Method and system for AI intelligent research and judgment and disposal matching of engineering monitoring equipment abnormal data

Abstract

This invention relates to an AI-powered intelligent analysis and handling matching method and system for abnormal data from engineering monitoring equipment. The method includes the following steps: data acquisition and preprocessing; AI model training and feature binding; multi-dimensional progressive analysis; abnormal handling matching and output; and knowledge accumulation and iteration. Based on a dual-core logic of "data-driven + knowledge binding," this invention trains an AI model by collecting typical abnormal data, establishes a correlation library between abnormal features and operational knowledge, and then verifies it through multi-dimensional progressive analysis. Finally, it outputs an integrated result of "abnormal identification - cause location - measure matching," achieving an intelligent upgrade from "superficial recognition" to "precise handling."

Description

Technical Field

[0001] This invention belongs to the field of intelligent engineering monitoring technology, specifically relating to an AI-based intelligent analysis and handling matching method and system for abnormal data from engineering monitoring equipment. Background Technology

[0002] Engineering monitoring refers to the continuous measurement and monitoring of data such as deformation, stress, and state of buildings, structures, soil and rock masses, or their surrounding environment using specialized instruments, equipment, and methods. Data analysis is then used to assess their safety and stability. When abnormal data is detected, a maintenance plan must be developed to address the malfunction.

[0003] Existing engineering monitoring methods have the following limitations: 1. In terms of anomaly detection, the system analyzes the single-point data sequence of the monitoring equipment using a preset formula to identify superficial features such as exceeding the threshold and abnormal trends in the data curve, triggering alarms. However, it cannot provide targeted troubleshooting steps and maintenance suggestions based on the equipment type and operating conditions. Anomaly identification and handling are disconnected, resulting in low alarm usability.

[0004] 2. In terms of data presentation, it only outputs basic information such as monitoring data trend charts and raw data listings, lacking in-depth analysis of the potential causes of anomalies and failing to accurately pinpoint the root cause of the problem.

[0005] 3. Regarding the causes and handling, the analysis of potential causes of the anomalies is not involved, and no specific investigation steps or maintenance suggestions are provided. It relies entirely on manual experience for judgment, resulting in low data processing efficiency and high time costs, which hinders the implementation of intelligent engineering monitoring.

[0006] 4. In terms of knowledge reuse, the operational experience of domain experts has not been systematically bound with the specific abnormal characteristics of the equipment, resulting in insufficient knowledge reuse, strong generalization of analysis results, and inability to directly guide on-site operations. 5. In terms of analysis dimensions, the analysis only focuses on the data sequence analysis of single-point equipment, without combining multi-dimensional information such as comparison of multiple measurement points in the same project, original equipment parameters, and working environment for comprehensive judgment, which is prone to misjudgment and omission. Summary of the Invention

[0007] The problem this invention aims to solve is to provide an AI-based intelligent analysis and matching method and system for abnormal data from engineering monitoring equipment. This system constructs a closed-loop system of "data feeding - feature binding - multi-dimensional progressive analysis - matching of handling measures". Using typical monitoring equipment as a carrier, it trains an AI model to achieve accurate binding of abnormal features with equipment abnormalities, causes, and handling measures. It combines multi-dimensional data to carry out progressive analysis and outputs integrated analysis results of "identification - location - measures".

[0008] To solve the above-mentioned technical problems, one technical solution adopted by the present invention is: an AI-based intelligent judgment and handling matching method for abnormal data from engineering monitoring equipment, comprising the following steps: S1: Data acquisition and preprocessing. Collect basic data, monitoring data, raw data and auxiliary data of the target equipment, and preprocess them to form standardized analytical data. S2: AI model training and feature binding: The AI ​​model is trained using typical abnormal data curves labeled with equipment abnormalities, causes and handling measures. Abnormal curve features are extracted, and a mapping relationship library between the abnormal curve features and "equipment abnormality - cause - handling measures" is constructed. S3: Multi-dimensional progressive analysis, based on the mapping relationship library, performs layer-by-layer analysis on the preprocessed data; S4: Anomaly handling matching and output, integrate analysis results, generate and output an integrated analysis report containing anomaly conclusions, cause location and specific handling suggestions; S5: Knowledge accumulation and iteration, storing the raw data, process data, result data and handling effect feedback in the cloud; using the handling effect feedback to optimize and update the mapping relationship library, and using new abnormal scenario data to iteratively train the AI ​​model.

[0009] Furthermore, in S1, the basic parameters include structure name, measuring point number, equipment model, algorithm type, core formula, initial value, and threshold; the monitoring data is real-time monitoring data or historical monitoring data of the target equipment, used to generate a data trend chart; the raw data is unprocessed underlying data originally collected by the equipment; the auxiliary data includes synchronous monitoring data of other measuring points in the same project, equipment installation information, and working environment data.

[0010] Furthermore, S2 includes the following steps: S21: Collect typical abnormal data curves for at least the following scenarios: equipment failure, external interference, and installation and commissioning. S22: Label each of the typical abnormal data curves with the corresponding "equipment abnormality - cause - handling measures" triplet information; S23: Use deep learning algorithms to train an AI model, so that the AI ​​model learns and establishes a mapping relationship between the abnormal curve features and the triplet information, and finally constructs a four-dimensional association library of "abnormal features - equipment abnormality - cause - handling measures".

[0011] Furthermore, S3 includes the following steps: S31: The first layer performs single-point data feature recognition, uses the trained AI model to identify abnormal features of the target device monitoring data, and initially matches candidate abnormal situations and their causes. S32: The second layer performs a multi-point comparative analysis of the same project, comparing the data of the target equipment with other measuring points in the same project to determine whether the anomaly is a local anomaly or a system-level anomaly. S33: The third layer performs in-depth analysis of the original data of the equipment, retrieves the original data of the target equipment, verifies and narrows down the cause range of the preliminary match; S34: The fourth layer performs expert experience and working condition matching analysis, and combines equipment characteristics and working environment data to finally pinpoint the remaining candidate causes.

[0012] Furthermore, in S33, for the hydrostatic level, at least the analysis is performed on whether its pressure value is continuously negative, the stability of its data form, and the auxiliary temperature parameters; for the strain gauge, at least the analysis is performed on its amplitude, frequency, signal-to-noise ratio, and temperature-stress fitting correlation coefficient.

[0013] Furthermore, in S34, the cause is finally determined and optimized treatment suggestions are made by combining the equipment sensing principle, historical operation and maintenance experience, construction activity records, and temperature change data.

[0014] Furthermore, in S4, the integrated analysis report includes a basic monitoring parameter table, a monitoring data trend chart of measuring points, a raw data chart of the equipment, clear conclusions about anomalies, precise location of the causes, and actionable step-by-step handling suggestions.

[0015] Furthermore, it also includes multi-device expansion and adaptation. For different types of monitoring devices, it trains customized AI sub-models for devices by configuring exclusive device parameters and building exclusive anomaly feature-operation and maintenance knowledge association library to adapt to the analysis of new devices.

[0016] Furthermore, this invention also provides an intelligent inspection system for video surveillance systems based on browser automation and visual AI, utilizing the aforementioned AI-based intelligent judgment and handling matching method for abnormal data from engineering monitoring equipment, including: The data acquisition and preprocessing module is used to acquire and preprocess multi-source data from the target device; The AI ​​model training and feature binding module is used to train AI models and build a mapping relationship library between abnormal features and operation and maintenance knowledge. The multi-dimensional progressive analysis module is used to perform progressive analysis on the preprocessed data to accurately locate the cause of anomalies. The anomaly handling matching and output module is used to generate and output an integrated analysis report containing handling recommendations; The knowledge accumulation and iteration module is used to store data, provide feedback on results, and iteratively optimize the model and knowledge base.

[0017] Furthermore, it also includes the S6 multi-device expansion and adaptation module, which can be used to quickly adapt to different types of monitoring devices by configuring dedicated parameters and related libraries.

[0018] 1. This invention achieves integrated "identification-location-measures" for anomalies, solving the problem of disconnected handling: Through a three-dimensional precise binding mechanism, AI can not only identify anomalies, but also accurately locate the causes and provide specific handling steps. After testing, the pertinence of anomaly handling is improved, and the on-site operation guidance is enhanced.

[0019] 2. This invention improves the accuracy of judgment through multi-dimensional analysis: The four-layer progressive analysis logic avoids misjudgment from a single dimension, accurately locates the cause of anomalies, and reduces the rate of missed judgments.

[0020] 3. This invention enables the systematic reuse of expert experience: Through a standardized association library and knowledge accumulation mechanism, scattered expert experience is transformed into reusable intelligent rules, allowing new operation and maintenance personnel to quickly carry out handling without accumulating experience, thus reducing reliance on manual labor.

[0021] 4. This invention significantly improves data processing efficiency: it automates data collection, analysis, and report generation, reducing the time for single-device anomaly analysis from 1 hour in the existing technology to less than 10 minutes, and improving overall operation and maintenance efficiency by 600%.

[0022] 5. This invention supports expansion to multiple devices and has wide applicability: The modular design can be quickly adapted to various engineering monitoring equipment such as static level, strain gauge, and vibrating wire equipment, without the need to repeatedly develop core logic, thus reducing the cost of technology implementation.

[0023] 6. This invention reduces engineering monitoring risks: It accurately identifies abnormalities such as equipment failure and construction interference in advance, avoiding monitoring data failure or engineering safety hazards caused by untimely handling of abnormalities.

[0024] 7. This invention is based on the dual core logic of "data-driven + knowledge binding". It trains AI models by collecting typical abnormal data, establishes a correlation library between abnormal features and operation and maintenance knowledge, and then verifies through multi-dimensional progressive analysis. Finally, it outputs an integrated result of "abnormal identification - cause location - measure matching", realizing an intelligent upgrade from "appearance recognition" to "precise handling". Attached Figure Description

[0025] Figure 1 This is an overall flowchart of an embodiment of the present invention.

[0026] Figure 2 This is a flowchart of AI model training and feature binding according to an embodiment of the present invention.

[0027] Figure 3 This is a flowchart of the multi-dimensional progressive anomaly assessment process according to an embodiment of the present invention. Detailed Implementation

[0028] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] The embodiments of the present invention will be further described below with reference to the accompanying drawings: like Figure 1 As shown, the AI-powered intelligent analysis and handling matching method for abnormal data from engineering monitoring equipment includes the following steps: S1: Data acquisition and preprocessing, including the following: Collect basic data of the target equipment: structure name, measuring point number, equipment model, algorithm type, core formula, initial value, threshold (such as the range of the hydrostatic level and the temperature-stress fitting formula of the strain gauge).

[0030] Monitoring data: Real-time / historical monitoring data of the target equipment (such as pressure data from a hydrostatic level and strain values ​​from a strain gauge), generating data trend charts.

[0031] Raw data: Data originally acquired by the equipment (such as relative reference point data for each water channel of the hydrostatic level, voltage values ​​of strain gauges, and frequency data).

[0032] Auxiliary data: synchronous monitoring data from other measuring points in the same project, equipment installation information, and working environment data (temperature, construction activity records).

[0033] Preprocessing includes the following: Data cleaning: Remove invalid data (such as extreme outliers caused by sensor malfunctions) and fill in missing data.

[0034] Standardized formatting: Organize parameters and data from different devices according to a unified template (such as monitoring basic parameter tables and data trend chart specifications) to support cross-device comparison.

[0035] S2: AI model training and feature binding, such as Figure 2 As shown, an AI model is trained using typical abnormal data curves labeled with equipment abnormalities, causes, and handling measures. Abnormal curve features are extracted, and a mapping relationship library between the abnormal curve features and "equipment abnormality - cause - handling measures" is constructed.

[0036] Among them, the AI ​​model can use random forest and LSTM neural network to replace CNN convolutional neural network, which is suitable for different types of abnormal curve features (such as LSTM is more suitable for long-term prediction of trend anomalies).

[0037] Specifically, S2 includes the following steps: S21: For the target equipment (such as a hydrostatic level), collect historical data curves under various abnormal scenarios. Typical abnormal data curves should include at least three categories: equipment failure, external interference, and installation / commissioning. Specifically, equipment failure curves include negative pressure curves for pipe leaks, spur curves for sensor malfunctions, and continuously decreasing curves for leaks. External interference curves include abrupt changes in construction disturbances and fluctuations in environmental vibrations. Installation / commissioning curves include curves showing excessively high values ​​before initial adjustment and unstable curves for pipe bends. When historical abnormal data is insufficient, typical abnormal curves (such as simulated negative pressure change curves for pipe leaks) can be generated through simulation to supplement training data.

[0038] S22: Label each typical abnormal data curve with the corresponding "Equipment Abnormality - Cause - Remedial Measures" triplet information. For example, "Negative pressure curve → Pipeline leakage → Check pipeline sealing + Repressurize and debug".

[0039] S23: Using deep learning algorithms (such as CNN convolutional neural networks), the labeled abnormal curve data is input to train an AI model to learn curve features (trends, spikes, abrupt changes, numerical ranges, etc.), and to establish a mapping relationship between abnormal curve features and triplet information. Specifically, the model automatically extracts key features of the abnormal curve, such as "continuously below 0 kPa (negative pressure)," "irregular small fluctuations (spics)," "sudden numerical changes in a short period of time (abrupt changes)," and "continuous decline (leakage)," etc. Finally, a four-dimensional association library of "abnormal features - equipment abnormality - cause - handling measures" is constructed, as shown in Table 1 below: Table 1. Data Table of Four-Dimensional Relationship Library

[0040] S3: Multi-dimensional progressive analysis, such as Figure 3 As shown, based on the mapping relation library, the preprocessed data is analyzed layer by layer. Specifically, S3 includes the following steps: S31: The first layer performs single-point data feature identification. Specifically, the trained AI model is invoked to analyze the monitoring data trend chart of the target device, identify abnormal features (such as trend type: continuous decline / no change / positive and negative trend; fluctuation status: normal / large / small; whether there are spikes, sudden changes, negative pressure, etc.), and initially match candidate abnormal situations and causes. Specifically, based on the identified abnormal features, candidate "equipment abnormal situation-cause-handling measures" are retrieved from the four-dimensional association library.

[0041] S32: The second layer performs multi-point comparative analysis of the same project, that is, extracts synchronous monitoring data from other similar monitoring points under the same monitoring project, compares the consistency between the target equipment data and the data from other monitoring points in the same project, and determines whether the anomaly is a local anomaly or a system-level anomaly. The scope of the anomaly is determined as follows. Abnormal data only in the target device → Preliminary judgment is that it is a local anomaly in a single device (such as equipment failure or single-point installation problem). Multiple devices experiencing data synchronization anomalies → This is determined to be a system-level / environment-level anomaly (such as external construction interference or changes in overall operating conditions).

[0042] Preferably, for some simple devices (such as simple displacement gauges), the analysis process can be simplified, skipping the "comparison of multiple measurement points in the same project" layer and directly moving from "single point identification" to "in-depth mining of raw data".

[0043] S33: The third layer performs in-depth analysis of the original equipment data, retrieves the original data of the target equipment (such as the water path relative benchmark data of the hydrostatic level and the voltage / frequency data of the strain gauge), eliminates candidate results that do not meet the requirements in the association library, and narrows down the range of causes for the initial matching.

[0044] Specifically, focusing on the details of device-level anomalies, the content is as follows: Static level: Focus on analyzing whether the pressure value is continuously in the negative pressure range, whether the data pattern is stable, and whether the auxiliary parameters (temperature) are abnormal.

[0045] Strain gauges: Focus on analyzing amplitude (whether it is below 80mV), frequency (whether it is close to 0Hz), signal-to-noise ratio (whether it is >0dB), temperature-stress fitting correlation coefficient, and whether there is a risk of exceeding the range.

[0046] S34: The fourth layer conducts expert experience and working condition matching analysis, combines equipment characteristics (such as the pressure sensing principle of the hydrostatic level and the temperature compensation mechanism of the strain gauge) with historical operation and maintenance experience to make a final judgment on the remaining candidate causes; and combines working environment data (such as construction activity records and temperature change data) to optimize disposal measures (such as suspending monitoring rather than equipment maintenance for abnormalities during construction) and finally accurately locate the remaining candidate causes.

[0047] Preferably, a "historical data comparison" dimension can be added (e.g., comparing data from the same period last year for the target device) to help determine whether the anomaly is caused by seasonal factors. A rule engine + expert system can be used to replace pure AI training. By manually pre-setting a rule base for anomaly features and related information, it is suitable for anomaly analysis in small-scale and simple scenarios.

[0048] S4: Anomaly Handling Matching and Output. Integrate analysis results to generate and output an integrated analysis report containing anomaly conclusions, cause location, and specific handling recommendations. Specifically, the integrated analysis report includes a basic monitoring parameter table, monitoring data trend charts for measurement points, raw equipment data charts, clear anomaly conclusions, precisely located causes, and actionable step-by-step handling recommendations.

[0049] Specifically, the basic information includes a table of basic monitoring parameters (structure, measuring points, equipment model, formula, initial values, etc.). Data charts include trend charts of measuring point monitoring data and raw equipment data charts. When determining anomalies, the abnormal characteristics must be clearly identified (e.g., "negative pressure < 0 kPa, large fluctuations, continuous decline"). When locating the cause, the exact cause of the anomaly must be determined (e.g., "leakage in the hydrostatic leveling instrument pipeline + construction interference").

[0050] The recommended steps are as follows: 1. Check the sealing of the pipeline joints; 2. Repressurize to the standard pressure; 3. Avoid monitoring during construction periods.

[0051] Output methods include direct display on AI analysis platforms (such as "Anxin Cloud AI"), export of PDF reports, and mobile push notifications.

[0052] Preferably, a "visualization of handling steps" function can be added (such as showing the hydrostatic level pipeline inspection process through animation) to provide more intuitive guidance for on-site operations.

[0053] S5: Knowledge Accumulation and Iteration. The raw data, process data, result data, and feedback on the handling effect (such as whether the anomaly was resolved after the recommended operation) of the analysis process are stored in the cloud. The handling effect feedback is used to optimize and update the mapping relationship library (such as adjusting the matching weight of features and causes), and new abnormal scenario data (such as uncovered abnormal curves) are used to iteratively train the AI ​​model to improve the recognition accuracy.

[0054] The standardized document update process is as follows: Regularly update the standardized document "Abnormal Curve Type - Equipment Abnormality - Cause - Measures" to achieve knowledge accumulation.

[0055] It also includes multi-device expansion and adaptation steps: based on the core technical solution, for different types of monitoring equipment (such as vibrating wire equipment and displacement gauges), by configuring exclusive equipment parameters and building an exclusive abnormal feature-operation and maintenance knowledge association library, a customized AI sub-model for the equipment is trained to adapt to the analysis of new equipment.

[0056] The newly added equipment parameter configuration includes inputting the core formulas, thresholds, and anomaly judgment indicators of the target equipment (such as the frequency threshold of vibrating wire equipment and the stroke range of displacement gauges).

[0057] Build a dedicated relational library: collect typical anomaly curves of target devices and corresponding operation and maintenance knowledge, and train customized AI sub-models for the devices.

[0058] Implementation method: A modular design is adopted. When adding a new device, only the exclusive parameters and related libraries need to be configured, without the need to reconstruct the core analysis logic.

[0059] This invention also provides an intelligent inspection system for video surveillance based on browser automation and visual AI, comprising the following modules: The data acquisition and preprocessing module is used to acquire and preprocess multi-source data from the target device.

[0060] The AI ​​model training and feature binding module is used to train AI models and build a mapping relationship library between abnormal features and operation and maintenance knowledge.

[0061] The multi-dimensional progressive analysis module is used to perform progressive analysis on the preprocessed data to accurately locate the cause of anomalies.

[0062] The anomaly handling matching and output module is used to generate and output an integrated analysis report that includes handling recommendations.

[0063] The knowledge accumulation and iteration module is used to store data, provide feedback on results, and iteratively optimize the model and knowledge base.

[0064] Preferably, it also includes a multi-device expansion and adaptation module, which is used to quickly adapt to different types of monitoring devices by configuring dedicated parameters and related libraries.

[0065] The following uses the "Wuzhong Yellow River Bridge Static Level Instrument Monitoring Project" as an example to illustrate the invention in detail: This project deploys static leveling equipment (e.g., measuring point ND-06-New, equipment number 38812) on the left side of the Wuzhong Yellow River Bridge to monitor bridge settlement. The equipment adopts a water-channel grouping design, with initial values ​​of 2.73 kPa and 0 m. The algorithm is a limited amplitude algorithm, and the measurement range conforms to industry standards. The project also deploys strain gauges (e.g., measuring point YB-24), with the core formula ε=K(F). i ²-F o ²)+K t (T i -T o The initial value is 1533.788Hz. There is a risk of construction interference at the project site, and the ambient temperature fluctuates significantly during winter and spring, requiring accurate identification of equipment malfunctions and external interference.

[0066] Specifically, the following steps are included: S1: Data acquisition and preprocessing, specifically including the following: Equipment basic parameter acquisition: Static level instrument parameters are synchronized from the project monitoring platform via API interface to generate a basic monitoring parameter table, as shown in Table 2.

[0067] Table 2 Basic Parameters of Monitoring Static Level

[0068] Monitoring data and raw data acquisition: hourly synchronous hydrostatic level pressure monitoring data (generating trend charts), raw data of relative base points for each waterway, strain values, voltage, and frequency data of strain gauges.

[0069] Auxiliary data collection: Manually enter construction activity records (e.g., construction on the left lane section from X month X day to X month X day) and daily ambient temperature data.

[0070] Data preprocessing: Remove invalid data with sudden changes in hydrostatic pressure values ​​exceeding 5 kPa, and format the data trend chart according to a unified template.

[0071] S2: AI model training and feature binding, specifically including the following: Training data preparation: Collect hydrostatic leveling anomaly curves for this project and similar projects, including: Pipeline leakage: The curve shows the pressure continuously dropping to -1.2 kPa.

[0072] Sensor malfunction: The curve shows irregular spikes (fluctuation range ±0.5Kpa).

[0073] Construction interference: The data curve suddenly changed to 3.5 kPa during the construction period on [Date].

[0074] Unadjusted initial value: The data remains at 5.8 kPa (far beyond the normal range).

[0075] Each curve is labeled with associated information (e.g., “pipeline leakage → cause: loose connection → measure: tighten connection + pressurize”).

[0076] AI model training: Using a CNN convolutional neural network, 1000 labeled abnormal curves were input for training. After 50 iterations, the model's accuracy in extracting abnormal features reached 88%.

[0077] Association Library Construction: Establish a four-dimensional association library containing 12 types of anomaly characteristics, 8 types of equipment anomalies, 10 types of causes, and 15 handling measures, as shown in Table 3 below: Table 3. Data table for building related libraries

[0078] S3: Multi-dimensional progressive analysis, taking the hydrostatic leveling point "Left Aspect ND-06-New" as an example, specifically includes the following: First layer: Single point data feature recognition: The AI ​​model analyzes the trend chart of the measurement point data and identifies "pressure continuously -0.8Kpa (negative pressure), stable trend, no burrs", and initially matches the candidate result of "pipeline leakage" in the association library.

[0079] Second layer: Comparison of multiple measuring points in the same project: Data from three other measuring points in the same waterway were retrieved, and it was found that only the pressure at this measuring point was negative, while the pressure at the other measuring points was stable at 2.5-2.8 kPa, which was determined to be a local anomaly of a single device.

[0080] Third layer: In-depth analysis of raw data: By examining the raw data of the water circuit relative to the base point at this measuring point, it was found that the pressure value was consistently below 0 kPa without fluctuation, thus ruling out sensor false alarms and verifying the candidate result of "pipeline leakage".

[0081] Fourth layer: Expert experience and working condition matching: Based on the pressure sensing principle of the hydrostatic level (it should normally maintain positive pressure) and the lack of recent construction records, the cause was determined to be "loose pipe joints", and the treatment measures were optimized to "tighten the joints + pressurize and calibrate".

[0082] S4: Anomaly Handling Matching and Output. The system generates an integrated analysis report, the core content of which is as follows: [Anomaly Analysis Report of the Static Level Instrument at Wuzhong Yellow River Bridge] Basic Information: Structure: Wuzhong Yellow River Bridge; Measuring point: Left span ND-06-New; Equipment number: 38812; Initial value: 2.73Kpa, 0m; Data charts: Monitoring data trend chart: Pressure remains at -0.8 kPa, with no significant fluctuations; Original data chart: The relative baseline data of the waterway are all <0 kPa, with no abnormal fluctuations.

[0083] Abnormal conclusion: The equipment has a continuous negative pressure (<0 kPa) abnormality, without burrs or sudden changes.

[0084] Cause of the leak: Loose pipe connections leading to leakage.

[0085] Recommendations for handling: Close the water valve and disconnect the pipe at the connection point.

[0086] Inspect the interface gasket and replace any worn-out gaskets.

[0087] Tighten the interface and pressurize to the standard initial value of 2.73 kPa.

[0088] ④ Monitor continuously for 2 hours to confirm that the pressure is stable.

[0089] The report was displayed through the "Anxin Cloud AI" platform and pushed to the mobile devices of operations and maintenance personnel.

[0090] S5: Knowledge Accumulation and Iteration. Raw data, process data, result data, and feedback on the handling effects are stored in the cloud. The mapping relationship library is optimized and updated using the handling effect feedback, and the AI ​​model is iteratively trained using new abnormal scenario data.

[0091] The advantages and positive effects of the invention are: 1. This invention achieves integrated "identification-location-measures" for anomalies, solving the problem of disconnected handling: Through a three-dimensional precise binding mechanism, AI can not only identify anomalies, but also accurately locate the causes and provide specific handling steps. After testing, the pertinence of anomaly handling is improved, and the on-site operation guidance is enhanced.

[0092] 2. This invention improves the accuracy of judgment through multi-dimensional analysis: The four-layer progressive analysis logic avoids misjudgment from a single dimension, accurately locates the cause of anomalies, and reduces the rate of missed judgments.

[0093] 3. This invention enables the systematic reuse of expert experience: Through a standardized association library and knowledge accumulation mechanism, scattered expert experience is transformed into reusable intelligent rules, allowing new operation and maintenance personnel to quickly carry out handling without accumulating experience, thus reducing reliance on manual labor.

[0094] 4. This invention significantly improves data processing efficiency: it automates data collection, analysis, and report generation, reducing the time for single-device anomaly analysis from 1 hour in the existing technology to less than 10 minutes, and improving overall operation and maintenance efficiency by 600%.

[0095] 5. This invention supports expansion to multiple devices and has wide applicability: The modular design can be quickly adapted to various engineering monitoring equipment such as static level, strain gauge, and vibrating wire equipment, without the need to repeatedly develop core logic, thus reducing the cost of technology implementation.

[0096] 6. This invention reduces engineering monitoring risks: It accurately identifies abnormalities such as equipment failure and construction interference in advance, avoiding monitoring data failure or engineering safety hazards caused by untimely handling of abnormalities.

[0097] 7. This invention is based on the dual core logic of "data-driven + knowledge binding". It trains AI models by collecting typical abnormal data, establishes a correlation library between abnormal features and operation and maintenance knowledge, and then verifies through multi-dimensional progressive analysis. Finally, it outputs an integrated result of "abnormal identification - cause location - measure matching", realizing an intelligent upgrade from "appearance recognition" to "precise handling".

[0098] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for AI-powered intelligent analysis and matching of abnormal data from engineering monitoring equipment, characterized by: Includes the following steps, S1: Data acquisition and preprocessing. Collect basic data, monitoring data, raw data and auxiliary data of the target equipment, and preprocess them to form standardized analytical data. S2: AI model training and feature binding: The AI ​​model is trained using typical abnormal data curves labeled with equipment abnormalities, causes and handling measures. Abnormal curve features are extracted, and a mapping relationship library between the abnormal curve features and "equipment abnormality - cause - handling measures" is constructed. S3: Multi-dimensional progressive analysis, based on the mapping relationship library, performs layer-by-layer analysis on the preprocessed data; S4: Anomaly handling matching and output, integrate analysis results, generate and output an integrated analysis report containing anomaly conclusions, cause location and specific handling suggestions; S5: Knowledge accumulation and iteration, storing the raw data, process data, result data and handling effect feedback in the cloud; using the handling effect feedback to optimize and update the mapping relationship library, and using new abnormal scenario data to iteratively train the AI ​​model.

2. The method of claim 1, wherein the method comprises: In S1, the basic parameters include structure name, measuring point number, equipment model, algorithm type, core formula, initial value, and threshold; the monitoring data is real-time monitoring data or historical monitoring data of the target equipment, used to generate a data trend chart; the raw data is the unprocessed underlying data originally collected by the equipment; the auxiliary data includes synchronous monitoring data of other measuring points in the same project, equipment installation information, and working environment data.

3. The method of claim 1 or 2, wherein the method is characterized by: S2 includes the following steps: S21: Collect typical abnormal data curves for at least the following scenarios: equipment failure, external interference, and installation and commissioning. S22: Label each of the typical abnormal data curves with the corresponding "equipment abnormality - cause - handling measures" triplet information; S23: Use deep learning algorithms to train an AI model, so that the AI ​​model learns and establishes a mapping relationship between the abnormal curve features and the triplet information, and finally constructs a four-dimensional association library of "abnormal features - equipment abnormality - cause - handling measures".

4. The method of claim 1 or 2, wherein the method is characterized by: S3 includes the following steps: S31: The first layer performs single-point data feature recognition, uses the trained AI model to identify abnormal features of the target device monitoring data, and initially matches candidate abnormal situations and their causes. S32: The second layer performs a multi-point comparative analysis of the same project, comparing the data of the target equipment with other measuring points in the same project to determine whether the anomaly is a local anomaly or a system-level anomaly. S33: The third layer performs in-depth analysis of the original data of the equipment, retrieves the original data of the target equipment, verifies and narrows down the cause range of the preliminary match; S34: The fourth layer performs expert experience and working condition matching analysis, and combines equipment characteristics and working environment data to finally pinpoint the remaining candidate causes.

5. The method of claim 4, wherein the method is characterized by: In S33, for the hydrostatic level, at least the analysis should be conducted on whether the pressure value is continuously negative, the stability of the data form, and the auxiliary temperature parameters; for the strain gauge, at least the analysis should be conducted on its amplitude, frequency, signal-to-noise ratio, and temperature-stress fitting correlation coefficient.

6. The method of claim 4, wherein the method is characterized by: In step S34, the cause is determined and optimized treatment suggestions are made by combining the equipment sensing principle, historical operation and maintenance experience, construction activity records, and temperature change data.

7. The method of claim 1 or 2, wherein the method is characterized by: In S4, the integrated analysis report includes a basic monitoring parameter table, a monitoring data trend chart of measuring points, a raw data chart of the equipment, clear conclusions about anomalies, precise location of the causes, and actionable step-by-step handling suggestions.

8. The method of claim 1 or 2, wherein the method is characterized by: It also includes S6 multi-device expansion and adaptation, which allows for the configuration of exclusive device parameters and the construction of exclusive anomaly feature-operation and maintenance knowledge association library for different types of monitoring devices, and the training of customized AI sub-models for devices to adapt to the analysis of new devices.

9. A smart inspection system for video surveillance based on browser automation and visual AI, characterized in that: The AI-powered intelligent analysis and handling matching method for abnormal data from engineering monitoring equipment as described in any one of claims 1 to 8 includes: The data acquisition and preprocessing module is used to acquire and preprocess multi-source data from the target device; The AI ​​model training and feature binding module is used to train AI models and build a mapping relationship library between abnormal features and operation and maintenance knowledge. The multi-dimensional progressive analysis module is used to perform progressive analysis on the preprocessed data to accurately locate the cause of anomalies. The anomaly handling matching and output module is used to generate and output an integrated analysis report containing handling recommendations; The knowledge accumulation and iteration module is used to store data, provide feedback on results, and iteratively optimize the model and knowledge base.

10. The intelligent inspection system for video surveillance based on browser automation and visual AI as described in claim 9, characterized in that: It also includes a multi-device expansion and adaptation module, which can be used to quickly adapt to different types of monitoring devices by configuring dedicated parameters and related libraries.

Images