Power station steel structure installation and health monitoring early warning method and system based on AI
By using AI-driven CNN-LSTM models and digital twin technology, combined with multi-source sensor monitoring, high-precision installation and full life-cycle health monitoring of heavy steel structures in power plants have been achieved. This solves the shortcomings of existing technologies in terms of installation accuracy and efficiency, monitoring data acquisition, and fault diagnosis, and enables efficient and accurate fault early warning and adaptive management.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- POWERCHINA SEPCO1 ELECTRIC POWER CONSTR CO LTD
- Filing Date
- 2025-12-16
- Publication Date
- 2026-04-24
AI Technical Summary
Existing technologies are insufficient to achieve micron-level installation precision control, accurate assessment of health status throughout the entire life cycle, and early warning of faults for heavy steel structures in power plants. They suffer from limitations in installation accuracy and efficiency, limitations in monitoring data acquisition, lag in fault diagnosis, and insufficient adaptability.
The system employs an AI-based CNN-LSTM hybrid neural network model to fuse basic static and dynamic real-time data, combined with the visualization guidance of a digital twin, to achieve intelligent installation and dynamic monitoring. It acquires full-dimensional monitoring data through multi-source sensors and utilizes deep learning algorithms for fault diagnosis and early warning.
It has achieved high-precision installation of heavy steel structures for power plants, shortened the construction period by 20-30%, reduced the incidence of safety accidents, improved the accuracy of fault identification to 98.5%, reduced unplanned downtime by more than 60%, and improved adaptability to different structural forms and working conditions.
Smart Images
Figure CN121920192A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the interdisciplinary field of artificial intelligence technology and heavy steel structure engineering for power plants, and in particular to an AI-based method and system for the installation and health monitoring and early warning of power plant steel structures. Background Technology
[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.
[0003] As the core load-bearing components of various power plant units, including ultra-high-parameter thermal power plants and nuclear power plants, the installation accuracy and long-term operational stability of heavy steel structures directly determine the safe operation level and service life of the power plant. With the development of power plant units towards ultra-high parameters and larger scale, stringent requirements have been placed on the micron-level precision control of steel structure installation, accurate assessment of health status throughout the entire life cycle, and early warning of faults. Against this backdrop, intelligent solutions integrating advanced technologies such as artificial intelligence, digital twins, and multi-source data sensing have become a key direction for solving the challenges of installation and health monitoring of heavy steel structures in power plants.
[0004] In existing technologies, the installation of heavy steel structures in power plants largely relies on manual experience to determine construction procedures. Key parameters such as crane operating radius and component installation logic are determined through subjective judgment. Installation accuracy measurements mainly use traditional equipment such as levels and theodolites. Health monitoring primarily relies on single-type sensors to collect local data, data transmission depends on wired networks, and fault diagnosis is mostly based on fixed threshold judgments. Furthermore, the installation process and monitoring system are independent of each other. Specifically, during the installation phase, component lifting, positioning, and connection are completed through a combination of manual planning and traditional measuring tools. During the operation phase, strain or vibration data is acquired through a single sensor, and after simple processing, faults are judged based on preset thresholds, forming a "separate installation-monitoring" management model.
[0005] In summary, existing technologies are ill-suited to the high demands of modern power plant units and present numerous problems: Firstly, installation accuracy and efficiency are limited. Manual planning can easily lead to unreasonable installation sequence and project delays. Traditional measurement is affected by environmental interference and operational errors, with verticality and horizontality measurement errors reaching ±5mm or more, which cannot meet the micron-level installation requirements. Moreover, the incidence of safety accidents in high-altitude operations is relatively high. Industry statistics show that the accident rate of traditional installation mode is 3-5 times higher than that of intelligent mode.
[0006] Secondly, the monitoring data collection has limitations. A single sensor type cannot fully reflect the true state of the steel structure under the coupled effects of temperature, load, and vibration. Fixed acquisition frequency can easily lead to the omission of key features or data redundancy. Wired transmission is prone to signal interruption in the high temperature, high humidity, and strong electromagnetic interference environment of power plants.
[0007] Third, fault diagnosis and early warning are lagging behind. Manual inspections are time-consuming and it is difficult to detect potential hazards such as early fatigue cracks and micro-deformations. Relying solely on threshold judgments cannot identify complex fault modes and lacks the ability to predict fault evolution, thus missing the best maintenance opportunity and increasing maintenance costs by more than 40%.
[0008] Fourth, there is a lack of adaptability and coordination. Traditional solutions are mostly customized for specific units and cannot be adapted to power plant steel structures with different structural forms and operating conditions. Furthermore, errors during the installation phase cannot be fed back to the operation monitoring system for threshold calibration, leading to a decrease in the accuracy of health assessments. These problems seriously affect the installation quality and operational safety of heavy steel structures in power plants, and restrict the construction and development of ultra-high parameter power plant units. Summary of the Invention
[0009] To address the aforementioned issues, this invention proposes an AI-based method and system for power plant steel structure installation and health monitoring and early warning. This invention presents a comprehensive solution integrating artificial intelligence, digital twins, and multi-source sensor networks to achieve closed-loop management of heavy steel structures in power plants, encompassing "intelligent installation - dynamic monitoring - precise diagnosis - tiered early warning," adaptable to different generating units, structural forms, and environmental conditions.
[0010] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides an AI-based method for the installation and health monitoring and early warning of power plant steel structures, including: Acquire basic static data and dynamic real-time data, input them into a pre-trained CNN-LSTM hybrid neural network model, and output the optimal installation scheme; A digital twin of the power plant steel structure is constructed, and the optimal installation scheme is mapped to the digital twin of the power plant steel structure. Based on the visual guidance of the digital twin, the installation of the actual power plant steel structure is started, the measurement data during the installation process is obtained, and the installation parameters are adjusted based on the preset deviation threshold. Based on the actual power plant steel structure after installation, monitoring data during the operation phase is acquired, and feature parameters are extracted after preprocessing. The extracted feature parameters are then input into a pre-trained fault diagnosis model, and the output includes fault diagnosis results including whether a fault exists, fault type, fault location, and severity.
[0011] Further technical solutions include: synchronizing measurement data during installation to monitoring during operation, calibrating the characteristic parameter judgment benchmark for fault diagnosis during operation, and feeding back the fault diagnosis results to the installation process.
[0012] Further technical solutions include the following: the basic static data includes design drawings, material parameters, terrain features, and historical installation cases; the dynamic real-time data includes wind speed, temperature, real-time performance parameters of the crane, and component transportation progress; the measurement data includes the spatial position, angle, and stress of the components; and the monitoring data includes strain data, vibration data, temperature data, and deformation data.
[0013] A further technical solution involves acquiring measurement data during the installation process and then performing measurement data preprocessing. This preprocessing includes removing environmental noise using Kalman filtering and identifying outliers using the isolated forest algorithm.
[0014] A further technical solution, the specific method for adjusting installation parameters based on a preset deviation threshold, is as follows: When the spatial position deviates from the standard parameters by more than the preset position deviation threshold, the model predictive control algorithm is used to adjust the lifting speed and amplitude of the crane; when the angle is greater than the preset angle deviation threshold, the posture of the steel structure is adjusted through closed-loop feedback; when the stress is greater than the preset stress threshold, the operation is automatically suspended, a stress release plan is pushed, and the operation is resumed after the stress drops to a safe range.
[0015] A further technical solution is that the CNN-LSTM hybrid neural network model includes a CNN module and an LSTM module. The CNN module is used to extract spatial structural features from the design drawings, including component dimensions, connection relationships, and stress distribution areas. The LSTM module is used to learn the temporal logic in historical installation cases, including lifting sequence, work intervals, and environmental adaptability adjustment rules.
[0016] A further technical solution is that the preprocessing includes data cleaning, data normalization, and elimination of noise and inconsistencies in the monitoring data; the feature parameters extracted after preprocessing include stress amplitude, frequency of change, vibration frequency, amplitude, and phase.
[0017] Secondly, this invention provides an AI-based power plant steel structure installation and health monitoring and early warning system, comprising the following modules: The installation planning module is configured to: acquire basic static data and dynamic real-time data, input them into a pre-trained CNN-LSTM hybrid neural network model, and output the optimal installation plan; The digital twin collaborative installation module is configured to: construct a digital twin of the power plant steel structure, map the optimal installation scheme to the digital twin of the power plant steel structure, start the installation of the actual power plant steel structure based on the visual guidance of the digital twin, acquire measurement data during the installation process, and adjust the installation parameters based on the preset deviation threshold. The fault diagnosis module is configured to: acquire monitoring data during the operation phase based on the actual power plant steel structure after installation, extract feature parameters after preprocessing, input the extracted feature parameters into a pre-trained fault diagnosis model, and output diagnostic results including whether a fault exists, fault type, fault location, and severity.
[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention uses a CNN-LSTM hybrid neural network model to fuse basic static data and dynamic real-time data to generate the optimal installation scheme. Combined with the visualization guidance of a digital twin and real-time measurement and parameter adjustment during the installation process, it eliminates reliance on traditional manual experience, effectively avoids delays caused by unreasonable installation sequence, and allows for precise control of installation deviations. This shortens the installation period by 20-30%, with installation deviations controlled within ±0.5mm, achieving 10 times the accuracy of traditional manual installation. The digital twin and VR visualization reduce the learning cost for construction personnel, increase work efficiency by 40%, meet the high-precision installation requirements of ultra-high parameter power plants, and reduce the risks of manual operation at heights, thus reducing the incidence of safety accidents.
[0019] This invention is based on the deployment of a multi-source monitoring mechanism on the actual power plant steel structure after installation. The acquired monitoring data during the operation phase covers the full-dimensional status of the steel structure. Compared with the local data acquisition of traditional single sensors, it can more comprehensively reflect the real status of the steel structure under complex working conditions, providing sufficient data support for subsequent fault diagnosis and solving the problem of limitations in traditional monitoring data acquisition.
[0020] This invention extracts feature parameters from the monitoring data during the operation phase and inputs them into a pre-trained fault diagnosis model. It can quickly output diagnostic results such as whether a fault exists, its type, location, and severity. This overcomes the limitations of traditional manual inspections, which are characterized by long cycles and delayed fault detection. It also avoids the problem of being unable to identify complex fault modes by relying solely on fixed thresholds. This enables accurate identification of early potential hazards, with a fault identification accuracy rate of ≥98.5% and a fault evolution prediction error of ≤10%. It achieves a shift from "post-event alarm" to "pre-event prevention," reducing unplanned downtime by more than 60%.
[0021] The installation process and operation monitoring of this invention are closely integrated. The measurement data during the installation process provides an adaptation basis for the monitoring during the operation phase. The actual monitoring and diagnosis results during the operation phase can be used to optimize subsequent installation plans. Compared with traditional solutions that are customized for specific units and where installation and monitoring are disconnected, this invention significantly improves the adaptability of the technical solution to power plant steel structures with different structural forms and different operating conditions, while also improving the overall accuracy of health assessment and fault diagnosis. Attached Figure Description
[0022] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.
[0023] Figure 1 This is a schematic diagram of the power plant steel structure installation planning process of the present invention; Figure 2 This is a diagram of the power plant steel structure operation monitoring architecture of the present invention. Detailed Implementation The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0024] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0025] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.
[0026] Example 1 This embodiment provides an AI-based method for the installation and health monitoring and early warning of power plant steel structures. Specifically, it involves intelligent installation planning, full life-cycle health status assessment, and intelligent fault diagnosis and hierarchical early warning model construction for power plant steel structures, integrating multi-source data perception, digital twin mapping, and deep learning algorithms. This method is applicable to the installation and long-term operation monitoring of heavy steel structures in various power plant units, such as ultra-high-parameter thermal power plants and nuclear power plants. The specific method steps are as follows: S1: Acquire basic static data and dynamic real-time data, input them into a pre-trained CNN-LSTM hybrid neural network model, and output the optimal installation scheme.
[0027] In step S1, as Figure 1 As shown, a CNN-LSTM hybrid neural network model is introduced to perform in-depth fusion analysis on the basic static data of the heavy steel structure of the power station.
[0028] In this embodiment, the basic static data includes design drawings, material parameters, terrain features, and historical installation cases; the dynamic real-time data includes wind speed, temperature, real-time performance parameters of the crane, and component transportation progress.
[0029] The CNN-LSTM hybrid neural network model includes a CNN module and an LSTM module. The CNN module is used to extract spatial structural features from the design drawings, including component dimensions, connection relationships, and stress distribution areas. The LSTM module is used to learn the temporal logic in historical installation cases, including lifting sequence, work intervals, and environmental adaptation adjustment rules.
[0030] The CNN-LSTM hybrid neural network model incorporates dynamic constraints: it receives real-time performance parameters of the crane, on-site environmental monitoring data (wind speed, temperature), and component transportation progress, and generates the optimal installation scheme through more than 500 iterations.
[0031] Optimal installation scheme output accuracy: component lifting time error ≤ 10 minutes, installation position deviation prediction ≤ ±0.3mm, connection sequence logic accuracy 100%.
[0032] S2: Construct a digital twin of the power station steel structure, map the optimal installation scheme into the digital twin of the power station steel structure, start the installation of the actual power station steel structure based on the visual guidance of the digital twin, obtain measurement data during the installation process, and adjust the installation parameters based on the preset deviation threshold.
[0033] In step S2, a digital twin of the power station steel structure is constructed, mapping it 1:1 to the physical entity. The optimal installation scheme is mapped onto the digital twin, recreating the entire installation process at a 1:1 scale. This allows construction personnel to immerse themselves in the installation path of components, the crane's movement trajectory, and the safe area for high-altitude operations using VR devices. Simultaneously, multi-terminal collaboration is supported, with construction guidance documents being pushed to a mobile app, on-site display screen, and crane operator's cab. The documents include key parameters such as component weight, lifting angle, and connecting bolt torque, and support offline viewing and real-time updates. Furthermore, the digital twin of the power station steel structure continuously verifies potential collision risks along the installation path, optimizing the work plan in advance.
[0034] In step S2, the measurement data during the installation process includes the spatial position, angle, and stress of the components. In this embodiment, a multi-dimensional measurement network consisting of lidar, inertial measurement unit (IMU), and intelligent strain sensors is deployed to achieve real-time sensing and closed-loop control of the installation process. Specifically: LiDAR accurately acquires the spatial position of components, inertial measurement unit (IMU) captures changes in angle and angular velocity in real time, and intelligent strain sensor is used to acquire stress changes during installation.
[0035] In step S2, after acquiring the measurement data during the installation process, the measurement data is preprocessed. The measurement data preprocessing includes removing environmental noise through Kalman filtering and identifying outliers using the isolated forest algorithm.
[0036] The specific method for adjusting installation parameters based on a preset deviation threshold is as follows: Position deviation: When the spatial position (measured position) deviates from the standard parameter by more than the preset position deviation threshold (±0.5mm), control commands are sent to the crane PLC via industrial Ethernet. The model predictive control algorithm is used to adjust the crane's lifting speed and amplitude to achieve micron-level positioning.
[0037] Angle deviation: When the angle (verticality / horizontality) is greater than the preset angle deviation threshold (±0.2°), the command is sent to the servo control system, and the steel structure posture is adjusted through closed-loop feedback. The adjustment response time is ≤0.5 seconds.
[0038] Stress deviation: When the stress exceeds the preset stress threshold (80% of the material's yield strength) during installation, the operation will be automatically paused, a stress release plan will be pushed, and the operation will resume after the stress drops to a safe range.
[0039] S3: Based on the actual power plant steel structure after installation, acquire monitoring data during the operation phase, extract feature parameters after preprocessing, and input the extracted feature parameters into the pre-trained fault diagnosis model to output fault diagnosis results including whether a fault exists, fault type, fault location, and severity.
[0040] In step S3, the monitoring data during the operation phase includes strain data, vibration data, temperature data, and deformation data. To obtain the monitoring data during the operation phase, in this embodiment, multiple types of sensors are arranged on the heavy steel structure of the power station, including strain gauge sensors, acceleration sensors, temperature sensors, and displacement sensors. Specifically: like Figure 2 As shown, strain gauge sensors are placed at stress concentration points, such as near welds and joints, to monitor strain changes in the steel structure; accelerometers are placed in the main supporting structures and vibration-prone areas of the steel structure to monitor vibration; temperature sensors are evenly distributed on the surface of the steel structure to measure the effect of ambient temperature on the performance of the steel structure; and displacement sensors are used to monitor the overall deformation of the steel structure.
[0041] In this embodiment, based on the operational characteristics and monitoring requirements of the power station's steel structure, data acquisition frequencies are set for different locations, different stress environments, and different sensors. Specifically: For strain gauge and accelerometer sensors, a sampling frequency of 100 times per second is set to promptly capture the response changes of the steel structure under dynamic loads. For temperature and displacement sensors, the sampling frequency is relatively low, set to once every 3 minutes, and the peak value of the change is monitored through the change curve. A dual-mode transmission system of fiber optic network + 5G is adopted, with fiber optics used for stable transmission of massive amounts of data and 5G serving as a backup link. Data transmission uses AES-256 encryption, and access permissions employ a three-level authentication method involving administrators, maintenance personnel, and visitors to prevent data leakage and tampering.
[0042] In step S3, after acquiring the monitoring data during the operation phase, preprocessing is required before feature parameters are extracted. Preprocessing includes data cleaning, data normalization, and intelligent analysis to eliminate noise and inconsistencies in the monitoring data. Then, signal processing and machine learning algorithms are used to extract feature parameters from the preprocessed monitoring data. Specifically, features such as stress amplitude and frequency of change are extracted from strain data; and features such as vibration frequency, amplitude, and phase are extracted from vibration data.
[0043] In step S3, a fault diagnosis model is constructed based on a convolutional neural network (CNN) or recurrent neural network (RNN) using deep learning. The model is trained using a large amount of historical monitoring data and known fault case data, and the model parameters are optimized and adjusted to enable the model to accurately identify different fault modes. By self-learning from a large amount of fatigue crack fault strain data and vibration data, the characteristic change patterns of the data when fatigue cracks occur are mastered.
[0044] First, real-time monitoring data is collected and feature parameters are extracted, then transmitted to the fault diagnosis model. The model outputs fault diagnosis results including whether a fault exists, the type of fault, the location of the fault, and its severity (quickly analyzing and determining whether a fault exists in the steel structure and the type, location, and severity of the fault). The system immediately issues warnings to maintenance personnel via SMS, voice alarms, etc., and displays the specific location and type of the fault, along with corresponding handling suggestions, at the monitoring center.
[0045] S4: Synchronize the measurement data during the installation process to the operation phase monitoring, calibrate the characteristic parameter judgment benchmark for fault diagnosis during the operation phase, and feed back the fault diagnosis results to the installation process.
[0046] In step S4, new operational data and fault cases are automatically collected every quarter to incrementally train the installation planning model and fault diagnosis model, update network parameters, and make the model adapt to the aging characteristics and operating condition changes of the power station steel structure, thereby improving the diagnostic accuracy by ≥1% annually.
[0047] This embodiment constructs an intelligent management system for the entire life cycle of heavy steel structures in power plants through the deep integration of artificial intelligence, digital twins, and multi-source sensing. It effectively solves the pain points of traditional technologies such as low precision, slow response, and poor adaptability, and provides core technical support for the safe and stable operation of ultra-high parameter power plant units.
[0048] Example 2 This embodiment provides an AI-based power plant steel structure installation and health monitoring and early warning system, including the following modules: The installation planning module is configured to: acquire basic static data and dynamic real-time data, input them into a pre-trained CNN-LSTM hybrid neural network model, and output the optimal installation plan; The digital twin collaborative installation module is configured to: construct a digital twin of the power plant steel structure, map the optimal installation scheme to the digital twin of the power plant steel structure, start the installation of the actual power plant steel structure based on the visual guidance of the digital twin, acquire measurement data during the installation process, and adjust the installation parameters based on the preset deviation threshold. The fault diagnosis module is configured to: acquire monitoring data during the operation phase based on the actual power plant steel structure after installation, extract feature parameters after preprocessing, input the extracted feature parameters into a pre-trained fault diagnosis model, and output diagnostic results including whether a fault exists, fault type, fault location, and severity.
[0049] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
[0050] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.
Claims
1. An AI-based method for power plant steel structure installation and health monitoring and early warning, characterized in that, include: Acquire basic static data and dynamic real-time data, input them into a pre-trained CNN-LSTM hybrid neural network model, and output the optimal installation scheme; A digital twin of the power plant steel structure is constructed, and the optimal installation scheme is mapped to the digital twin of the power plant steel structure. Based on the visual guidance of the digital twin, the installation of the actual power plant steel structure is started, the measurement data during the installation process is obtained, and the installation parameters are adjusted based on the preset deviation threshold. Based on the actual power plant steel structure after installation, monitoring data during the operation phase is acquired, and feature parameters are extracted after preprocessing. The extracted feature parameters are then input into a pre-trained fault diagnosis model, and the output includes fault diagnosis results including whether a fault exists, fault type, fault location, and severity.
2. The AI-based method for power plant steel structure installation and health monitoring and early warning as described in claim 1, characterized in that, Also includes: The measurement data during the installation process is synchronized to the monitoring during the operation phase, the characteristic parameter judgment benchmark for fault diagnosis during the operation phase is calibrated, and the fault diagnosis results are fed back to the installation process.
3. The AI-based method for power plant steel structure installation and health monitoring and early warning as described in claim 1, characterized in that, The basic static data includes design drawings, material parameters, terrain features, and historical installation cases; the dynamic real-time data includes wind speed, temperature, real-time performance parameters of the crane, and component transportation progress; the measurement data includes the spatial position, angle, and stress of the components; and the monitoring data includes strain data, vibration data, temperature data, and deformation data.
4. The AI-based method for power plant steel structure installation and health monitoring and early warning as described in claim 1, characterized in that, After acquiring the measurement data during the installation process, the measurement data is preprocessed. The preprocessing includes removing environmental noise using Kalman filtering and identifying outliers using the isolated forest algorithm.
5. The AI-based method for power plant steel structure installation and health monitoring and early warning as described in claim 1, characterized in that, The specific method for adjusting installation parameters based on a preset deviation threshold is as follows: When the spatial position deviates from the standard parameters by more than the preset position deviation threshold, the model predictive control algorithm is used to adjust the lifting speed and amplitude of the crane; when the angle is greater than the preset angle deviation threshold, the attitude of the steel structure is adjusted through closed-loop feedback. When the stress exceeds the preset stress threshold, the operation is automatically paused, a stress release plan is pushed out, and the operation resumes after the stress drops to a safe range.
6. The AI-based method for power plant steel structure installation and health monitoring and early warning as described in claim 1, characterized in that, The CNN-LSTM hybrid neural network model includes a CNN module and an LSTM module. The CNN module is used to extract spatial structural features from the design drawings, including component dimensions, connection relationships, and stress distribution areas. The LSTM module is used to learn the temporal logic in historical installation cases, including lifting sequence, work intervals, and environmental adaptability adjustment rules.
7. The AI-based method for power plant steel structure installation and health monitoring and early warning as described in claim 1, characterized in that, The preprocessing includes data cleaning, data normalization, and elimination of noise and inconsistencies in the monitoring data; the feature parameters extracted after preprocessing include stress amplitude, frequency of change, vibration frequency, amplitude, and phase.
8. An AI-based power plant steel structure installation and health monitoring and early warning system, characterized in that, Includes the following modules: The installation planning module is configured to: acquire basic static data and dynamic real-time data, input them into a pre-trained CNN-LSTM hybrid neural network model, and output the optimal installation plan; The digital twin collaborative installation module is configured to: construct a digital twin of the power plant steel structure, map the optimal installation scheme to the digital twin of the power plant steel structure, start the installation of the actual power plant steel structure based on the visual guidance of the digital twin, acquire measurement data during the installation process, and adjust the installation parameters based on the preset deviation threshold. The fault diagnosis module is configured to: acquire monitoring data during the operation phase based on the actual power plant steel structure after installation, extract feature parameters after preprocessing, input the extracted feature parameters into a pre-trained fault diagnosis model, and output diagnostic results including whether a fault exists, fault type, fault location, and severity.
9. A computer-readable storage medium having a program stored thereon, characterized in that, When executed by the processor, the program implements the steps in the AI-based power plant steel structure installation and health monitoring and early warning method as described in any one of claims 1-7.
10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the AI-based power plant steel structure installation and health monitoring and early warning method as described in any one of claims 1-7.