Civil engineering experiment detection system based on cloud computing technology

Abstract

The invention relates to the technical field of civil engineering experiment detection, in particular to a civil engineering experiment detection system based on a cloud computing technology. Comprising a self-adaptive detection execution module, an intelligent model conversion and association module, a dynamic threshold decision module, a multi-source data fusion verification module, an experiment monitoring simulation verification module, a visual result presentation module, a report intelligent generation module and a cloud computing interaction module. The self-adaptive detection execution module collects detection data, and the multi-source data fusion verification module outputs fusion data; the dynamic threshold decision module generates a threshold parameter, and the experiment monitoring simulation verification module generates a verification result; a verification result is transmitted to the intelligent model conversion and association module; the intelligent report generation module generates a standardized report; and the cloud computing interaction module is connected with the local computing node and the cloud computing node. Through sensor adaptive acquisition, multi-source data purification and BIM association, lightweight presentation and automatic reporting, accurate and stable detection, accurate data model correspondence and efficient and standard result presentation are realized.

Description

Technical Field

[0001] This invention relates to the field of civil engineering experimental testing technology, specifically to a civil engineering experimental testing system based on cloud computing technology. Background Technology

[0002] With the rapid development of the civil engineering industry, experimental testing technology has gradually evolved from traditional manual operation to informatization and intelligentization. Early testing work relied mainly on manual data collection, recording, and analysis, which was not only inefficient but also prone to errors due to human intervention. Later, the introduction of sensor and data processing technologies gradually automated data collection, and some laboratories began using digital equipment to record data and perform simple analyses. In recent years, the rise of BIM technology and cloud computing has provided new technical support for testing work. Some systems are attempting to combine testing data with models, leveraging cloud resources to handle some computational tasks, thus improving the convenience and data processing capabilities of testing work to some extent.

[0003] Current applications of related technologies still face numerous limitations. During data acquisition, sensors are susceptible to environmental factors, and there is a lack of efficient emergency switching mechanisms in case of equipment failure. Probe position and pressure adjustments are difficult to dynamically adapt to the actual conditions of components, resulting in insufficient accuracy and stability of the acquired data. In multi-source data processing, outlier removal methods are inadequate, the integration of different data types lacks a scientific basis for weight allocation, the correlation between data and models is often imprecise, format compatibility is poor during model conversion, and local computing resources struggle to support large-scale task processing. Regarding visualization, model loading is slow, update delays are high, and it is difficult to achieve an intuitive connection between detection results and real-world scenarios. Report generation relies on manual data integration, format standardization is difficult, annotation information is not synchronized in a timely manner during cross-regional collaboration, and effective anti-tampering measures are lacking. Summary of the Invention

[0004] To solve the above-mentioned technical problems, the technical solution provided by the present invention is as follows:

[0005] A civil engineering experimental testing system based on cloud computing technology includes an adaptive testing execution module, an intelligent model conversion and association module, a dynamic threshold decision module, a multi-source data fusion verification module, an experimental monitoring simulation verification module, a visualization result presentation module, an intelligent report generation module, and a cloud computing interaction module.

[0006] The adaptive detection execution module collects detection data and transmits it to the multi-source data fusion verification module. The multi-source data fusion verification module outputs fused data and transmits it to the dynamic threshold decision module and the experimental monitoring simulation verification module, respectively. The dynamic threshold decision module generates threshold parameters, and the experimental monitoring simulation verification module generates verification results. The threshold parameters and verification results are fed back to the adaptive detection execution module. The verification results are transmitted to the intelligent model conversion and association module. The intelligent model conversion and association module establishes the association between the detection data and the structured BIM model and transmits it to the visualization result presentation module. The visualization result presentation module transmits the detection information and anomaly markers to the intelligent report generation module. The intelligent report generation module generates a standardized report. The cloud computing interaction module connects the local computing node and the cloud computing node and feeds back the processing results from the cloud computing node to the adaptive detection execution module, the intelligent model conversion and association module, and the intelligent report generation module.

[0007] Furthermore, the adaptive detection execution module includes an equipment status monitoring module and a core control module. The equipment status monitoring module collects temperature, voltage, and probe pressure signals through analog sensors. The collected signals are processed by a filtering circuit and then transmitted to the core control module. The equipment status monitoring module determines whether the main sensor is faulty by comparing the main sensor data with a preset deviation threshold. After determining the fault, the core control module activates the backup sensor and disconnects the data transmission link of the main sensor. The component surface temperature signal is converted by an analog-to-digital converter and then transmitted to the pressure regulation module. The calculation method for the probe pressure adjustment is as follows:

[0008] ;

[0009] in, This represents the change in probe pressure. The pressure-temperature coupling coefficient is related to the elastic modulus of the component material and the contact area of ​​the probe. This refers to the change in surface temperature of the component.

[0010] When the temperature, voltage, or probe pressure signal value exceeds the safety threshold preset in the core control module, an audible and visual alarm is triggered and the sampling frequency of the data acquisition chip is adjusted.

[0011] The adaptive detection execution module also includes a dual-path air cleaning module for blowing and adsorption cleaning of the component surface; an industrial camera acquires images of the component surface, which are then processed by grayscale conversion and Gaussian filtering before being input into a locally pre-trained image recognition model; after the image recognition model outputs the recognition result, the core control module drives a stepper motor to move the probe along a preset path and adjust the angle between the probe and the component surface; equipment anomaly logs are stored in flash memory and uploaded to the cloud computing node via an Ethernet interface; the cloud computing node analyzes and optimizes the linkage control parameters based on the received log data, and the optimized parameters are verified and then sent back to the core control module.

[0012] Furthermore, the intelligent model conversion and association module includes a format parsing module, a semantic segmentation module, and a model matching module;

[0013] The format parsing module imports the non-parametric model, extracts geometric topology data, and removes redundant vertices; the semantic segmentation module extracts the contour, size, and spatial location features of components through a convolutional neural network; the extracted feature data is encoded and then used to generate a structured BIM model through the BIM model conversion interface.

[0014] The intelligent model conversion and association module uses an association rule mining algorithm to traverse the attribute information of the detection data and the structured BIM model to establish associations. The intelligent model conversion and association module also includes a local risk assessment module, which calculates the risk level of components based on the high-frequency association rules output by the association rule mining algorithm, and uses the risk level calculation results to update the weight coefficients of the rules in the association rule mining algorithm. The intelligent model conversion and association module has built-in conversion plugins that support multiple formats.

[0015] The verification results output by the experimental monitoring simulation verification module are transmitted to the model matching module. The model matching module corrects the mapping relationship between the structured BIM model and the detection data based on the received verification results. The batch model conversion task or historical data optimization task is determined by the local task scheduling module based on the local computing resource utilization rate. When the local computing resource utilization rate exceeds the preset threshold, the task is uploaded to the cloud computing node for parallel processing through the cloud computing interaction module, and the processing result is returned through an encrypted link.

[0016] Furthermore, the dynamic threshold decision module incorporates a local time-series prediction model, which is a long short-term memory network. This model uses historical detection data and environmental parameters as training samples, including temperature and humidity collected by digital sensors. The local time-series prediction model outputs real-time safety threshold predictions. These predictions are compared with the fused data in a comparator. When the fused data exceeds the range defined by the safety threshold prediction, an alarm command is triggered. This alarm command is transmitted to the BIM display module. The BIM display module adds a risk level icon to the corresponding coordinate position in the structured BIM model based on the component identifier associated with the detection data. The threshold adjustment log is encrypted and stored in an encrypted storage device. When performing local threshold adjustments, historical log data stored on cloud computing nodes is retrieved, processed, and input into the local time-series prediction model to optimize the weight parameters.

[0017] Furthermore, the multi-source data fusion verification module includes a multi-channel interface and a data preprocessing module; the multi-channel interface receives sensor data, image data, and laboratory data; the data preprocessing module performs outlier removal on the received data, and the outlier determination rule is: when satisfy or If it is, then it will be removed. Data point values, The mean of the data samples. The standard deviation of the data sample;

[0018] The multi-source data fusion verification module uses a weighted fusion algorithm to integrate the preprocessed multi-source data. The weighted fusion formula is as follows:

[0019] ;

[0020] in, The merged data values For sensor data weighting coefficients, These are the sensor data values ​​after mean filling and normalization. These are the image data weighting coefficients. The values ​​are the image data after grayscale conversion and edge enhancement. For laboratory data weighting coefficients, The values ​​are laboratory data after anomaly removal and unit standardization, and the weighting coefficients satisfy the following: And allocate them based on data credibility scores;

[0021] The data credibility score is calculated based on the data integrity verification and collection accuracy calibration results. Abnormal data that is removed is stored by associating the collection device number, timestamp, and data serial number.

[0022] Furthermore, the experimental monitoring and simulation verification module includes a data integration module and a simulation calculation module. The simulation calculation module generates simulation data. The data integration module extracts local experimental data, on-site monitoring data, and simulation data, and classifies and associates the three types of data according to the component identifier. The experimental monitoring and simulation verification module compares the classified and associated data pairwise through a cross-verification process and calculates the deviation value. When the deviation value is less than the preset allowable deviation threshold, the experimental monitoring and simulation verification module generates an adjustment instruction. The adjustment instruction is transmitted to the adaptive detection execution module via a communication link. The adjustment instruction includes calibration parameters and sampling frequency adjustment coefficients. For components that fail verification, the experimental monitoring and simulation verification module adds a mark with an encrypted digital watermark to the corresponding component in the structured BIM model. The mark is associated with a deviation cause code, which points to a preset cause library containing sensor drift, environmental interference, and equipment error. When the local computing resource occupancy rate exceeds a preset threshold, the large-scale simulation task executed by the simulation calculation module is processed in parallel by calling the computing resources of cloud computing nodes.

[0023] Furthermore, the visualization results presentation module includes a lightweight BIM module; the lightweight BIM module performs surface merging and vertex deletion on the structured BIM model through a mesh simplification algorithm to obtain a lightweight BIM model; the lightweight BIM model is displayed in real time on the local terminal, and the data update delay time of the lightweight BIM model is less than one-tenth of the inspection sampling cycle; when a query is triggered by touch or click operation, the visualization results presentation module calls the inspection data, operation logs and verification results associated with the selected component and presents them on the local terminal; the visualization results presentation module drives the augmented reality device to collect real scene images, and after calibration, overlays and projects the lightweight BIM model and anomaly markers onto the real scene.

[0024] Furthermore, the intelligent report generation module includes a data extraction and integration module. This module extracts detection data, verification results, visualization markers, and anomaly logs from the adaptive detection execution module, experimental monitoring simulation verification module, and visualization result presentation module, and completes data association and integration according to component identifiers. The intelligent report generation module calls pre-stored standardized report templates and automatically fills the integrated data into the corresponding positions of the standardized report templates. Modifications, exports, and annotations to the standardized report templates are all recorded by the intelligent report generation module, and the generated operation log includes the operator's identity, timestamp, and operation details. During cross-regional collaboration, annotations added to the report by collaborating parties are synchronized via cloud computing nodes, and all annotation information is stored using blockchain technology. The standardized report templates are updated from cloud computing nodes through an interface.

[0025] The advantages of this invention compared to the prior art are:

[0026] This invention acquires temperature, voltage, and probe pressure signals using analog sensors and filters them. In case of a main sensor failure, it automatically switches to a backup sensor. Based on the component surface temperature change and the pressure-temperature coupling coefficient, it dynamically calculates the probe pressure adjustment. After cleaning the component surface with dual air paths, an industrial camera captures images, which are then processed by grayscale and Gaussian filtering before being input into a recognition model. This drives the probe to move along a preset path and adjust the included angle. Abnormal logs are uploaded to the cloud to optimize control parameters. When the signal exceeds a threshold, an audible and visual alarm is triggered, and the sampling frequency is adjusted. This achieves intelligent adaptive operation of the detection equipment, effectively avoiding the impact of environmental interference and equipment failure, and significantly improving the accuracy and stability of the detection data acquisition process.

[0027] This invention removes outliers from multi-source data using the 3σ criterion, assigns weights based on data credibility scores to complete the weighted integration of preprocessed data, extracts component features using convolutional neural networks to generate structured BIM models, establishes attribute relationships between data and models using association rule mining, calculates and updates algorithm weight coefficients based on component risk levels, corrects mapping relationships based on verification results, and uploads batch tasks to the cloud for parallel processing when local computing resources are limited. This achieves efficient purification of multi-source data and intelligent association of models, enabling precise correspondence between data and models, and providing reliable data support and an intuitive analysis platform for engineering inspection.

[0028] This invention generates a lightweight BIM model through a grid simplification algorithm, ensuring low-latency real-time display on local terminals. It drives augmented reality devices to collect images of real scenes and overlay the model with anomaly markers. It automatically extracts data from multiple modules and fills them into standardized templates to reduce human error. In cross-regional collaboration, cloud-based annotations are synchronized and blockchain technology is used for evidence storage to prevent tampering. The standardized templates are updated from the cloud to ensure uniform format. This invention achieves intuitive presentation of inspection results and efficient automation of report generation, reducing errors caused by human intervention and making the viewing and collaboration of inspection results more convenient and standardized. Attached Figure Description

[0029] The accompanying drawings, which form part of this application, are used to provide a further understanding of the application and to make other features, objects, and advantages of the application more apparent. The illustrative embodiments and descriptions of this application are used to explain the application and do not constitute an undue limitation of the application.

[0030] In the attached diagram:

[0031] Figure 1 This is a system framework diagram of a civil engineering experimental testing system based on cloud computing technology in Example 1.

[0032] Figure 2 This is a flowchart of the adaptive detection execution module of a civil engineering experimental testing system based on cloud computing technology in Example 1.

[0033] Figure 3 This is a flowchart of the intelligent model conversion and association module of a civil engineering experimental testing system based on cloud computing technology in Example 1.

[0034] Figure 4 This is a flowchart of the multi-source data fusion verification and validation process of a civil engineering experimental testing system based on cloud computing technology in Example 1. Detailed Implementation

[0035] The following detailed description of the embodiments is used to exemplify the principles of this application, but should not be used to limit the scope of this application. That is, the civil engineering experimental testing system based on cloud computing technology in this application is not limited to the described embodiments.

[0036] The present invention will be further described below with reference to embodiments.

[0037] like Figure 1 As shown, a civil engineering experimental testing system based on cloud computing technology includes an adaptive testing execution module, an intelligent model conversion and association module, a dynamic threshold decision module, a multi-source data fusion verification module, an experimental monitoring simulation verification module, a visualization result presentation module, an intelligent report generation module, and a cloud computing interaction module.

[0038] The adaptive detection execution module collects detection data and transmits it to the multi-source data fusion verification module. The multi-source data fusion verification module outputs fused data and transmits it to the dynamic threshold decision module and the experimental monitoring simulation verification module, respectively. The dynamic threshold decision module generates threshold parameters, and the experimental monitoring simulation verification module generates verification results. The threshold parameters and verification results are fed back to the adaptive detection execution module. The verification results are transmitted to the intelligent model conversion and association module. The intelligent model conversion and association module establishes the association between the detection data and the structured BIM model and transmits it to the visualization result presentation module. The visualization result presentation module transmits the detection information and anomaly markers to the intelligent report generation module. The intelligent report generation module generates a standardized report. The cloud computing interaction module connects the local computing node and the cloud computing node and feeds back the processing results from the cloud computing node to the adaptive detection execution module, the intelligent model conversion and association module, and the intelligent report generation module.

[0039] In a specific embodiment, this system is applicable to the experimental testing of various components in civil engineering. Eight modules each undertake key functions in the entire testing process, ensuring smooth data flow and close cooperation. The adaptive testing execution module directly connects to the testing site, capturing raw data in real time to provide a foundation for testing analysis. The multi-source data fusion and verification module integrates scattered data from different sources, transforming fragmented information into a cohesive whole with analytical value. The dynamic threshold decision module, from a safety judgment perspective, and the experimental monitoring simulation verification module, from a data validity perspective, rigorously control the fused data. The results from both modules inversely influence the adaptive testing execution module, allowing for timely adjustments to the acquisition strategy and reducing unnecessary work. The intelligent model conversion and association module builds an intuitive model carrier for abstract testing data, clearly corresponding the data to the component entities. The visualization result presentation module lowers the professional threshold for data viewing, helping staff quickly grasp the core testing situation. The intelligent report generation module meets the standardized requirements of engineering testing, directly outputting standardized results that can be archived or reviewed. The cloud computing interaction module uses a dual-mode connection of Ethernet and wireless communication to connect local and cloud nodes, which ensures stable data transmission in fixed scenarios and meets the flexible needs of outdoor mobile detection. When the local system faces batch data processing or complex model calculations, it can seamlessly call on cloud computing power to significantly improve the overall processing speed.

[0040] Furthermore, such as Figure 2 As shown, the adaptive detection execution module includes an equipment status monitoring module and a core control module. The equipment status monitoring module collects temperature, voltage, and probe pressure signals through analog sensors. The collected signals are processed by a filtering circuit and then transmitted to the core control module. The equipment status monitoring module determines whether the main sensor is faulty by comparing the main sensor data with a preset deviation threshold. After determining the fault, the core control module activates the backup sensor and disconnects the data transmission link of the main sensor. The component surface temperature signal is converted by an analog-to-digital converter and then transmitted to the pressure regulation module. The calculation method for the probe pressure adjustment is as follows:

[0041] ;

[0042] in, This represents the change in probe pressure. The pressure-temperature coupling coefficient is related to the elastic modulus of the component material and the contact area of ​​the probe. This refers to the change in surface temperature of the component.

[0043] When the temperature, voltage, or probe pressure signal value exceeds the safety threshold preset in the core control module, an audible and visual alarm is triggered and the sampling frequency of the data acquisition chip is adjusted.

[0044] The adaptive detection execution module also includes a dual-path air cleaning module for blowing and adsorption cleaning of the component surface; an industrial camera acquires images of the component surface, which are then processed by grayscale conversion and Gaussian filtering before being input into a locally pre-trained image recognition model; after the image recognition model outputs the recognition result, the core control module drives a stepper motor to move the probe along a preset path and adjust the angle between the probe and the component surface; equipment anomaly logs are stored in flash memory and uploaded to the cloud computing node via an Ethernet interface; the cloud computing node analyzes and optimizes the linkage control parameters based on the received log data, and the optimized parameters are verified and then sent back to the core control module.

[0045] In a specific embodiment, the equipment status monitoring module uses high-precision analog sensors. These sensors have low signal drift and strong anti-electromagnetic interference capabilities, accurately capturing subtle changes in temperature, voltage, and probe pressure to ensure the accuracy of the raw data. The acquired signals are processed by an RC low-pass filter circuit, effectively filtering high-frequency noise from on-site construction machinery and ensuring signal purity. The primary and backup sensors are of the same specifications to ensure consistent performance parameters. The preset deviation threshold is set according to the sensor's factory calibration report. When the deviation between the primary sensor data and the historical average exceeds this threshold, the core control module quickly activates the backup sensor and disconnects the primary sensor link through a hardware trigger mechanism, preventing invalid data from the faulty sensor.

[0046] The analog-to-digital converter is a high-precision model, capable of accurately converting analog temperature signals into digital signals. The value needs to be calculated in advance based on the material properties of the components. Reflecting the magnitude of temperature change on the component surface, through and The linear correlation allows the probe pressure to adjust in real time according to changes in component temperature, ensuring stable contact between the probe and the component surface and reducing the impact of temperature changes on detection accuracy. The safety threshold is set with reference to civil engineering testing specifications. When the threshold is exceeded, an audible and visual alarm emits a warning light and a buzzer, while the sampling frequency of the data acquisition chip is correspondingly increased, improving the data acquisition density under abnormal conditions and facilitating the comprehensive capture of anomaly information.

[0047] The dual-path air cleaning module first uses high-pressure gas to blow away surface dust from components, then uses negative pressure to adsorb residual impurities, preventing surface stains from affecting detection accuracy. The industrial camera supports automatic exposure adjustment, adapting to complex lighting conditions in both laboratories and outdoors. Acquired images are compressed to grayscale and smoothed with Gaussian filtering before being input into a locally pre-trained image recognition model. This model accurately identifies features such as cracks and holes on component surfaces. The core control module drives a stepper motor based on the recognition results, moving the probe along a preset path with high angle adjustment precision, ensuring the probe is always aligned with the target area. Equipment anomaly logs are stored in industrial-grade flash memory according to a fixed directory structure and automatically uploaded to a cloud computing node via Ethernet during system idle periods. The cloud analyzes the fault types and frequencies in the logs to optimize the linkage control parameters. The optimized parameters undergo multiple verifications before being distributed, continuously improving module operational stability.

[0048] Furthermore, such as Figure 3 As shown, the intelligent model conversion and association module includes a format parsing module, a semantic segmentation module, and a model matching module;

[0049] The format parsing module imports the non-parametric model, extracts geometric topology data, and removes redundant vertices; the semantic segmentation module extracts the contour, size, and spatial location features of components through a convolutional neural network; the extracted feature data is encoded and then used to generate a structured BIM model through the BIM model conversion interface.

[0050] The intelligent model conversion and association module uses an association rule mining algorithm to traverse the attribute information of the detection data and the structured BIM model to establish associations. The intelligent model conversion and association module also includes a local risk assessment module, which calculates the risk level of components based on the high-frequency association rules output by the association rule mining algorithm, and uses the risk level calculation results to update the weight coefficients of the rules in the association rule mining algorithm. The intelligent model conversion and association module has built-in conversion plugins that support multiple formats.

[0051] The verification results output by the experimental monitoring simulation verification module are transmitted to the model matching module. The model matching module corrects the mapping relationship between the structured BIM model and the detection data based on the received verification results. The batch model conversion task or historical data optimization task is determined by the local task scheduling module based on the local computing resource utilization rate. When the local computing resource utilization rate exceeds the preset threshold, the task is uploaded to the cloud computing node for parallel processing through the cloud computing interaction module, and the processing result is returned through an encrypted link.

[0052] In specific embodiments, the format parsing module supports importing various common non-parametric model formats. When extracting geometric topology data, a greedy algorithm is used to traverse vertex connections, eliminating duplicates or redundant vertices that have minimal impact on the model shape. This reduces the amount of model data without losing core features, saving computational resources for detection processing. The semantic segmentation module uses a convolutional neural network. This network extracts features through encoder downsampling and recovers spatial information through decoder upsampling. It has high accuracy in component feature extraction tasks, accurately capturing the component's contour boundaries, actual dimensions, and three-dimensional spatial coordinates, providing accurate feature support for the generation of structured BIM models. After the extracted feature data is encoded, a structured BIM model is generated through a general BIM model conversion interface, ensuring that the model has good compatibility and editability. The association rule mining algorithm traverses parameters such as strength and moisture content in the detection data and attribute information such as component number, material type, and design parameters in the BIM model, establishing a one-to-one correspondence, allowing the data and model to correspond.

[0053] The local risk assessment module calculates risk levels based on high-frequency association rules output by the algorithm, combined with component design safety factors. Higher risk levels correspond to higher rule weights, ensuring the data-model correlation better aligns with actual engineering risk concerns. The module includes built-in conversion plugins supporting various common formats, directly adapting to model files exported from different design software without requiring additional format conversion tools, thus improving ease of use. The model matching module, after receiving verification results, iteratively optimizes the algorithm to correct the data-model mapping, eliminating deviations during data transmission or model conversion to ensure accurate association. The local task scheduling module presets resource usage thresholds and monitors CPU and memory usage in real time to determine resource status. When either indicator exceeds the threshold, resource-intensive batch tasks are uploaded to cloud computing nodes via an encrypted link. The cloud employs a multi-threaded parallel processing architecture, splitting tasks across multiple computing nodes for simultaneous execution. Processed data is returned via the same encrypted link, ensuring data transmission security and significantly reducing processing time.

[0054] Furthermore, the dynamic threshold decision module incorporates a local time-series prediction model, which is a long short-term memory network. This model uses historical detection data and environmental parameters as training samples, including temperature and humidity collected by digital sensors. The local time-series prediction model outputs real-time safety threshold predictions. These predictions are compared with the fused data in a comparator. When the fused data exceeds the range defined by the safety threshold prediction, an alarm command is triggered. This alarm command is transmitted to the BIM display module. The BIM display module adds a risk level icon to the corresponding coordinate position in the structured BIM model based on the component identifier associated with the detection data. The threshold adjustment log is encrypted and stored in an encrypted storage device. When performing local threshold adjustments, historical log data stored on cloud computing nodes is retrieved, processed, and input into the local time-series prediction model to optimize the weight parameters.

[0055] In a specific embodiment, the local time-series prediction model uses a Long Short-Term Memory (LSTM) network. This network effectively solves the gradient vanishing problem of traditional recurrent neural networks through a gating mechanism, accurately capturing the time-series change patterns of historical detection data. It is suitable for scenarios where data changes slowly over time in civil engineering inspections. Temperature and humidity are selected as environmental parameters because these two factors have the most direct and significant impact on component inspection results. Digital sensors offer high measurement accuracy, providing high-quality sample data for model training, making the prediction results more realistic. Model training employs a sliding window method, using recent inspection data as the training set and periodically updating the samples to ensure the prediction model can adapt to changes in the environment and component status. The model outputs real-time safety threshold predictions, meeting the timeliness requirements of real-time detection. The comparator uses a high-speed hardware comparison circuit with fast response speed, enabling timely judgment of whether the fused data exceeds the range of the safety threshold prediction, avoiding delayed alarms. After an alarm command is triggered, the BIM display module adds risk level icons of different colors to the corresponding coordinate positions in the structured BIM model. Staff can intuitively locate risky components through the model without having to check them one by one. The threshold adjustment logs are encrypted using an encryption algorithm and stored in a hardware encryption chip. This chip is tamper-proof and tamper-proof, effectively protecting the security of the log data. When performing local threshold adjustments, historical log data stored in the cloud computing node is retrieved via an interface. After data cleaning to remove outliers and normalization to unify dimensions, the data is input into the local time-series prediction model. The gradient descent method is used to optimize the model's weight parameters, making the security threshold prediction more closely reflect the changes in the actual detection scenario and improving the accuracy of threshold determination.

[0056] Furthermore, such as Figure 4 As shown, the multi-source data fusion verification module includes a multi-channel interface and a data preprocessing module; the multi-channel interface receives sensor data, image data, and laboratory data; the data preprocessing module performs outlier removal on the received data, and the outlier determination rule is: when... satisfy or If it is, then it will be removed. Data point values, The mean of the data samples. The standard deviation of the data sample;

[0057] The multi-source data fusion verification module uses a weighted fusion algorithm to integrate the preprocessed multi-source data. The weighted fusion formula is as follows:

[0058] ;

[0059] in, The merged data values For sensor data weighting coefficients, These are the sensor data values ​​after mean filling and normalization. These are the image data weighting coefficients. The values ​​are the image data after grayscale conversion and edge enhancement. For laboratory data weighting coefficients, The values ​​are laboratory data after anomaly removal and unit standardization, and the weighting coefficients satisfy the following: And allocate them based on data credibility scores;

[0060] The data credibility score is calculated based on the data integrity verification and collection accuracy calibration results. Abnormal data that is removed is stored by associating the collection device number, timestamp, and data serial number.

[0061] In specific embodiments, the multi-channel interface integrates multiple interface types, enabling simultaneous adaptation to the data output needs of different devices such as sensors, industrial cameras, and laboratory testing instruments. It supports parallel data reception, significantly improving data reception efficiency and avoiding data congestion. The data preprocessing module employs criteria to remove outliers. These criteria are based on the normal distribution characteristics of the data, representing the specific value of a single data point, reflecting the central tendency of the data sample, and demonstrating the degree of data dispersion. The combination of these three factors accurately defines the normal fluctuation range of the data, effectively filtering out extreme deviations caused by sudden equipment failures, instantaneous environmental interference, etc., ensuring the rationality of the processed detection data. In the weighted fusion algorithm, the data reliability score is calculated by combining the data integrity and acquisition accuracy calibration results. Sensor data, directly acquired at the component site, reflects the true state of the component and has the highest reliability, so it is usually set to a high value. Image data, after multiple processing steps, loses some detailed information and has the next highest reliability, so it is set to an intermediate value. Laboratory data, acquired in a controlled environment, has high accuracy but differs from the actual on-site conditions, so it is set to a lower value. The weight coefficients can be dynamically adjusted according to the characteristics of different detection items, but always meet the constraint that the sum of the three is 1, ensuring the scientific nature of the fusion calculation. When sensor data has missing values, the mean of the same batch of data is used to fill them, and then normalization is performed to eliminate the influence of dimensions. Image data is extracted after grayscale conversion and edge enhancement to highlight features. After outliers are removed from laboratory data, it is uniformly converted to the International System of Units (SI) to ensure data consistency. The removed outlier data is recorded in detail with the acquisition device number, a timestamp accurate to milliseconds, and a unique data serial number, and is also associated with the temperature and humidity environmental parameters at the time of acquisition, providing a complete basis for tracing the cause of the anomaly.

[0062] Furthermore, the experimental monitoring and simulation verification module includes a data integration module and a simulation calculation module. The simulation calculation module generates simulation data. The data integration module extracts local experimental data, on-site monitoring data, and simulation data, and classifies and associates the three types of data according to the component identifier. The experimental monitoring and simulation verification module compares the classified and associated data pairwise through a cross-verification process and calculates the deviation value. When the deviation value is less than the preset allowable deviation threshold, the experimental monitoring and simulation verification module generates an adjustment instruction. The adjustment instruction is transmitted to the adaptive detection execution module via a communication link. The adjustment instruction includes calibration parameters and sampling frequency adjustment coefficients. For components that fail verification, the experimental monitoring and simulation verification module adds a mark with an encrypted digital watermark to the corresponding component in the structured BIM model. The mark is associated with a deviation cause code, which points to a preset cause library containing sensor drift, environmental interference, and equipment error. When the local computing resource occupancy rate exceeds a preset threshold, the large-scale simulation task executed by the simulation calculation module is processed in parallel by calling the computing resources of cloud computing nodes.

[0063] In a specific embodiment, the simulation calculation module uses finite element analysis software to generate simulation data. Based on the component's design drawings and material mechanics parameters, a finite element model is established. Loads and boundary conditions consistent with actual testing are applied to generate test data for the component under ideal conditions, providing a reliable reference benchmark for actual data verification. The data integration module categorizes and associates local experimental data, field monitoring data, and simulation data according to component identification, ensuring that all three types of data pertain to the same test item for the same component, thus avoiding data confusion.

[0064] The cross-verification process employs a pairwise comparison method. First, local experimental data is compared with field monitoring data. Then, the average of the two is compared with simulated data. The deviation value is calculated using the mean square error formula, which effectively reflects the overall deviation between the two sets of data, making the verification results more convincing. The allowable deviation threshold is set according to the characteristics of the testing item and complies with the requirements of civil engineering testing specifications. When the deviation value is less than the allowable deviation threshold, the generated adjustment command is transmitted to the adaptive testing execution module via a communication link. Calibration parameters are used to correct the sensor's system error. The sampling frequency adjustment coefficient is dynamically set according to the deviation magnitude; the smaller the deviation, the closer the coefficient is to 1, reducing unnecessary sampling frequency fluctuations and balancing testing accuracy and resource consumption. For components that fail verification, an invisible encrypted digital watermark is used for marking. This watermark is deeply bound to the component identification and cannot be tampered with or removed. The deviation cause code uses a two-digit code, corresponding to common causes such as sensor drift, environmental interference, and equipment error. Staff can quickly query a preset cause database using the code to preliminarily determine the possible source of the deviation, improving problem-solving efficiency. When the local computing resource utilization rate exceeds the preset threshold, the large-scale simulation task calls cloud computing power through the cloud computing interaction module and adopts multi-threaded parallel processing technology to split the complex task into multiple computing nodes for synchronous execution, which greatly shortens the simulation computing time.

[0065] Furthermore, the visualization results presentation module includes a lightweight BIM module; the lightweight BIM module performs surface merging and vertex deletion on the structured BIM model through a mesh simplification algorithm to obtain a lightweight BIM model; the lightweight BIM model is displayed in real time on the local terminal, and the data update delay time of the lightweight BIM model is less than one-tenth of the inspection sampling cycle; when a query is triggered by touch or click operation, the visualization results presentation module calls the inspection data, operation logs and verification results associated with the selected component and presents them on the local terminal; the visualization results presentation module drives the augmented reality device to collect real scene images, and after calibration, overlays and projects the lightweight BIM model and anomaly markers onto the real scene.

[0066] In a specific embodiment, the lightweight BIM module employs the Quadric Error Metrics (QEM) mesh simplification algorithm. This algorithm calculates the quadratic error matrix for each vertex, prioritizing the deletion of vertices with smaller errors while merging adjacent surfaces. This significantly reduces the model's data volume and improves loading and rendering speed, ensuring smooth display on the local terminal, while preserving key model features. The local terminal can be an industrial tablet or desktop computer, supporting touch and mouse click operations. After setting the sampling period, the data update delay of the lightweight BIM model is controlled within one-tenth of the sampling period, ensuring that staff see real-time updated inspection status and avoiding misjudgments due to delays. When a staff member touches or clicks on a component in the model, the visualization results module quickly retrieves the corresponding inspection data, operation logs, and verification results through an internal data association interface. Specific values ​​are displayed in tabular form, and data trends are shown in line graphs, facilitating detailed data viewing for staff. The augmented reality equipment uses industrial-grade devices with high-definition image acquisition capabilities and real-time projection functions. It performs spatial calibration by recognizing preset optical landmarks on site, achieving high calibration accuracy. After calibration, the lightweight BIM model and anomaly markers are accurately superimposed and projected onto the real scene. On-site staff can intuitively see the correspondence between the virtual model of the component and its actual location, quickly locate abnormal parts, and improve inspection efficiency without repeatedly comparing drawings and the site.

[0067] Furthermore, the intelligent report generation module includes a data extraction and integration module. This module extracts detection data, verification results, visualization markers, and anomaly logs from the adaptive detection execution module, experimental monitoring simulation verification module, and visualization result presentation module, and completes data association and integration according to component identifiers. The intelligent report generation module calls pre-stored standardized report templates and automatically fills the integrated data into the corresponding positions of the standardized report templates. Modifications, exports, and annotations to the standardized report templates are all recorded by the intelligent report generation module, and the generated operation log includes the operator's identity, timestamp, and operation details. During cross-regional collaboration, annotations added to the report by collaborating parties are synchronized via cloud computing nodes, and all annotation information is stored using blockchain technology. The standardized report templates are updated from cloud computing nodes through an interface.

[0068] In a specific embodiment, the data extraction and integration module extracts test data, verification results, visualization markers, and anomaly logs from relevant modules via an internal high-speed data bus. These are then integrated according to component identifiers to form a complete data package containing basic component information, test process data, result judgments, and anomaly descriptions, ensuring the comprehensiveness and relevance of the report data. The standardized report template strictly adheres to industry standards, including fixed sections such as project overview, testing basis, test results, and conclusions / recommendations. The data extraction and integration module automatically fills the integrated data into the corresponding positions using template variable mapping, eliminating the need for manual input and reducing human error. Report modifications, exports, and annotations are all recorded in real-time by the system. The operation log includes the operator's account, name, timestamp accurate to the second, and specific operation content, facilitating traceability of the operation process. In cross-regional collaboration scenarios, staff from all parties can view the report online simultaneously. Annotations added by collaborating parties are synchronized to all viewing terminals in real-time via cloud computing nodes. All annotation information is uploaded to the blockchain for notarization, and each annotation generates a unique hash value, ensuring that the annotation content is tamper-proof and traceable. The standardized report template is updated from the cloud computing node via an interface. The cloud regularly updates the template version according to the latest industry standards and testing requirements. The module automatically detects updates and downloads and installs them to ensure that the report format always conforms to the latest specifications.

[0069] It should be noted that the combination of the technical features in this case is not limited to the combination methods described in the claims of this case or the combination methods described in the specific embodiments. All technical features described in this case can be freely combined or combined in any way, unless they contradict each other.

[0070] It should also be noted that the embodiments listed above are merely specific embodiments of the present invention. Obviously, the present invention is not limited to the above embodiments, and similar changes or modifications made thereto are those that can be directly derived or easily conceived by those skilled in the art from the content disclosed in the present invention, and should all fall within the protection scope of the present invention.

Claims

1. A civil engineering experimental testing system based on cloud computing technology, characterized in that, It includes an adaptive detection execution module, an intelligent model conversion and association module, a dynamic threshold decision module, a multi-source data fusion verification module, an experimental monitoring and simulation verification module, a visualization result presentation module, an intelligent report generation module, and a cloud computing interaction module; The adaptive detection execution module collects detection data and transmits it to the multi-source data fusion verification module; the multi-source data fusion verification module outputs fused data and transmits it to the dynamic threshold decision module and the experimental monitoring simulation verification module, respectively. The dynamic threshold decision module generates threshold parameters, and the experimental monitoring simulation verification module generates verification results; the threshold parameters and verification results are fed back to the adaptive detection execution module. The verification results are transmitted to the intelligent model conversion and association module; The intelligent model conversion and association module establishes the association between the detection data and the structured BIM model, and transmits it to the visualization result presentation module. The visualization results presentation module transmits the detection information and anomaly markers to the report intelligent generation module; The intelligent report generation module generates standardized reports; The cloud computing interaction module connects the local computing node and the cloud computing node, and feeds back the processing results from the cloud computing node to the adaptive detection execution module, the intelligent model conversion and association module, and the intelligent report generation module.

2. The civil engineering experimental testing system based on cloud computing technology according to claim 1, characterized in that: The adaptive detection execution module includes an equipment status monitoring module and a core control module. The equipment status monitoring module collects temperature, voltage, and probe pressure signals through analog sensors. The collected signals are processed by a filtering circuit and then transmitted to the core control module. The equipment status monitoring module determines whether the main sensor is faulty by comparing the main sensor data with a preset deviation threshold. After determining the fault, the core control module activates the backup sensor and disconnects the data transmission link of the main sensor. The component surface temperature signal is converted by an analog-to-digital converter and then transmitted to the pressure adjustment module. The probe pressure adjustment is calculated as follows: ； in, This represents the change in probe pressure. The pressure-temperature coupling coefficient is related to the elastic modulus of the component material and the contact area of ​​the probe. This refers to the change in surface temperature of the component. When the temperature, voltage, or probe pressure signal value exceeds the safety threshold preset in the core control module, an audible and visual alarm is triggered and the sampling frequency of the data acquisition chip is adjusted. The adaptive detection execution module also includes a dual-path air cleaning module for blowing and adsorption cleaning of the component surface; an industrial camera acquires images of the component surface, which are then processed by grayscale conversion and Gaussian filtering before being input into a locally pre-trained image recognition model; after the image recognition model outputs the recognition result, the core control module drives a stepper motor to move the probe along a preset path and adjust the angle between the probe and the component surface; equipment anomaly logs are stored in flash memory and uploaded to a cloud computing node via an Ethernet interface; the cloud computing node analyzes and optimizes the linkage control parameters based on the received log data, and the optimized parameters are verified and then sent back to the core control module.

3. The civil engineering experimental testing system based on cloud computing technology according to claim 2, characterized in that: The intelligent model conversion and association module includes a format parsing module, a semantic segmentation module, and a model matching module; The format parsing module imports the non-parametric model, extracts geometric topology data, and removes redundant vertices; the semantic segmentation module extracts the contour, size, and spatial location features of the components through a convolutional neural network; the extracted feature data is encoded and then used to generate a structured BIM model through the BIM model conversion interface. The intelligent model conversion and association module uses an association rule mining algorithm to traverse the attribute information of the detection data and the structured BIM model to establish an association. The intelligent model conversion and association module also includes a local risk assessment module, which calculates the risk level of the component based on the high-frequency association rules output by the association rule mining algorithm, and uses the risk level calculation results to update the weight coefficients of the rules in the association rule mining algorithm. The intelligent model conversion and association module has built-in conversion plugins that support multiple formats; The verification results output by the experimental monitoring simulation verification module are transmitted to the model matching module. The model matching module corrects the mapping relationship between the structured BIM model and the detection data based on the received verification results. The batch model conversion task or historical data optimization task is determined by the local task scheduling module based on the local computing resource utilization rate. When the local computing resource utilization rate exceeds the preset threshold, the task is uploaded to the cloud computing node for parallel processing through the cloud computing interaction module, and the processing result is returned through an encrypted link.

4. The civil engineering experimental testing system based on cloud computing technology according to claim 3, characterized in that: The dynamic threshold decision module incorporates a local time-series prediction model, which is a long short-term memory network. This model uses historical detection data and environmental parameters as training samples, including temperature and humidity collected by digital sensors. The model outputs a real-time safety threshold prediction value. This prediction value is compared with the fused data in a comparator. When the fused data exceeds the range defined by the safety threshold prediction value, an alarm command is triggered. This alarm command is transmitted to the BIM display module. The BIM display module adds a risk level icon to the corresponding coordinate position in the structured BIM model based on the component identifier associated with the detection data. Threshold adjustment logs are encrypted and stored in encrypted storage. When performing local threshold adjustments, historical log data stored on cloud computing nodes is retrieved, processed, and then input into the local time series prediction model to optimize weight parameters.

5. The civil engineering experimental testing system based on cloud computing technology according to claim 4, characterized in that: The multi-source data fusion verification module includes a multi-channel interface and a data preprocessing module; the multi-channel interface receives sensor data, image data, and laboratory data. The data preprocessing module performs outlier removal on the received data. The outlier determination rule is: when... satisfy or If it is, then it will be removed. Data point values, The mean of the data samples. The standard deviation of the data sample; The multi-source data fusion and verification module uses a weighted fusion algorithm to integrate the preprocessed multi-source data. The weighted fusion formula is as follows: ； in, The merged data values For sensor data weighting coefficients, These are the sensor data values ​​after mean filling and normalization. These are the image data weighting coefficients. The values ​​are the image data after grayscale conversion and edge enhancement. These are the weighting coefficients for laboratory data. The weighting coefficients are the values ​​of laboratory data after anomaly removal and unit standardization, and satisfy the following conditions: And allocate them based on data credibility scores; The data credibility score is calculated based on the data integrity verification and collection accuracy calibration results. Abnormal data that is removed is stored by associating the collection device number, timestamp, and data serial number.

6. The civil engineering experimental testing system based on cloud computing technology as described in claim 5, characterized in that: The experimental monitoring simulation verification module includes a data integration module and a simulation calculation module. The simulation calculation module generates simulation data. The data integration module extracts local experimental data, on-site monitoring data, and simulation data, and classifies and associates the three types of data according to component identification. The experimental monitoring simulation verification module compares the classified and associated data pairwise through a cross-verification process and calculates the deviation value. When the deviation value is less than a preset allowable deviation threshold, the experimental monitoring simulation verification module generates an adjustment instruction. The adjustment instruction is transmitted to the adaptive detection execution module via a communication link. The adjustment instruction includes calibration parameters and sampling frequency adjustment coefficients. For components that fail verification, the experimental monitoring simulation verification module adds a mark with an encrypted digital watermark to the corresponding component in the structured BIM model. The mark is associated with a deviation cause code, which points to a preset cause library containing sensor drift, environmental interference, and equipment errors. When the local computing resource occupancy rate exceeds a preset threshold, the large-scale simulation task executed by the simulation calculation module is processed in parallel by calling the computing resources of the cloud computing node.

7. A civil engineering experimental testing system based on cloud computing technology according to claim 6, characterized in that: The visualization result presentation module includes a lightweight BIM module; the lightweight BIM module performs surface merging and vertex deletion on the structured BIM model through a mesh simplification algorithm to obtain a lightweight BIM model; the local terminal displays the lightweight BIM model in real time, and the data update delay time of the lightweight BIM model is less than one-tenth of the detection sampling period; When a query is triggered by touch or click, the visualization result presentation module calls the detection data, operation logs and verification results associated with the selected component and presents them on the local terminal; the visualization result presentation module drives the augmented reality device to collect real scene images, and after calibration, the lightweight BIM model and anomaly markers are superimposed and projected onto the real scene.

8. A civil engineering experimental testing system based on cloud computing technology according to claim 7, characterized in that: The intelligent report generation module includes a data extraction and integration module; the data extraction and integration module extracts detection data, verification results, visualization tags and abnormal logs from the adaptive detection execution module, experimental monitoring simulation verification module and visualization result presentation module, and completes data association and integration according to component identifiers; the intelligent report generation module calls a pre-stored standardized report template and automatically fills the integrated data into the corresponding position of the standardized report template; Modifications, exports, or annotations to standardized report templates are all recorded by the intelligent report generation module, and the generated operation log includes the operator's identity, timestamp, and operation details. When collaborating across regions, the annotations added to the report by the collaborating parties are synchronized through the cloud computing nodes, and all annotation information is stored using blockchain technology; the standardized report template is updated from the cloud computing nodes through an interface.

Images