mode function; an anomaly detection model based on a deep autoencoder is developed, which learns the reconstruction of normal vibration signals through the encoding and decoding process; the reconstruction error is calculated as an anomaly indicator, and internal damage is determined when the error exceeds a threshold; an attention mechanism is used to identify the feature dimension most sensitive to anomalies, and the damage type is determined through feature contribution analysis to generate a third diagnostic result.

[0100] Step 3: Generation of four-dimensional spatiotemporal defect probability distribution

[0101] By fusing multi-source diagnostic results into a unified probability distribution model, comprehensive defect status information is provided for decision-making.

[0102] Spatiotemporal registration of multi-source data:

[0103] A unified spatiotemporal coordinate system was established, and a feature-point-based registration method was used to align diagnostic results from different stages. Temporal registration employed interpolation to unify asynchronously acquired data to the same timestamp. Spatial registration calculated coordinate transformation parameters by identifying common feature points to ensure all data were in the same three-dimensional coordinate system.

[0104] Multi-scale feature extraction and fusion:

[0105] A multi-scale signal analysis algorithm was developed to perform wavelet packet decomposition on various defect signals, extracting energy, entropy, and statistical features from different frequency bands. A high-dimensional feature vector space was constructed, and principal component analysis was used for dimensionality reduction, preserving feature components in the main variation directions. A deep belief network was designed for feature-level fusion. The network was constructed using a multi-layer restricted Boltzmann machine, unsupervised pre-training was performed using the contrastive divergence algorithm, and then supervised fine-tuning was performed using the backpropagation algorithm. Finally, a low-dimensional comprehensive defect feature vector was output, which both preserved the original information and eliminated redundancy.

[0106] Defect probability field modeling:

[0107] Based on the comprehensive defect feature vector, a nonparametric kernel density estimation method is used to calculate the probability of defect occurrence at each point in three-dimensional space. An appropriate Gaussian kernel function is selected, and the optimal bandwidth parameter is determined through cross-validation. A mathematical expression for the probability field is established, interpolating the discrete detection point probabilities into a continuous probability distribution. Moving least squares is used for surface fitting to ensure the smoothness and accuracy of the probability field.

[0108] Time dimension expansion and evolution prediction:

[0109] A time series analysis method is introduced to establish a defect state transition model. A Hidden Markov Model (HMM) is used to describe the defect evolution process, and the state transition probability is calculated using a forward-backward algorithm. Combining a material fatigue model and environmental influencing factors, the changing trend of defect probability over time is predicted. Finally, a four-dimensional spatiotemporal defect probability distribution is constructed, where three dimensions represent spatial location and one dimension represents time. The value at each spatiotemporal point represents the probability of defect occurrence at that location at that moment.

[0110] Step 4: Intelligent Decision Engine and Instruction Generation

[0111] Based on the four-dimensional spatiotemporal defect probability distribution, an intelligent decision engine was developed to generate targeted control commands.

[0112] Material disposal instruction generation:

[0113] During the component arrival phase, a material acceptance model based on Bayesian decision theory is established. The prior probability comes from historical quality data, and the likelihood function is based on the current defect probability distribution. The posterior probability is calculated and compared with a preset defect tolerance threshold. When the posterior probability exceeds the threshold, a material disposal instruction is generated. The instruction includes specific disposal measures, such as replacement or repair, execution time limit, and responsible personnel, and the disposal process is automatically triggered through the material management system.

[0114] Construction guidance instruction generation:

[0115] During the installation and construction phase, a path planning algorithm based on the potential field method was developed. High-risk areas were modeled as repulsive potential fields, with the defect probability value determining the potential field strength. The target location was modeled as an attractive potential field, guiding the equipment movement through the resultant force direction. The optimal path was calculated using a numerical integration method to ensure that the path was smooth and met the equipment motion constraints. The continuous path was discretized into a sequence of path points, generating detailed pose commands and motion parameters for each path point. The commands were transmitted to the equipment control system in real time via industrial Ethernet to guide the installation operation to avoid risky areas.

[0116] Maintenance warning command generation:

[0117] During the final acceptance phase, a multi-indicator comprehensive evaluation system is established. The defect probability value reflects the current risk level, the probability gradient reflects the rate of risk change, and the defect type affects the urgency of maintenance. The weight of each indicator is determined using the analytic hierarchy process (AHP), and a comprehensive risk score is calculated through weighted summation. Risk levels are classified according to the score, and corresponding response strategies are defined for each level. Maintenance early warning instructions include specific maintenance suggestions, time windows, and resource requirements, and are pushed to relevant responsible persons through the project management platform.

[0118] Step 5: Defect Evolution Modeling and Trend Prediction

[0119] Establish a defect evolution prediction model that combines data-driven approaches with physical models to enable proactive maintenance decisions.

[0120] Construction of a time-series prediction neural network:

[0121] Design a time-series prediction model based on a long short-term memory network. The network input includes a historical defect probability sequence, an environmental parameter sequence, and a device state sequence. The network adopts an encoder-decoder architecture. The encoder learns the implicit features of the input sequence, and the decoder generates future predictions based on these features. An attention mechanism is added between the encoder and the decoder so that the model can focus on the historical moments that are most important for the prediction.

[0122] Physical model coupling correction:

[0123] A crack propagation model based on fracture mechanics is established to describe the relationship between stress intensity factor and crack growth rate; a fatigue life model based on cumulative damage theory is established to consider load sequence effects; the neural network prediction results are fused with the physical model predictions, and the Kalman filter algorithm is used to achieve data assimilation; the defect evolution process is described through a state-space model, and the model parameters are continuously corrected using observation data to improve prediction accuracy.

[0124] Remaining useful life assessment:

[0125] Based on the corrected defect evolution trajectory, the remaining service life of the component is calculated using the first-crossing probability theory. A critical state for the defect is defined; the component is considered to have failed when the defect parameters exceed this state. Different combinations of loads and environmental conditions are simulated using Monte Carlo simulations to statistically analyze the distribution of failure times and calculate the confidence intervals for reliability and remaining service life.

[0126] Step 6: System closed-loop optimization and digital twin update

[0127] Establish a self-learning mechanism based on feedback data to achieve continuous system improvement.

[0128] Runtime security threshold optimization:

[0129] Collect feedback data after command execution, including the actual development of defects, maintenance effects, and cost data; establish a multi-objective optimization model that considers safety, economy, and construction progress; use a genetic algorithm to solve for the Pareto optimal solution set and select the most suitable threshold combination for the current engineering conditions; update the safety rule base in the digital twin to achieve threshold adjustment.

[0130] Material property data update:

[0131] Develop a material property update framework based on generative adversarial networks; the generator learns the mapping from latent variables to material properties, and the discriminator distinguishes between real properties and generated properties; new detection data is encoded as latent variables, and the updated material properties are obtained through the generator; an adversarial training method is adopted to ensure that the updated properties are both consistent with the new data and maintain physical rationality; the significance of the update is verified by hypothesis testing, and only updates that pass the test will be formally adopted.

[0132] Quality control strategy optimization:

[0133] Analyze historical quality data to identify key control points and sensitive parameters in the construction process; establish a quality traceability chain to track the causes and development of defects; and optimize construction process parameters and quality control standards based on the traceability results to form a continuous improvement quality management system.

[0134] Example 2: As Figure 2 As shown, specifically: This embodiment describes a lightweight implementation scheme based on an edge computing architecture for scenarios with high real-time requirements.

[0135] Step 1: Deployment of Distributed Edge Nodes

[0136] Edge computing nodes are deployed at key locations on the construction site, with each node equipped with dedicated computing hardware and communication modules.

[0137] Edge node hardware configuration:

[0138] It adopts an industrial-grade embedded system as the edge computing platform, equipped with a multi-core processor, large-capacity memory, and solid-state storage; it integrates multiple industrial communication interfaces, supporting Ethernet, wireless LAN, and fieldbus. A protective enclosure is installed to meet the waterproof, dustproof, and vibration-resistant requirements of the construction site.

[0139] Software architecture design:

[0140] Develop a lightweight operating system, removing unnecessary system services and optimizing real-time performance; design modular application software, with each functional module running independently and communicating via message queues. Implement a dynamic load balancing mechanism to allocate computing resources based on task urgency.

[0141] Step 2: Lightweight Digital Twin Construction

[0142] To adapt to the computing and storage limitations of edge nodes, a simplified version of the digital twin was developed.

[0143] Data simplification strategy:

[0144] A feature selection algorithm is used to select the most discriminative feature subset from the complete reference waveguide characteristic data; a data compression algorithm is used to reduce storage space requirements while ensuring that key information is not lost; a hierarchical storage mechanism is established to store frequently used data locally and infrequently used data in the cloud.

[0145] Quickly query the index:

[0146] Spatial and attribute indexes are established for material property data to support fast location-based and attribute-based retrieval. A caching mechanism is used to store recently accessed data, reducing disk read / write operations; a data prefetching algorithm is developed to predict the next required data based on the construction progress and load it in advance.

[0147] Step 3: Optimization of Real-time Defect Diagnosis Algorithm

[0148] To optimize the diagnostic algorithm to meet real-time requirements, the computing power of edge nodes is considered.

[0149] Signal processing algorithm optimization:

[0150] A fast filtering algorithm was developed, employing a cascaded integrator-comb filter to achieve efficient low-pass filtering. Fourier transform calculations were optimized by using sparse Fourier transform techniques to reduce computational load; an incremental feature extraction algorithm was designed, calculating only the changed features when new data arrives, rather than recalculating the entire feature set.

[0151] Lightweighting machine learning models:

[0152] Model pruning techniques are employed to remove connections in the neural network that have little impact on the output. Weight quantization is used to convert floating-point weights into low-precision representations; knowledge distillation is applied to guide the training of smaller models with larger models, thereby improving the performance of the smaller models.

[0153] Real-time decision-making mechanism:

[0154] Establish a rule-based reasoning engine to encode expert knowledge into production rules. Develop a case-based reasoning system to assist decision-making by retrieving similar historical cases; implement streaming data processing, processing and generating output immediately upon data arrival, without waiting for the entire batch.

[0155] Step 4: Edge-Cloud Collaborative Computing

[0156] Rationally allocate computing tasks between the edge and the cloud to achieve efficient collaboration.

[0157] Task allocation strategy:

[0158] Assign tasks with high real-time requirements to edge nodes, such as signal preprocessing, simple feature extraction, and emergency decision-making; assign computationally intensive tasks to the cloud, such as model training, complex data analysis, and long-term prediction; and establish a task migration mechanism to transfer some tasks to the cloud when the edge nodes are overloaded.

[0159] Data synchronization mechanism:

[0160] Develop an incremental data synchronization algorithm that transmits only changed data instead of all data. Employ data compression and differential coding techniques to reduce the amount of data transmitted; implement a resume function to automatically save progress when the network is interrupted and resume transmission upon recovery.

[0161] Consistency maintenance:

[0162] Establish a data version management mechanism to track the data modification history; use optimistic locking to resolve concurrent access conflicts and automatically coordinate and resolve conflicts when they are detected; implement an eventual consistency model, which allows for short-term data inconsistency but will eventually reach a consistent state.

[0163] Step 5: Adaptive Control and Dynamic Optimization

[0164] The system parameters and control strategies are dynamically adjusted according to changes in the on-site conditions.

[0165] Parameter adaptive adjustment:

[0166] Establish a system performance monitoring module to track diagnostic accuracy, response time, and resource utilization in real time. Design a controller to automatically adjust algorithm parameters based on performance indicators, such as filter cutoff frequency, feature extraction window size, and decision threshold. Employ reinforcement learning algorithms to autonomously learn the optimal parameter configuration.

[0167] Dynamic resource allocation:

[0168] Monitor the load of each edge node, and migrate some of its tasks to nodes with lighter loads when a node is overloaded; allocate computing resources according to the urgency of tasks, with high-priority tasks receiving more resources; implement elastic scaling of computing resources, automatically allocating more resources when demand increases and releasing excess resources when demand decreases.

[0169] Network adaptive transmission:

[0170] Monitor network bandwidth and latency changes and adaptively adjust data transmission strategies. Transmit high-precision data when bandwidth is sufficient, and transmit compressed or feature-rich data when bandwidth is limited. Implement multi-path transmission and use both wired and wireless networks simultaneously to improve reliability.

[0171] Example 3: Figure 3 As shown, specifically, this embodiment focuses on how to integrate multiple sensor modal data to improve the accuracy of defect diagnosis and achieve more comprehensive construction quality control.

[0172] Step 1: Multimodal data acquisition and preprocessing

[0173] Deploy multiple types of sensors to monitor construction quality from different physical dimensions.

[0174] Visual data acquisition:

[0175] High-resolution industrial cameras are used to acquire images of component surfaces, with polarizing filters to eliminate glare interference; structured light 3D scanners are used to acquire 3D topographic data of component surfaces; and thermal imagers are used to monitor temperature distribution and identify abnormal hot areas.

[0176] Acoustic data acquisition:

[0177] A microphone array is deployed to collect ambient sound and beamforming technology is used to focus on specific sound sources; acoustic emission sensors are used to monitor damage signals inside the material, and the sampling frequency meets the requirements of high-frequency acoustic emission signals; an ultrasonic probe is installed to actively excite and receive ultrasonic signals for internal defect detection.

[0178] Vibration data acquisition:

[0179] High-sensitivity accelerometers are used to measure structural vibrations, with a frequency response range covering a wide frequency band from low to high frequencies; dynamic strain gauges are installed to measure the strain response of key parts, with a sampling frequency that meets the requirements for dynamic strain measurement; and laser vibrometers are used for non-contact vibration measurement, which is particularly suitable for rotating parts and high-temperature environments.

[0180] Data preprocessing and standardization:

[0181] Develop a multimodal data alignment algorithm to ensure that data from different sensors are synchronized in time and space; design a data standardization process to convert data with different physical dimensions into a unified numerical range; implement abnormal data detection and repair, and identify and handle abnormal values ​​caused by sensor failures.

[0182] Step 2: Multimodal Feature Extraction and Fusion

[0183] Complementary feature information is extracted from multi-source data, and a more comprehensive state representation is obtained through fusion.

[0184] Cross-modal feature learning:

[0185] Design a cross-modal neural network, which includes multiple dedicated subnetworks to process data from different modalities; add a cross-modal attention mechanism to the high-level layers of each subnetwork to learn the correlation between features of different modalities; and use a contrastive learning method to bring the features of different modalities of the same sample closer together and push the features of different samples further apart.

[0186] Multi-scale feature fusion:

[0187] In feature-level fusion, bilinear pooling is used to fuse features from different modalities while preserving the interaction information between modalities. In decision-level fusion, an ensemble learning algorithm is used to combine the classification results of each modality, and a final decision is generated through weighted voting. An adaptive fusion weight mechanism is designed to dynamically adjust the weight of each modality in the fusion process based on the quality of the data.

[0188] Feature selection and dimensionality reduction:

[0189] The maximum correlation and minimum redundancy criterion is used to select the most discriminative and complementary feature subsets; multimodal principal component analysis is used for feature dimensionality reduction to eliminate redundancy while retaining the main information; and a feature importance evaluation algorithm is developed to identify the feature combinations that contribute the most to defect diagnosis.

[0190] Step 3: Multimodal Defect Diagnosis and Localization

[0191] Accurate defect diagnosis and localization are achieved by utilizing the fused multimodal features.

[0192] Multimodal anomaly detection:

[0193] A multimodal deep autoencoder is established to learn the reconstruction of the normal state through the encoding and decoding process; the reconstruction error of each mode is calculated, and when multiple modes have high reconstruction errors at the same time, they are judged as defects; a multimodal isolation forest algorithm is used to identify outliers in the high-dimensional feature space.

[0194] Defect Classification and Identification:

[0195] Design a multimodal convolutional neural network to extract local features from different types of data such as images and waveforms; use graph neural networks to process non-Euclidean data, such as point cloud data and sensor network data; and develop a multi-task learning framework to simultaneously complete multiple related tasks such as defect detection, classification, and localization.

[0196] Precise defect location:

[0197] By combining visual positioning, sound source positioning, and vibration transmission path analysis, precise three-dimensional positioning of defects is achieved; the time difference positioning method is used to process propagating signals, and the defect location is calculated by the signal arrival time difference; beamforming technology is used to process array signals, and the signal in a specific direction is enhanced by phase adjustment.

[0198] Step 4: Multi-objective decision optimization

[0199] By comprehensively considering multiple optimization objectives, an optimal decision that balances the needs of all parties is generated.

[0200] Establishment of a multi-objective optimization model:

[0201] Define multiple optimization objectives, including maximizing safety, minimizing cost, minimizing project time, and optimizing quality; identify the trade-offs between the objectives, establish the objective function and constraints; and adopt the Pareto optimality concept to seek a non-dominated solution set rather than a single optimal solution.

[0202] Optimize the algorithm implementation:

[0203] Develop a multi-objective evolutionary algorithm to explore the solution space through selection, crossover, and mutation operations. Employ non-dominated sorting and crowding calculation to maintain population diversity.

[0204] Decision support and visualization:

[0205] Develop an interactive decision support system that visualizes the Pareto front and the performance of each solution. Provide a solution comparison function to compare the advantages and disadvantages of various candidate solutions from different dimensions; implement sensitivity analysis to assess the impact of changes in key parameters on the decision outcome.

[0206] Step 5: Full Lifecycle Quality Traceability

[0207] Establish a quality data chain covering the entire lifecycle of design, construction, operation and maintenance.

[0208] Quality data standardization:

[0209] Define a unified quality data model to ensure data consistency and interoperability across all stages. Develop quality data collection specifications, clarifying collection frequency, accuracy requirements, and storage formats; establish a quality data dictionary to standardize terminology and coding rules.

[0210] Blockchain Quality Profile:

[0211] Blockchain technology is used to establish tamper-proof quality archives, recording data throughout the entire process from material procurement to final acceptance; this enables reliable traceability of quality data, with any modifications leaving a permanent record. Smart contracts are established to automatically execute quality acceptance standards, reducing human intervention.

[0212] Big Data Analysis and Mining:

[0213] Association rule mining techniques were applied to discover implicit relationships between construction parameters and quality results. Cluster analysis was used to identify typical patterns of quality problems, providing a basis for targeted improvements; a predictive model was established to predict potential future quality problems based on historical data.

[0214] The foregoing has shown and described the basic principles, main features, and advantages of this application. Those skilled in the art should understand that this application is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of this application. Various changes and modifications can be made to this application without departing from the spirit and scope thereof, and all such changes and modifications fall within the scope of this application as claimed. The scope of protection of this application is defined by the appended claims and their equivalents.

Claims

1. A three-dimensional visualization-based technical method for the construction of power transmission and transformation projects, characterized in that: The method includes: Based on the tower material, insulator model, and hardware design drawings, a digital twin including material reference waveguide characteristic data is established; Based on the digital twin, defect diagnosis is performed at different stages of the entire construction process; During the tower erection phase, micro-cracks in the tower material are identified by micro-impact excitation and comparison with the reference waveguide characteristic data of the material, generating the first diagnostic result. During the conductor deployment stage, by sweeping frequency vibration and comparing the material reference waveguide characteristic data, the delamination defects of the insulator string are identified, and a second diagnostic result is generated. During the equipment installation phase, internal damage to the fittings is identified by vibration monitoring and comparison with the material reference waveguide characteristic data, generating a third diagnostic result. The first, second, and third diagnostic results are fused together to generate a four-dimensional spatiotemporal defect probability distribution that characterizes the spatial location, occurrence probability, and evolution over time of the defect. The four-dimensional spatiotemporal defect probability distribution is used as the input of the decision engine. The spatial location, occurrence probability and evolution over time are mapped and matched to the corresponding links in the construction process to generate instructions that are adapted to the quality control requirements of different construction stages. The instructions include material disposal instructions for guiding material replacement or repair, construction guidance instructions for adjusting construction technology or sequence, and maintenance early warning instructions for risk warning. A defect evolution model is established based on the four-dimensional spatiotemporal defect probability distribution to predict the defect expansion trend; Execute the material handling instructions, construction guidance instructions, and maintenance early warning instructions, collect the generated execution feedback data, and use the feedback data to optimize the operational safety thresholds in the digital twin; Based on the optimized operational safety threshold and defect expansion trend, a pre-maintenance plan is generated, the pre-maintenance plan is executed and the actual execution effect data is collected, and the results of strategy adjustments taken in response to maintenance warnings are recorded. The actual implementation data of the pre-maintenance scheme and the adjustment results of the early warning strategy are fed back to the digital twin to update the material reference waveguide characteristic data.

2. The method according to claim 1, characterized in that, The process uses the four-dimensional spatiotemporal defect probability distribution as input to the decision engine, mapping and matching spatial location, occurrence probability, and evolution over time to corresponding stages of the construction process. This generates instructions adapted to the quality control needs of different construction stages. These instructions include material handling instructions for guiding material replacement or repair, construction guidance instructions for adjusting construction techniques or sequences, and maintenance early warning instructions for risk alerts. During the component arrival stage, the defect probability in the four-dimensional spatiotemporal defect probability distribution is evaluated and compared with the defect tolerance threshold of the incoming material. When the defect probability value exceeds the tolerance threshold, a material handling instruction to guide material replacement or repair is generated. During the installation and construction phase, the spatial coordinate information in the four-dimensional spatiotemporal defect probability distribution and the risk level represented by the defect probability value are integrated with the physical entity status data. When a high-risk area is identified, construction guidance instructions for adjusting the construction process or sequence are generated. During the final acceptance phase, the time evolution data and defect probability values ​​in the four-dimensional spatiotemporal defect probability distribution are comprehensively analyzed and compared with the operational safety threshold. When the predicted defect development trend exceeds the safety threshold, a maintenance warning instruction for risk alert is generated.

3. The method according to claim 1, characterized in that, Based on the digital twin, defect diagnosis is performed at different stages of the entire construction process, including: During the tower erection phase, the stress wave propagation characteristics in the tower material are analyzed and compared with the propagation characteristics in the reference waveguide characteristic data of the material. When the difference between the stress wave propagation characteristics and the corresponding reference data in the reference waveguide characteristic data of the material exceeds a preset threshold, it is determined that there is a microcrack defect and a first diagnostic result is generated. During the conductor deployment stage, the vibration response spectrum characteristics of the insulator string are analyzed and the matching degree is calculated with the vibration spectrum characteristics in the material reference waveguide characteristic data. When the matching degree is lower than the preset matching threshold, it is determined that there is a delamination defect and a second diagnostic result is generated. During the equipment installation phase, the vibration signal characteristics of the fittings are analyzed, and the deviation from the normal operation characteristics in the material reference waveguide characteristic data is analyzed. When the deviation exceeds the normal range, it is determined that there is internal damage, and a third diagnostic result is generated.

4. The method according to claim 3, characterized in that, The process of fusing the first, second, and third diagnostic results to generate a four-dimensional spatiotemporal defect probability distribution characterizing the spatial location, occurrence probability, and evolution over time includes: Defect diagnosis results from different construction stages are spatiotemporally aligned, and a multi-scale signal decomposition algorithm is used to extract the energy distribution, spectral characteristics and temporal statistics of various defect signals after alignment, forming a feature vector group. The feature vector group is input into a deep belief network for feature fusion and dimensionality reduction to obtain a comprehensive defect feature vector. Based on the comprehensive defect feature vector, the probability of defect occurrence at spatial points is calculated using a probability density estimation model to form a three-dimensional defect probability distribution. By introducing the time dimension and combining it with the state transition model of defect evolution, the trend of defect probability change over time is predicted, and a four-dimensional spatiotemporal defect probability distribution including spatial coordinates, defect probability, and time evolution data is constructed.

5. The method according to claim 2, characterized in that, The generation of construction guidance instructions for adjusting construction techniques or sequences includes: Based on the spatial coordinates and defect probabilities in the four-dimensional spatiotemporal defect probability distribution, a density clustering algorithm is used to identify the boundaries of high-risk areas and generate a three-dimensional risk restricted area model. The three-dimensional risk restricted area model is compared with the device model in the digital twin using Boolean operations to obtain the passable space; A path optimization algorithm is used to generate a safety installation path in the passable space; The safety installation path is discretized into a sequence of path points, and pose adjustment commands and motion control parameters are generated for each path point. The posture adjustment command and motion control parameters are transmitted to the installation equipment control system through an industrial network to form construction guidance commands.

6. The method according to claim 1, characterized in that, The method of establishing a defect evolution model based on the four-dimensional spatiotemporal defect probability distribution to predict defect expansion trends includes: Based on the four-dimensional spatiotemporal defect probability distribution, the defect probability value of each spatial point at consecutive time points is obtained to form a defect probability time series. A multivariate defect evolution model is established using a time-series predictive neural network. The various input features of the time-series predictive neural network include defect probability time series, environmental parameters, and equipment operating status parameters. By weighted fusing various input features of the temporal prediction neural network through an attention mechanism, a comprehensive feature representation is obtained; Based on the comprehensive feature representation, the time-series prediction neural network trained end-to-end is used to predict the trend of defect probability changes in future stages. The predicted results of the defect probability change trend are compared with the material aging model data in the digital twin; Based on the physical constraints provided by the material aging model, the prediction results are corrected to obtain the defect probability prediction after physical model correction. Based on the defect probability prediction corrected by the physical model, the remaining service life of the component is calculated as the final defect propagation trend prediction.

7. The method according to claim 6, characterized in that, The process of predicting defect expansion trends, executing material handling instructions, construction guidance instructions, and maintenance early warning instructions, collecting generated execution feedback data, and using the feedback data to optimize operational safety thresholds in the digital twin includes: Based on the prediction results of the defect evolution model, a stochastic simulation method is used to simulate the defect development path under different combinations of environmental parameters and equipment operating state parameters to obtain the probability distribution of defects exceeding the critical state. Based on the probability distribution, a predefined risk rule base is queried, which defines the risk level and adjustment strategy corresponding to different probability intervals; Execute adjustment strategies matched from the risk rule base, including raising or lowering the operational security threshold; The adjusted operational safety thresholds are associated with equipment identification, component location, and service time and stored in the security rule base of the digital twin, and then updated to the alarm configuration of the field monitoring system.

8. The method according to claim 1, characterized in that, The process of feeding back the actual implementation effect data of the pre-maintenance scheme and the adjustment results of the early warning strategy to the digital twin to update the material reference waveguide characteristic data includes: A material property update framework based on a generative model is established, and the monitored defect evolution data is encoded as latent variables; The latent variables are fused with the material reference waveguide property data stored in the digital twin, and the updated material property parameters are obtained through decoding and reconstruction. The confidence intervals of the updated material property parameters are calculated using statistical inference methods, and the update is accepted when the confidence intervals are stable. The updated material reference waveguide property data is stored in a digital twin for subsequent defect diagnosis and analysis.

9. The method according to claim 2, characterized in that, The generation of maintenance warning instructions for risk alerts includes: Based on the time evolution data of the four-dimensional spatiotemporal defect probability distribution, the urgency index of the defect location is calculated. Assess the complexity of maintenance operations based on the spatial location and type of the defect; The maintenance priority score is calculated by combining the defect probability value, the urgency index, and the maintenance operation complexity. Maintenance levels are divided based on the maintenance priority score, and a corresponding response time window and resource configuration scheme are defined for each level; Maintenance warning instructions are generated based on the maintenance level and its corresponding response time window and resource configuration scheme. The maintenance warning instructions include defect coordinates, type, maintenance level, suggested maintenance period and required resource information.

10. The method according to claim 9, characterized in that, After executing the material handling instructions, construction guidance instructions, and maintenance early warning instructions, the method further includes: Collect and maintain data; The maintenance data is compared with the predicted data of the defect evolution model to calculate the model deviation; The defect evolution model is updated based on the incremental learning algorithm and the model bias. Analyze the correlation between maintenance effectiveness and quality control during the construction phase, and establish a quality traceability chain; The diagnostic thresholds and control strategies for the construction phase are optimized based on the aforementioned quality traceability chain.

Images