GIS equipment installation quality intelligent evaluation method and system

Abstract

The invention discloses a GIS equipment installation quality intelligent evaluation method and system, and the method comprises the steps: constructing a quality element list database according to GIS installation specifications and quality requirements, and enabling the database to serve as a target space for quality evaluation, so that various types of data collected subsequently can be mapped to corresponding quality inspection items, and the quality inspection items are matched with the quality inspection items. The system comprises a data acquisition layer, a data processing layer, an analysis decision-making layer and an application layer, the data acquisition layer acquires key data in a GIS installation process in real time through various sensors and equipment interfaces, and is also integrated with a programmable logic controller and a data interface of an intelligent tool; the data processing layer transmits real-time data streams by adopting an efficient distributed message queue, and the analysis decision-making layer fuses an artificial intelligence algorithm. According to the invention, multi-source sensing data, quality factor standards and an intelligent algorithm can be fused, and the whole-process, quantitative and traceable management of the installation environment, personnel operation, equipment attitude and key test results is realized.

Description

Technical Field

[0001] This invention belongs to the field of power high-voltage switchgear installation quality management technology, specifically referring to a method and system for intelligent assessment of GIS equipment installation quality. Background Technology

[0002] Current GIS (Gas Insulated Metal Enclosed Switchgear) installation quality management primarily relies on manual experience and post-installation inspections. This fragmented and subjective quality control process makes it difficult to achieve comprehensive, quantitative, and traceable management of the installation environment, personnel operations, equipment posture, and key test results. In particular, issues such as fluctuations in environmental temperature and humidity, cleanliness, installation alignment deviations, and personnel violations during installation are often difficult to detect in a timely manner, with quality risks only gradually emerging after commissioning. Existing installation assistance systems based on video surveillance and sensors largely remain at the monitoring and recording level, lacking a systematic quality evaluation model integrated with installation quality standards. A technical system has not yet been established to map multi-source data into unified quality indicators and to perform multi-timescale analysis and trend judgment of quality status. Therefore, there is an urgent need for an intelligent management and evaluation system for GIS equipment installation quality that can integrate multi-source sensing data, quality element standards, and intelligent algorithms. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for intelligent evaluation of the installation quality of GIS equipment. This method and system can integrate multi-source sensing data, quality element standards and intelligent algorithms to achieve full-process, quantitative and traceable management of the installation environment, personnel operation, equipment posture and key test results.

[0004] To achieve the above objectives, this invention provides an intelligent assessment method for the installation quality of GIS equipment. Based on GIS installation specifications and quality requirements, a quality element list database is constructed. This database serves as the target space for quality evaluation, enabling all subsequently collected data to be mapped to corresponding quality inspection items. The method includes the following steps: The first step will be within the time window. t The environmental features, posture features, behavioral features identified by the camera, and experimental features collected internally are mapped into a unified quality feature vector. The second step is to calculate the compliance of each quality inspection item based on the quality element list database after obtaining the quality feature vector. The third step is to weight and sum up the compliance of all inspection items corresponding to each installation process according to their weights to form a process-level quality score. The fourth step is to aggregate all process scores according to their importance weights to obtain the equipment-level quality score; The fifth step involves using an isolated forest unsupervised anomaly identification model to intelligently identify anomalies in the quality feature vectors. The isolated forest model outputs anomaly scores, and deduction values ​​are calculated based on the degree of anomaly to form a corrected process-level score. The sixth step involves constructing a multi-timescale quality trend model based on a sliding window for quality performance across different time periods.

[0005] As a further aspect of the present invention: in the first step, the quality feature vector Q t The mapping process is implemented using the following linear weighted model: ; in W E The environmental quality mapping weight matrix represents environmental characteristics. E t =[ T t , H t , P t ] T , W S The weight matrix maps device attitude to installation accuracy, and the attitude features are... , W A A weight matrix is ​​used to map personnel behavior to operational procedures, and behavioral characteristics are represented. , W X The weight matrix maps experimental and validation results, and the experimental features... X t =[ x t,1 , x t,2 ,… T ; The second step includes checking any item. i Its time window t The measured value is M i,t The standard value is S i The allowable deviation is Δ i Its quality compliance is defined as: ; If the deviation exceeds the allowable range, then directly set =0; In the third step, process-level quality scores are generated. The process-level scoring model is as follows: ; Step 4: Equipment-level quality assessment : ; in, α t The importance weights are derived from the process sensitivity ranking in the quality element database; Step 5: The Isolation Forest model outputs anomaly scores. A t ∈[0,1], deduction value D t = βA t The revised process-level scores are as follows: ; Step 6: Set the time window length to... h The quality trend is then defined as: ; By selecting different window lengths, quality trends at the process level, equipment level, and project level can be obtained respectively.

[0006] The above method is applied to an intelligent assessment system for GIS equipment installation quality. This system includes a data acquisition layer, a data processing layer, an analysis and decision-making layer, and an application layer. The data acquisition layer collects key data in real time during the GIS installation process through various sensors and equipment interfaces. This layer also integrates data interfaces for programmable logic controllers and intelligent tools. The data processing layer uses an efficient distributed message queue to transmit real-time data streams and utilizes a big data processing framework to perform data cleaning, noise reduction, and feature extraction, providing standardized input for subsequent intelligent analysis. The analysis and decision-making layer integrates artificial intelligence algorithms to perform in-depth analysis of the collected data, detects abnormal patterns through the isolated forest algorithm, predicts potential risks by combining historical failure cases, and constructs an installation quality evaluation model to generate dynamic scores. The application layer provides visual monitoring, early warning information push, and electronic report generation functions through a web or mobile interface, realizing remote monitoring and management of the GIS installation site.

[0007] As a further aspect of the present invention: the hardware of the system includes a high-definition camera, an environmental sensor module, an electronic theodolite, an intelligent control host, a human-computer interaction device, a voice warning unit, an alarm light, and a communication interface. The intelligent control host is connected to the high-definition camera, the environmental sensor module, the electronic theodolite, the human-computer interaction device, the voice warning unit, the alarm light, and the communication interface. The intelligent control host includes an AI processor, a neural network acceleration unit, and data storage. The communication interface includes a wired communication interface and a wireless communication interface.

[0008] As a further aspect of the present invention: the intelligent control host has a built-in target detection algorithm, which is based on an improved Yolov5 deep learning model and uses the GhostBottleneckCSP module to replace the original C3 module; the SENet attention mechanism is introduced to enhance the model's feature recognition ability for different targets and backgrounds; and the Mosaic data augmentation strategy is used to expand the dataset of targets related to abnormal behavior.

[0009] As a further aspect of the present invention: the intelligent control host also incorporates a personnel behavior recognition and posture analysis algorithm. First, the raw video data acquired by the high-definition camera is preprocessed, including fusing IMU sensor data using a Kalman filter algorithm to clean, denoise, and calibrate the motion trajectory signal. Then, a convolutional neural network is used to extract spatial features from the image sequence, and a long short-term memory network is combined to capture the pattern of motion changes over time, thereby identifying the operator's skeletal posture and continuous motion trajectory. Based on a pre-trained machine learning model, the extracted motion features are classified and analyzed to identify non-standard or dangerous operating behaviors.

[0010] Compared with existing technologies, the beneficial effects of this invention are as follows: First, this invention combines IoT sensing, high-definition video surveillance, and artificial intelligence algorithms to achieve real-time collection of the entire process of GIS equipment installation environment, construction behavior, and key process parameters. On this basis, a quality element list database is introduced to uniformly map the collected multi-source data to standardized quality inspection items. This ensures that the environmental conditions, operational behaviors, and equipment postures during the installation process are no longer merely monitored but directly participate in quality evaluation calculations, thereby significantly improving the intelligence level, real-time performance, and accuracy of installation quality control.

[0011] Secondly, this invention constructs a closed-loop quality management system spanning pre-construction, construction, and post-construction phases. Environmental data, behavioral records, tool usage data, and test data are all automatically collected and digitally stored in a standardized manner, ensuring the complete traceability of the installation process and effectively solving the problems of incomplete records and fragmented information in traditional solutions. This invention comprehensively evaluates the quality status of different processes and stages during GIS installation through process-level, equipment-level, and multi-time-scale quality scoring and trend analysis models. It not only reflects the quality level at a single moment but also identifies quality fluctuation trends and potential deterioration risks, enabling early warning and process correction of quality problems. This reduces the probability of rework and potential post-operation failures, significantly improving overall installation reliability.

[0012] Furthermore, the system of this invention adopts a modular and scalable architecture, which can flexibly connect to different types of sensors and algorithms, and is suitable for various substations and different models of GIS equipment, with good scalability and application value.

[0013] Overall, this invention significantly improves the traceability, controllability, construction efficiency, and management transparency of installation quality, reduces equipment failure rates caused by installation defects, and provides technical support for the long-term safe operation of GIS equipment. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the four-layer architecture in the system of the present invention.

[0015] Figure 2 This is a schematic diagram of the hardware modules of the system of the present invention.

[0016] Figure 3 This is a flowchart of the intelligent algorithm used in the system of this invention.

[0017] Figure 4 This is a schematic diagram of the software interface of the system of the present invention. Detailed Implementation

[0018] The invention will now be further described with reference to the accompanying drawings.

[0019] like Figure 1 As shown, this system adopts a four-layer architecture, including a data acquisition layer, a data processing layer, an analysis and decision-making layer, and an application layer. The data acquisition layer collects key data in real time during the GIS installation process through various sensors and device interfaces, such as ambient temperature, humidity, air cleanliness (dust particle concentration), and equipment installation posture parameters. This layer also integrates data interfaces for programmable logic controllers (PLCs) and intelligent tools, ensuring comprehensive data access to physical quantities and operation records during the installation process. The data processing layer uses an efficient distributed message queue to transmit real-time data streams and utilizes big data processing frameworks (Spark or Flink) to perform data cleaning, noise reduction, and feature extraction, providing standardized input for subsequent intelligent analysis. The analysis and decision-making layer integrates artificial intelligence algorithms to perform in-depth analysis of the collected data, detects abnormal patterns using the isolated forest algorithm, predicts potential risks by combining historical failure cases, and constructs an installation quality evaluation model to generate dynamic scores. The application layer provides functions such as visual monitoring, early warning information push, and electronic report generation through a web or mobile interface, enabling remote monitoring and management of the GIS installation site.

[0020] like Figure 2As shown, the hardware of this system includes several key modules: a high-definition camera, an environmental sensor module, an electronic theodolite, an intelligent control host, a human-machine interface device, a voice warning unit, alarm lights, and a communication interface. The high-definition camera is preferably 8 megapixels with 4K resolution, capable of capturing high-precision images at the GIS installation site, collecting and monitoring information on personnel status (identity information, protective clothing, actions) and equipment status (installation progress, completion and acceptance status). The camera is connected to the intelligent control host with a built-in AI chip, enabling real-time analysis of video data and target recognition. The intelligent control host uses a high-performance artificial intelligence processor (such as Rockchip RV1126), integrating a multi-core CPU and an independent neural network acceleration unit, possessing a computing power of over one trillion operations per second, accelerating the algorithm model, and achieving a target detection frame rate of no less than 40 FPS. The environmental sensor module includes temperature sensors, humidity sensors, and air cleanliness sensors, used to accurately measure environmental parameters at the installation site, ensuring that the equipment installation accuracy and conditions meet standard requirements. Due to the large size of GIS equipment and the limited measurement range of general displacement sensors, this system uses an electronic theodolite for measurement. The measurement results are combined with image recognition technology to monitor and control the equipment installation accuracy (such as alignment deviation and levelness). High-definition cameras are installed in key locations such as the personnel operation area, flange alignment area, and sealing strip installation area to collect image information on personnel movements, tool usage, and component status. Temperature and humidity sensors and air cleanliness sensors are placed in the equipment installation area, spare parts storage area, and busbar connection area to monitor the ambient temperature in real time. T ,humidity H Particle concentration P Key environmental parameters, such as pitch angle, are continuously monitored by electronic theodolites at critical docking points of GIS equipment to obtain attitude parameters. Electronic theodolites are deployed at key docking locations of GIS equipment to continuously monitor component alignment angles, horizontal deviations, and axis consistency, thereby acquiring attitude parameters such as pitch angle. θ Offset angle φ and alignment deviation d Insulation testing equipment, pressure testing equipment, and partial discharge detection equipment are used to acquire test data for key testing stages after installation, such as insulation resistance, chamber pressure, and partial discharge amplitude. All data is uniformly accessed through the message queue module within the intelligent control host, enabling synchronous acquisition and data standardization across devices, forming the foundational data source for subsequent quality calculations in this invention.

[0021] Human-machine interface devices, such as touch screens, are used to display real-time data including current camera feeds, various sensor parameters, personnel identification information, and equipment status, and provide an operation and control interface. Each hardware module connects to the monitoring host via wired or wireless means to achieve synchronous data transmission and centralized processing. Wired interfaces may include USB, RJ45 Ethernet, RS485, etc., while wireless communication interfaces may include 4G, WiFi, Bluetooth, etc. The acquisition system has data verification and anomaly detection functions to ensure the accuracy and reliability of data transmission.

[0022] The flowchart of the intelligent algorithm in this invention is as follows: Figure 3 As shown, the software platform of this invention incorporates multiple intelligent algorithms to achieve automatic identification and analysis of personnel, equipment, and the environment. Among them, the target detection algorithm is based on an improved YOLOv5 deep learning model, optimized for the complex and ever-changing scenarios at GIS installation sites. Due to the numerous targets, complex backgrounds, and significant lighting effects at installation sites, the original YOLOv5 model is insufficient for real-time detection tasks. This invention improves model accuracy and robustness through the following improvements: replacing the original C3 module with the GhostBottleneckCSP module, obtaining rich features with fewer parameters, thus making the model more lightweight and efficient; introducing the SENet attention mechanism to enhance the model's feature recognition ability for different targets and backgrounds, enabling the extraction of key features even when target objects differ significantly in different anomaly detection tasks, thereby improving detection accuracy; and utilizing the Mosaic data augmentation strategy to expand the dataset of targets related to abnormal behavior, increasing sample diversity and improving the model's generalization ability to rare working conditions. Through the above improvements, the target detection algorithm can accurately identify targets such as personnel, protective equipment, tools, and equipment parts at the construction site, and automatically determine whether operators are wearing safety helmets properly, whether tools are being used correctly, and whether parts are installed in place.

[0023] This invention also introduces personnel behavior recognition and posture analysis algorithms to supervise the standardization of operations during installation. The system captures the operator's movements by deploying high-definition cameras (combined with wearable sensors when necessary) and integrates computer vision and deep learning technologies for intelligent analysis of the behavior. First, the raw video data acquired by the cameras is preprocessed, including using a Kalman filter algorithm to fuse data from sensors such as IMUs, and cleaning, denoising, and calibrating the motion trajectory signals to improve the accuracy and stability of the data. Then, a convolutional neural network (CNN) is used to extract spatial features from the image sequence, and a long short-term memory network (LSTM) is combined to capture the patterns of movement changes over time, thereby identifying the operator's skeletal posture and continuous movement trajectory. Based on a pre-trained machine learning model, the extracted movement features are classified and analyzed to identify non-standard or dangerous operating behaviors, such as smoking, using incorrect tools, or other potentially unsafe actions. Once abnormal behavior is detected, the system will issue corrective prompts to the operator in real time via voice or interface and record the violation. The combination of the above behavior recognition algorithm and target detection algorithm enables the system to not only monitor the environment and equipment status, but also intelligently monitor unsafe human factors, thereby comprehensively ensuring the quality and safety of the GIS installation process.

[0024] This system utilizes multi-sensor collaboration to achieve real-time monitoring and anomaly detection mechanisms for the installation environment. Data such as temperature, humidity, and cleanliness collected by environmental sensors are processed and compared with preset safety threshold ranges. When the ambient temperature or humidity exceeds the permissible range, or the air cleanliness (e.g., dust particle concentration) exceeds the standard limit, the system automatically determines an environmental anomaly and immediately triggers an early warning mechanism. Warning methods include playing a voice alarm such as "Environmental parameters are abnormal, please stop installation" via a voice prompt unit, and displaying a warning dialog box on the screen to remind the operator. For example, when the indoor temperature is outside the range of -10℃ to 40℃, the interface will display a warning message "Temperature abnormal, installation prohibited"; when humidity exceeds the limit, a prompt "Humidity abnormal, installation prohibited" will appear, ensuring that installation work is promptly suspended when environmental conditions are not met. Simultaneously, the system automatically saves information such as abnormal parameters and the time of occurrence to the alarm log, generating an alarm record for post-event traceability and analysis. Furthermore, the analysis and decision-making layer utilizes the isolated forest machine learning algorithm to perform pattern analysis on multi-dimensional sensor data. This can detect abnormal trends or subtle fault signs that are difficult to identify in a timely manner using simple threshold judgments, and issue early warnings. Through the above environmental monitoring and alarm mechanisms, it is ensured that the GIS installation process is always carried out under controlled and safe conditions.

[0025] This method constructs a quality element list database based on GIS installation specifications and quality requirements. It categorizes and organizes installation accuracy, installation cleanliness, sealing performance, humidity control, and insulation test results into structured quality element sets. Each quality element includes fields such as inspection items, standard values, permissible deviations, testing methods, and weights. This database serves as the target space for quality evaluation, enabling all subsequently collected data to be mapped to corresponding quality inspection items, thus establishing a standardized and comparable basis for quality judgment.

[0026] The first step will be within the time window. t Environmental features collected internally E t =[ T t , H t , P t ] T Posture characteristics Behavioral characteristics obtained from camera recognition and experimental characteristics X t =[ x t,1 , x t,2 ,… T Mapped to a uniform quality feature vector Q t This mapping process is implemented using the following linear weighted model: ; in W E This is an environmental quality mapping weight matrix, used to characterize the influence weight of environmental quality elements in the overall installation quality evaluation. The weight values ​​are derived from the importance ranking of environmental quality indicators in the quality element list. The corresponding quality element types include environmental constraints such as installation environment temperature, humidity, and air cleanliness. W S This is a weighted matrix mapping equipment posture and installation accuracy, used to reflect the weight relationship of structural quality elements such as equipment installation posture and alignment accuracy in quality evaluation. Key corresponding features include alignment angle deviation, levelness, coaxiality deviation, flange connection offset, and equipment geometric installation error. W A A weight matrix is ​​mapped to personnel behavior and operational procedures to characterize the influence of human factors such as personnel operation behavior and the standardization of tool use on quality evaluation. The main content includes whether protective equipment is worn according to specifications, whether tools are used correctly, whether the operation sequence meets process requirements, and whether there are any violations or dangerous operational behaviors. WX A weight matrix is ​​used to map the test and verification results, representing the weight relationship of various test results in the overall quality evaluation after installation. Corresponding data includes loop resistance measurement results, gas moisture content and gas ratio, insulation resistance, partial discharge detection results, pressure and sealing test results, etc. This model unifies features from different sources, with different physical meanings, and different data structures into a single quality vector space, enabling quality analysis to leap from a single dimension to multi-dimensional fusion, thus providing a more comprehensive description of the quality status of GIS installation. In this process, the invention normalizes, denoises, aligns the time series, and imputs data to ensure the stability of the feature mapping process under harsh field conditions.

[0027] The second step is to obtain the quality feature vector. Q t Then, compliance calculations are performed for each quality inspection item based on the quality element list database. For any inspection item... i Its time window t The measured value is M i,t The standard value is S i The allowable deviation is Δ i Its quality compliance is defined as: ; If the deviation exceeds the allowable range, then directly set =0. This compliance model embodies the logic of "the greater the deviation, the lower the score," and can be applied to various quality factors such as installation accuracy deviation, environmental deviation, and test result deviation, achieving unified quantification of quality items.

[0028] The third step is to weight and sum the compliance rates of all inspection items corresponding to each installation process to form a process-level quality score. The process-level scoring model is as follows: ; The process-level scoring model can reflect the comprehensive performance of a single process in various quality elements, so that each node in the installation process has a clear scoring basis.

[0029] Fourthly, since the GIS installation process involves multiple key steps such as cleaning, alignment, sealing, tightening, and testing, this invention further aggregates the scores of all processes according to their importance weights to obtain an equipment-level quality score. : ; in, α tThe importance weights are derived from the sensitivity ranking of processes in the quality element database. Since processes such as sealing installation and insulation testing have a higher impact on equipment safety, this invention assigns them greater weights, making the equipment-level quality score more consistent with the characteristics of engineering risks.

[0030] The fifth step involves using an unsupervised anomaly detection model of isolated forests to analyze the quality feature vectors. Q t Intelligent anomaly detection is used to identify quality risks such as sudden environmental changes, abnormal postures, behavioral violations, and experimental fluctuations. The Isolation Forest model outputs anomaly scores. A t ∈[0,1], and calculate the deduction value according to the degree of abnormality. D t = βA t This results in the revised process-level score: ; This mechanism can effectively reflect real-world anomalies, making the quality score more sensitive and robust.

[0031] Step 6: For quality performance across time periods, a multi-time-scale quality trend model based on a sliding window was constructed. Let the time window length be... h The quality trend is then defined as: ; By selecting different window lengths, process-level, equipment-level, and project-level quality trends can be obtained respectively, enabling the present invention to not only analyze instantaneous quality but also capture quality fluctuation trends and potential signs of degradation.

[0032] like Figure 4 As shown, to facilitate user operation and viewing, the software platform of this invention provides a highly integrated and user-friendly graphical interface. Each functional module is clearly divided, supporting data entry, real-time monitoring, anomaly alerts, and result visualization. The system homepage adopts an intuitive navigation layout, featuring eight core functional modules: equipment installation information, historical test data, test information, environmental confirmation, alarm records, environmental monitoring records, system settings, and tool registration ledger. Users can quickly access the corresponding functional interface by clicking the icons.

[0033] The equipment installation information module is used to register basic information about the GIS equipment installation process, including bay name, air chamber name, docking location, construction personnel, etc. The system provides standardized data input controls to avoid formatting errors. This module is also linked to the "Environmental Confirmation" and "Environmental Testing" sub-interfaces to realize the process management of pre-construction condition review.

[0034] The historical test data module enables overall retrospective analysis of construction quality. This module integrates and saves content such as "environmental confirmation records," "environmental monitoring records," "tool and equipment registration ledger," and "test records." Users can filter and view by date or project, and export the data to Excel reports with one click for archiving or analysis.

[0035] The test information module is used to enter key test results after the GIS installation is completed, including loop resistance measurement, SF6 / N2 mixed gas moisture content test, gas ratio test and partial discharge test. Users can fill in the test values, test locations and test dates one by one. Each test interface supports scrolling display of a large number of record items and provides a date selection control to ensure uniform format and convenient data entry.

[0036] The environmental verification module is used to verify the conditions before construction. The interface supports filling in basic information such as interval name, air chamber name, and docking location, and inputting the start and end time of the operation through the time selection control, eliminating the need for manual input by the user. For verification items such as cleanliness measures, ground treatment, and tool cleaning before construction, radio buttons can be used to mark them, and the "Completed / No such item" option ensures the completeness and standardization of the records.

[0037] The alarm logging module is used to centrally display all abnormal alarms automatically generated by the system. When abnormal parameters or violations occur in environmental monitoring, test data, or installation information records, the system will automatically create an alarm record, including the time of occurrence, object name, and description of the abnormality. Users can delete false alarm records caused by incorrect input, ultimately forming an accurate database of construction anomalies, and supporting the export of reports for subsequent quality analysis.

[0038] The environmental monitoring module is responsible for displaying real-time data on indoor and outdoor temperature, humidity, and air cleanliness (particle concentration) uploaded by hardware devices. It also allows users to manually input calibration values ​​under special circumstances. When environmental parameters exceed the system's preset thresholds, the interface will immediately pop up a warning dialog box, such as "Abnormal temperature, installation prohibited" or "Abnormal humidity, installation prohibited," and will prevent the user from proceeding to the next step, ensuring that installation is carried out in a compliant environment.

[0039] The system settings module provides auxiliary functions such as language switching, software version viewing, and system exit, making it easy for users to adjust the interface language or view the system running status according to their needs.

[0040] The tool and equipment registration ledger module is used to record the issuance and return of construction tools. Users can enter the name, quantity, recipient, and issuance time of the tools and equipment. The system provides a drop-down list to automatically complete the names of commonly used tools, improving data entry efficiency. After each save, the tools and equipment are automatically imported into the "Tool and Equipment Registration Ledger" interface. The input area can be cleared to quickly enter the next entry. When returning tools and equipment, users can click on the corresponding entry to fill in the return time, forming a complete closed-loop management of tool and equipment usage.

[0041] Through the synergy of the above eight functional modules, the software platform of this invention realizes visualized management, structured recording and intelligent prompts for the entire GIS installation process, enabling construction personnel to complete the installation process under the guidance of a clear interface, and ensuring that all key data is accurately entered and all abnormalities are promptly reported, thereby greatly improving the efficiency, standardization and traceability of construction quality management.

Claims

1. A method for intelligently assessing the installation quality of GIS equipment, characterized in that, A quality element list database is constructed based on GIS installation specifications and quality requirements. This database serves as the target space for quality evaluation, enabling all subsequently collected data to be mapped to the corresponding quality inspection items. The process includes the following steps: The first step will be within the time window. t The environmental features, posture features, behavioral features identified by the camera, and experimental features collected internally are mapped into a unified quality feature vector. The second step is to calculate the compliance of each quality inspection item based on the quality element list database after obtaining the quality feature vector. The third step is to weight and sum up the compliance of all inspection items corresponding to each installation process according to their weights to form a process-level quality score. The fourth step is to aggregate all process scores according to their importance weights to obtain the equipment-level quality score; The fifth step involves using an isolated forest unsupervised anomaly identification model to intelligently identify anomalies in the quality feature vectors. The isolated forest model outputs anomaly scores, and deduction values ​​are calculated based on the degree of anomaly to form a corrected process-level score. The sixth step involves constructing a multi-timescale quality trend model based on a sliding window for quality performance across different time periods.

2. The method according to claim 1, characterized in that, In the first step, the quality feature vector Q t The mapping process is implemented using the following linear weighted model: ； in W E The environmental quality mapping weight matrix represents environmental characteristics. E t =[ T t , H t , P t ] T , W S The weight matrix maps device attitude to installation accuracy, and the attitude features are... , W A A weight matrix is ​​used to map personnel behavior to operational procedures, and behavioral characteristics are represented. , W X The weight matrix maps experimental and validation results, and the experimental features... X t =[ x t,1 , x t,2 ,… T ; The second step includes checking any item. i Its time window t The measured value is M i,t The standard value is S i The allowable deviation is Δ i Its quality compliance is defined as: ； If the deviation exceeds the allowable range, then directly set =0; In the third step, process-level quality scores are generated. The process-level scoring model is as follows: ； Step 4: Equipment-level quality assessment : ； in, α t The importance weights are derived from the process sensitivity ranking in the quality element database; Step 5: The Isolation Forest model outputs anomaly scores. A t ∈[0,1], deduction value D t = βA t The revised process-level scores are as follows: ； Step 6: Set the time window length to... h The quality trend is then defined as: ； By selecting different window lengths, quality trends at the process level, equipment level, and project level can be obtained respectively.

3. The method according to claim 2, when applied to a GIS equipment installation quality intelligent assessment system, is characterized in that... The system comprises a data acquisition layer, a data processing layer, an analysis and decision-making layer, and an application layer. The data acquisition layer collects key data in real time during the GIS installation process through various sensors and device interfaces. This layer also integrates data interfaces for programmable logic controllers and intelligent tools. The data processing layer uses an efficient distributed message queue to transmit real-time data streams and leverages a big data processing framework to perform data cleaning, noise reduction, and feature extraction, providing standardized input for subsequent intelligent analysis. The analysis and decision-making layer integrates artificial intelligence algorithms to perform in-depth analysis of the collected data, detects abnormal patterns using the isolated forest algorithm, predicts potential risks by combining historical failure cases, and constructs an installation quality evaluation model to generate dynamic scores. The application layer provides visual monitoring, early warning information push, and electronic report generation functions through a web or mobile interface, enabling remote monitoring and management of the GIS installation site.

4. The system according to claim 3, characterized in that, The system's hardware includes a high-definition camera, an environmental sensor module, an electronic theodolite, an intelligent control host, a human-machine interaction device, a voice warning unit, an alarm light, and a communication interface. The intelligent control host is connected to the high-definition camera, the environmental sensor module, the electronic theodolite, the human-machine interaction device, the voice warning unit, the alarm light, and the communication interface. The intelligent control host includes an AI processor, a neural network acceleration unit, and data storage. The communication interface includes a wired communication interface and a wireless communication interface.

5. The system according to claim 4, characterized in that, The intelligent control host has a built-in target detection algorithm based on an improved Yolov5 deep learning model. It replaces the original C3 module with the GhostBottleneckCSP module, introduces the SENet attention mechanism to enhance the model's ability to identify features of different targets and backgrounds, and uses the Mosaic data augmentation strategy to expand the dataset of targets related to abnormal behavior.

6. The system according to claim 5, characterized in that, The intelligent control host also incorporates human behavior recognition and posture analysis algorithms. First, it preprocesses the raw video data acquired by the high-definition camera, including using a Kalman filter algorithm to fuse IMU sensor data, and cleaning, denoising, and calibrating the motion trajectory signal. Then, it uses a convolutional neural network to extract spatial features from the image sequence and combines a long short-term memory network to capture the patterns of motion changes over time, thereby identifying the operator's skeletal posture and continuous motion trajectory. Based on a pre-trained machine learning model, it classifies and analyzes the extracted motion features to identify non-standard or dangerous operating behaviors.

Images