Building engineering safety monitoring method and system based on machine learning

Abstract

The invention relates to the technical field of constructional engineering safety monitoring, and discloses a constructional engineering safety monitoring method and system based on machine learning, and the method comprises the steps: obtaining an original detection data set; according to the original detection data set, adopting a noise reduction algorithm to obtain a monitoring data set; according to the monitoring data set, a clustering algorithm is adopted to divide a building structure into a plurality of areas, and a high-risk area is obtained; according to the high-risk area, time sequence data are extracted, a time sequence analysis method is adopted to calculate a quality index change trend, and a trend analysis result is obtained; according to the trend analysis result, analyzing quality index relevance between different areas through a relevance mining algorithm to obtain a potential chain reaction risk assessment report; and inputting the potential chain reaction risk assessment report into a pre-trained building engineering safety monitoring model to obtain a building structure health trend report. According to the method, self-adaptive building quality safety dynamic global analysis can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of building engineering safety monitoring, and particularly to a building engineering safety monitoring method and system based on machine learning. Background Art

[0002] At present, the field of building engineering safety monitoring faces the dual challenges of a sharp increase in data scale and dynamic risk evolution. With the popularization of complex structures such as super high-rise buildings and long-span bridges, the traditional monitoring methods relying on manual inspections and regular spot checks can no longer meet the requirements of real-time and accuracy. In the prior art, the data volume of vibration, settlement, cracks, etc. collected by monitoring devices has increased exponentially, but due to the limitations of data processing capabilities and the lag of analysis methods, a large amount of data has not been effectively mined and utilized. Especially in the construction stage or extreme environments, the mechanical state changes of building structures have strong time-varying and non-linear characteristics, and the current technology lacks the ability to dynamically track the chain reactions caused by local minor defects, resulting in potential risks being difficult to warn in time.

[0003] In an existing technology, building engineering safety monitoring usually adopts a fixed sensor network combined with static threshold alarm. By deploying sensors such as inclinometers and crack meters to collect index data such as the displacement and crack width of building structures, and setting a unified threshold based on experience. Although such methods can achieve basic anomaly detection, they have significant defects. For example, the static threshold lacks adaptability, and the sensitivities of different building structures to the same index vary greatly. The fixed threshold is prone to false alarms or missed alarms. At the same time, the sensor data is only analyzed independently at a single point, without considering the mechanical correlations between regions, and can only identify the current abnormal state, unable to predict the risk evolution path through historical data modeling. The monitoring models in the prior art cannot dynamically adjust the analysis logic according to the building structure characteristics and environmental factors, resulting in low monitoring accuracy for special-shaped structures, ignoring the relevance between local risks and the overall safety state, being difficult to warn of the butterfly effect events caused by small defects leading to large accidents, lacking the ability to continuously model the structural health trend, unable to provide decision support for preventive maintenance, and unable to perform global dynamic safety detection on building engineering.

[0004] There is a problem in the prior art that it is difficult to achieve adaptive dynamic global analysis of building quality and safety. Summary of the Invention

[0005] The present invention provides a building engineering safety monitoring method and system based on machine learning to achieve adaptive dynamic global analysis of building quality and safety.

[0006] In a first aspect, to solve the above technical problems, the present invention provides a building engineering safety monitoring method based on machine learning, including: Obtaining an original detection data set containing building dynamic information and engineering quality indicators; Based on the original detection dataset, a noise reduction algorithm is used to remove environmental noise and measurement errors, obtaining a denoised monitoring dataset; Based on the monitoring dataset, a clustering algorithm is used to divide the building structure into several regions and perform risk analysis, obtaining high-risk regions; Based on the high-risk regions, time series data of key quality indicators for building engineering safety are extracted, and a time series analysis method is used to calculate the change trend of the quality indicators, obtaining a trend analysis result; Based on the trend analysis result, diagnosis is performed in combination with a preset regional quality threshold, and the correlation of quality indicators between different regions is analyzed through an association mining algorithm, obtaining a potential chain reaction risk assessment report; The potential chain reaction risk assessment report is input into a pre-trained building engineering safety monitoring model to obtain a building structure health trend report.

[0007] In an alternative implementation, the obtaining of the original detection dataset including building dynamic information and engineering quality indicators includes: Obtain the dynamic information of the building structure, the foundation settlement data and the wall crack monitoring data at a preset acquisition frequency; Based on the dynamic information, the foundation settlement data and the wall crack monitoring data, analyze and extract the quality indicator data of building engineering safety, obtaining an initial monitoring dataset; Based on the initial monitoring dataset, perform integrity analysis and use an interpolation algorithm to fill in missing values, obtaining a complete detection dataset; Based on the complete detection dataset, perform data format standardization processing to obtain the original detection dataset including building dynamic information and engineering quality indicators.

[0008] In an alternative implementation, the based on the original detection dataset, using a noise reduction algorithm to remove environmental noise and measurement errors, obtaining a denoised monitoring dataset, includes: Based on the original detection dataset, identify and filter out the outliers and invalid values in the original detection dataset, obtaining a preliminary screening dataset; Based on the preliminary screening dataset, analyze the statistical characteristics of data deviation and establish a linear regression model to perform linear trend correction on the preliminary screening dataset, obtaining a corrected dataset; Based on the corrected dataset, use a moving average algorithm to smooth the fluctuations of the corrected dataset, and calculate the local average value through a moving window, obtaining a smoothed dataset; Based on the smoothed dataset, perform time series decomposition, separate the trend term and the residual term, and retain the long-term trend data that conforms to engineering physical laws, obtaining a denoised monitoring dataset.

[0009] In an alternative embodiment, based on the monitoring data set, a clustering algorithm is used to divide the building structure into several regions, and risk analysis is performed to obtain high-risk regions, including: Based on the monitoring data set, various features of the building structure are extracted to generate a multi-dimensional feature matrix; Based on the multi-dimensional feature matrix, standardization processing is performed for different feature dimensions to obtain a standardized feature matrix; Based on the standardized feature matrix, a clustering algorithm is used for region division to obtain building safety region categories; Based on the building safety region categories, statistical analysis is performed to calculate the dispersion degree of data distribution, and regions with a dispersion degree exceeding a preset risk threshold are marked as high-risk regions.

[0010] In an alternative embodiment, based on the high-risk regions, time series data of key quality indicators for building engineering safety are extracted, and a time series analysis method is used to calculate the change trend of the quality indicators to obtain a trend analysis result, including: Based on the high-risk regions, time series data of quality indicators are extracted to obtain a time series data set; Based on the time series data set, a sliding window method is used for segmentation processing, and the mean value, variance, and extreme value of each time period are extracted to obtain risk time series statistical features; Based on the risk time series statistical features, the principal component analysis method is used to reduce the feature dimension to obtain a region feature matrix; Based on the region feature matrix, a time series prediction algorithm is used to calculate the change trend of the quality indicators of each region to obtain a trend analysis result.

[0011] In an alternative embodiment, based on the trend analysis result, diagnosis is performed in combination with a preset regional quality threshold, and the association between quality indicators of different regions is analyzed through an association mining algorithm to obtain a potential chain reaction risk assessment report, including: The trend analysis result is compared with a preset quality trend threshold. When the quality indicator of the trend analysis result is greater than the quality trend threshold, abnormal diagnosis is performed to obtain an abnormal diagnosis result; Based on the abnormal diagnosis result, quality indicator data of different regions are extracted for feature analysis to obtain a regional quality association matrix; Based on the regional quality association matrix, an association mining algorithm is used to calculate the association strength of quality indicators between different regions to obtain the potential chain reaction relationship between regions; According to the potential chain reaction relationship and combined with historical monitoring data, predict the risk diffusion path to obtain a potential chain reaction risk assessment report.

[0012] In an alternative embodiment, inputting the potential chain reaction risk assessment report into a pre-trained building engineering safety monitoring model to obtain a building structure health trend report, including: Using the historical potential chain reaction risk assessment report as the input and the historical building structure health trend report as the output, constructing an initial building engineering safety monitoring model and training it. When the number of training times is greater than or equal to the preset number of training times, it is determined that the training is completed, and a trained building engineering safety monitoring model is obtained; Inputting the potential chain reaction risk assessment report into the trained building engineering safety monitoring model to obtain a building structure health trend report.

[0013] In a second aspect, the present invention provides a building engineering safety monitoring system based on machine learning, including: A data acquisition module for acquiring Feature 1 and Feature 2; A data acquisition module for acquiring an original detection data set containing building dynamic information and engineering quality indicators; A noise reduction processing module for removing environmental noise and measurement errors from the original detection data set using a noise reduction algorithm to obtain a denoised monitoring data set; A risk division module for dividing the building structure into several regions using a clustering algorithm according to the monitoring data set and performing risk analysis to obtain high-risk regions; A trend analysis module for extracting time series data of key quality indicators for building engineering safety from the high-risk regions and calculating the change trend of the quality indicators using a time series analysis method to obtain a trend analysis result; A chain diagnosis module for diagnosing according to the trend analysis result in combination with a preset regional quality threshold and analyzing the correlation of quality indicators between different regions through an association mining algorithm to obtain a potential chain reaction risk assessment report; A result output module for inputting the potential chain reaction risk assessment report into a pre-trained building engineering safety monitoring model to obtain a building structure health trend report.

[0014] In a third aspect, the present invention further provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the method for building engineering safety monitoring based on machine learning described in any one of the above.

[0015] Fourthly, the present invention also provides a computer-readable storage medium, which includes a stored computer program. When the computer program runs, it controls the device where the computer-readable storage medium is located to execute the above-mentioned machine learning-based building engineering safety monitoring method according to any one of the above.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The present invention discloses a machine learning-based building engineering safety monitoring method, which includes obtaining an original detection data set containing building dynamic information and engineering quality indicators; removing environmental noise and measurement errors from the original detection data set by using a noise reduction algorithm to obtain a denoised monitoring data set; dividing the building structure into several regions by using a clustering algorithm according to the monitoring data set and performing risk analysis to obtain high-risk regions; extracting time series data of key quality indicators for building engineering safety according to the high-risk regions, and calculating the change trend of the quality indicators by using a time series analysis method to obtain a trend analysis result; diagnosing according to the trend analysis result in combination with a preset regional quality threshold, and analyzing the correlation of quality indicators between different regions by using an association mining algorithm to obtain a potential chain reaction risk assessment report; inputting the potential chain reaction risk assessment report into a pre-trained building engineering safety monitoring model to obtain a building structure health trend report.

[0017] The present invention constructs an adaptive dynamic global analysis system for building quality and safety step by step through the following steps: First, obtain the original detection data set containing building dynamic information and engineering quality indicators to provide a comprehensive data basis for subsequent analysis; Second, adopt a noise reduction algorithm to remove environmental noise and measurement errors, filter out invalid data caused by equipment failures or instantaneous interferences through outlier filtering, smooth short-term fluctuations by combining with a moving average algorithm, and use time series decomposition technology to separate the long-term trend term and the random residual term, retaining the denoised data reflecting the true structural response to ensure the physical consistency and temporal continuity of the monitoring data set; Then, use a clustering algorithm to divide the building structure into several regions, perform clustering analysis based on the denoised multi-dimensional feature matrix, dynamically divide the monitoring regions according to the data space distribution density, break through the rigid limitations of traditional manual preset partitioning, make the region division results adapt to the structural characteristics of different buildings, and at the same time calculate the data dispersion within each clustering cluster, mark the regions with dispersion exceeding the threshold as high-risk regions to achieve risk positioning; Subsequently, extract the time series data of the key quality indicators in the high-risk regions and conduct trend analysis, construct sliding window statistical features for core indicators such as settlement rate and crack expansion amount, reduce the dimension and eliminate feature redundancy through principal component analysis, then use the ARIMA model to fit the index change trend, dynamically adjust the prediction parameters by combining with Bayesian optimization, so that the trend analysis results can not only capture short-term mutations and reflect long-term evolution laws, but also introduce an elastic threshold mechanism to dynamically set the diagnostic threshold according to the historical data distribution and structural type to avoid the adaptability defects of fixed thresholds for heterogeneous structures; After that, analyze the correlation of quality indicators between regions through an association mining algorithm, construct a regional quality correlation matrix, use the Apriori algorithm to mine high-frequency association rules, verify the statistical significance of the association relationship by combining with Granger causality test, identify potential chain reaction paths, and quantify the risk conduction probability based on the historical accident case library to construct a risk evolution map; Finally, input the chain reaction risk assessment into the LSTM model to generate a health trend report, train the long short-term memory network with historical monitoring data, embed engineering physical constraints in the model loss function, update the input sequence iteratively through the sliding time window mechanism, predict the trend of key indicators in the future time period, and generate a health trend report containing risk hot spot distribution, trend comparison curves and maintenance suggestions by integrating the association mining results to form a closed-loop decision-making link. Thus, the present invention realizes an adaptive dynamic global analysis of building quality and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 FIG. is a schematic flow chart of a building engineering safety monitoring method based on machine learning provided by the first embodiment of the present invention; Figure 2 FIG. is a schematic structural diagram of a building engineering safety monitoring system based on machine learning provided by the second embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts shall fall within the protection scope of the present invention.

[0020] Referring to Figure 1 , the first embodiment of the present invention provides a building engineering safety monitoring method based on machine learning, including the following steps: S11. Obtain an original detection data set including building dynamic information and engineering quality indicators; S12. According to the original detection data set, use a noise reduction algorithm to remove environmental noise and measurement errors to obtain a denoised monitoring data set; S13. According to the monitoring data set, use a clustering algorithm to divide the building structure into several regions and perform risk analysis to obtain high-risk regions; S14. According to the high-risk regions, extract time series data of key quality indicators for building engineering safety, and use time series analysis methods to calculate the change trend of quality indicators to obtain a trend analysis result; S15. According to the trend analysis result, combine a preset regional quality threshold for diagnosis, and analyze the correlation of quality indicators between different regions through an association mining algorithm to obtain a potential chain reaction risk assessment report; S16. Input the potential chain reaction risk assessment report into a pre-trained building engineering safety monitoring model to obtain a building structure health trend report.

[0021] In step S11, it is necessary to obtain an original detection data set including building dynamic information and engineering quality indicators.

[0022] In one implementation, the obtaining of the original detection data set including building dynamic information and engineering quality indicators includes: Obtain the dynamic information of the building structure, the foundation settlement data and the wall crack monitoring data at a preset acquisition frequency; according to the dynamic information, the foundation settlement data and the wall crack monitoring data, analyze and extract the quality index data of building engineering safety to obtain an initial monitoring data set; according to the initial monitoring data set, perform integrity analysis and use an interpolation algorithm to fill in missing values to obtain a complete detection data set; according to the complete detection data set, perform data format standardization processing to obtain an original detection data set including building dynamic information and engineering quality indicators.

[0023] In step S12, it is necessary to use a noise reduction algorithm to remove environmental noise and measurement errors based on the original detection data set, so as to obtain a denoised monitoring data set.

[0024] In one implementation manner, the using a noise reduction algorithm to remove environmental noise and measurement errors based on the original detection data set, so as to obtain a denoised monitoring data set includes: Based on the original detection data set, identify and filter outliers and invalid values in the original detection data set to obtain a preliminary screening data set; based on the preliminary screening data set, analyze the statistical characteristics of data deviation and establish a linear regression model to perform linear trend correction on the preliminary screening data set to obtain a corrected data set; based on the corrected data set, use a moving average algorithm to smooth the fluctuations of the corrected data set, calculate the local average through a moving window to obtain a smoothed data set; based on the smoothed data set, perform time series decomposition, separate the trend term and the residual term, and retain the long-term trend data that conforms to the engineering physical law to obtain a denoised monitoring data set.

[0025] It should be noted that the denoised monitoring data set is a high-quality data set obtained by performing multi-level noise reduction processing on the original detection data set. Its core feature is that it removes environmental noise and measurement errors and retains long-term trend data that reflects the true response of the building structure. Specifically, the denoised monitoring data set is obtained through the following steps: first, identify and filter outliers and invalid values ​​from the original detection data set to generate a preliminary screening data set; second, analyze the statistical characteristics of data deviations based on the preliminary screening data set, establish a linear regression model to correct the data trend, eliminate systematic errors, and obtain a corrected data set; then, use a sliding average algorithm to smooth the fluctuations of the corrected data set, calculate the local average through a sliding window, and use a dynamic window width mechanism to set the window span to 2N+1 continuous sampling points based on the main frequency characteristics of the structural vibration, where the N value is based on the monitored object. The natural frequency is dynamically adjusted, and 10 sampling intervals are taken by default. The N points before and after the current data point are symmetrically selected to form a sliding window. The arithmetic mean of all data points in the window is calculated as the smoothing value of the current point. The window slides point by point along the time axis to complete the processing of the entire sequence. The mirror symmetry expansion method is used to supplement virtual data points at both ends of the data sequence to ensure boundary integrity and suppress short-term noise, and generate a smooth data set; finally, the smooth data set is decomposed into a time series, and the long-term trend term and random residual term are separated. The residual data that does not conform to the laws of engineering physics are eliminated, and the long-term trend data is retained as the denoised monitoring data set. The monitoring data set is used for regional division in subsequent analysis. The monitoring data set is used to extract various features of the building structure to generate a multidimensional feature matrix, and then the multidimensional feature matrix is ​​clustered to ensure that the regional division results reflect the real structural response. The denoised monitoring data set provides a high-quality data foundation for the dynamic global analysis of building quality and safety.

[0026] In step S13, it is necessary to divide the building structure into several areas using a clustering algorithm based on the monitoring data set, and perform risk analysis to obtain high-risk areas.

[0027] In one implementation, the building structure is divided into several areas using a clustering algorithm based on the monitoring data set, and risk analysis is performed to obtain high-risk areas, including: According to the monitoring data set, various features of the building structure are extracted to generate a multidimensional feature matrix; according to the multidimensional feature matrix, standardization is performed on the differences in different feature dimensions to obtain a standardized feature matrix; according to the standardized feature matrix, a clustering algorithm is used to divide the area to obtain the building safety area category; according to the building safety area category, statistical analysis is performed to calculate the degree of dispersion of the data distribution, and the area whose dispersion exceeds the preset risk threshold is marked as a high-risk area.

[0028] It should be noted that the high-risk area represents the area with abnormal mechanical properties identified through multi-dimensional data analysis of the building structure, and its judgment basis is that the statistical dispersion of the monitoring parameters exceeds the allowable range of the engineering safety code. The high-risk area is obtained through the following steps. First, according to the monitoring data set, a twelve-dimensional feature vector of the building structure is extracted, including the vibration fundamental frequency reflecting the overall stiffness characteristics, the maximum amplitude characterizing the structural deformation limit ability, the stress gradient revealing the internal stress distribution state, the crack propagation rate quantifying the development process of material damage, the environmental temperature and humidity coupling coefficient evaluating the interactive influence of the external environment, the displacement change rate monitoring the dynamic response sensitivity, the strain accumulation amount tracking the development degree of plastic deformation, the inclination angle deviation detecting the structural imbalance trend, the material fatigue coefficient characterizing the durability attenuation of components, the natural frequency offset diagnosing the stiffness degradation phenomenon, the load distribution uniformity evaluating the coordination of the bearing system, and the corrosion rate quantifying the deterioration process of metal components. Then, aiming at the dimension difference of the characteristic parameters, the Z-score standardization method is used to eliminate the dimension influence. By calculating the mean and standard deviation of the data of each characteristic dimension, the original data is converted into the distance from the mean value in units of the standard deviation, so that all characteristic values are distributed in the standard normal interval from negative three to positive three. Subsequently, an improved DBSCAN density clustering algorithm is adopted to dynamically optimize the neighborhood radius parameter and the minimum sample number threshold, and the clustering density parameter is automatically adjusted based on the silhouette coefficient evaluation index to effectively identify the spatial units with similar mechanical responses and overcome the algorithm defects of the traditional preset number of partitions. Finally, by calculating the coefficient of variation of each region, that is, the ratio of the standard deviation to the absolute value mean, when the coefficient of variation of the region exceeds the threshold set according to the building structure reliability design standard, the region is determined as a high-risk area, providing decision support for the structural safety assessment.

[0029] In step S14, it is necessary to extract the time series data of the key quality indicators for building engineering safety according to the high-risk area, and use the time series analysis method to calculate the change trend of the quality indicators to obtain the trend analysis result.

[0030] In one implementation, the extracting the time series data of the key quality indicators for building engineering safety according to the high-risk area, using the time series analysis method to calculate the change trend of the quality indicators, and obtaining the trend analysis result includes: According to the high-risk area, extract the time series data of the quality indicators to obtain a time series data set; according to the time series data set, use the sliding window method for segmented processing, extract the mean, variance and extreme values of each time period to obtain the risk time series statistical characteristics; according to the risk time series statistical characteristics, use the principal component analysis method to reduce the feature dimension to obtain a regional feature matrix; according to the regional feature matrix, use the time series prediction algorithm to calculate the change trend of the quality indicators of each region to obtain the trend analysis result.

[0031] It should be noted that the trend analysis result is a quantitative trend description obtained through multi-stage processing of the time series data of key quality indicators in high-risk areas. Its core content includes the evolution direction, change rate, and future trend prediction of quality indicators. Specifically, the process of obtaining the trend analysis result is as follows: First, extract the time series data of key quality indicators such as foundation settlement rate and wall crack width from high-risk areas to form a time series data set; then use the sliding window method to segment the time series data. For example, with a window length of 24 hours and a one-hour sliding interval, extract the mean value within each time window to reflect the overall level, the variance to characterize the fluctuation intensity, and the extreme value to identify abnormal peaks, generating risk time series statistical features; subsequently, use the principal component analysis method to reduce the dimension of multi-dimensional statistical features, eliminate redundant information between features, and extract the main change patterns to form a low-dimensional regional feature matrix; finally, based on the regional feature matrix, use a time series prediction algorithm such as the ARIMA model to calculate the future change trends of quality indicators in each region, and output an analysis result including trend slope, acceleration, and confidence interval. The trend analysis result is used for anomaly diagnosis and global risk assessment in subsequent processes: In the anomaly diagnosis stage, compare the trend slope with a preset threshold to identify high-risk indicators with continuous deterioration; in the association mining stage, analyze the mechanical interaction between regions by combining the trend evolution patterns of different regions; in the health trend report, generate a visualization curve by integrating the trend prediction result and historical data to guide engineering maintenance decisions. The trend analysis result realizes the conversion from data description to risk prediction, providing a core basis for dynamic global analysis.

[0032] In step S15, it is necessary to make a diagnosis based on the trend analysis result, combined with a preset regional quality threshold, and analyze the quality indicator correlation between different regions through an association mining algorithm to obtain a potential chain reaction risk assessment report.

[0033] In one implementation, the making a diagnosis based on the trend analysis result, combined with a preset regional quality threshold, and analyzing the quality indicator correlation between different regions through an association mining algorithm to obtain a potential chain reaction risk assessment report includes: Compare the trend analysis result with a preset quality trend threshold. When the quality indicator of the trend analysis result is greater than the quality trend threshold, perform anomaly diagnosis to obtain an anomaly diagnosis result; according to the anomaly diagnosis result, extract the quality indicator data of different regions and perform feature analysis to obtain a regional quality correlation matrix; according to the regional quality correlation matrix, use an association mining algorithm to calculate the quality indicator correlation intensity between different regions to obtain the potential chain reaction relationship between regions; according to the potential chain reaction relationship, combined with historical monitoring data, predict the risk diffusion path to obtain a potential chain reaction risk assessment report.

[0034] It should be noted that the potential chain reaction risk assessment report is a risk conduction path prediction document generated by analyzing the dynamic correlation of quality indicators between multiple regions of the building structure, which is used to identify the systemic risk diffusion pattern that may be caused by local structural abnormalities. The potential chain reaction risk assessment report is obtained through the following steps. First, the trend analysis results are compared with the preset quality trend threshold. When it is detected that the quality indicator of the trend analysis result is greater than the quality trend threshold, the abnormal diagnosis mechanism is triggered and an abnormal diagnosis result containing the abnormal location and severity is generated. When the quality indicator is less than the quality trend threshold, the data credibility verification process is automatically executed, and the reliability of the monitoring system is confirmed through sensor cross-checking and time series data stationarity analysis. At the same time, the structural health baseline parameters are updated and the subsequent monitoring frequency is optimized; then, according to the abnormal diagnosis results, the quality indicator data of different regions are extracted, including eight core parameters such as node stress conduction coefficient, vibration energy transfer ratio, deformation coordination factor, material damage accumulation, environmental coupling sensitivity, load redistribution index, stiffness degradation rate, and damping characteristic variability, to construct a spatial The regional quality association matrix of the topological relationship between the two regions is obtained; then the improved Apriori association rule mining algorithm is used to screen the frequent item sets by setting the minimum support threshold, and the confidence and improvement indicators are calculated to identify the strong association rules. The specific algorithm process includes: scanning the regional quality association matrix to generate candidate item sets, iteratively pruning low-support items based on the prior principle, verifying the reliability of the rules through confidence calculation, and finally extracting the strong association rule set such as "the abnormal increase in the stiffness degradation rate of region A will lead to an 85% increase in the probability of stress concentration in region B", and obtaining the potential chain reaction relationship between regions; then, combined with historical monitoring data, the directed graph model is used to construct the risk propagation network, and the risk diffusion probability calculation model based on the random walk algorithm is used to predict the risk diffusion path, and a potential chain reaction risk assessment report containing key conduction nodes, potential failure modes, and risk cascade probability matrix is ​​obtained. The potential chain reaction risk assessment report shows the risk thermal distribution, and links the structural health monitoring system to trigger targeted detection plans, providing data support for the formulation of preventive reinforcement strategies.

[0035] In step S16, the potential chain reaction risk assessment report needs to be input into a pre-trained construction engineering safety monitoring model to obtain a building structure health trend report.

[0036] In one implementation, the potential chain reaction risk assessment report is input into a pre-trained construction engineering safety monitoring model to obtain a building structure health trend report, including: Taking the historical potential chain reaction risk assessment report as the input and the historical building structure health trend report as the output, an initial building engineering safety monitoring model is constructed and trained. When the number of training times is greater than or equal to the preset number of training times, it is determined that the training is completed, and a trained building engineering safety monitoring model is obtained. The potential chain reaction risk assessment report is input into the trained building engineering safety monitoring model to obtain a building structure health trend report.

[0037] To facilitate the understanding of the present invention, some preferred embodiments of the present invention will be further described below.

[0038] The working process of the present invention will be described below by taking a relatively common scenario as an example. Please also refer to Figure 2 , which is Figure 1 a schematic diagram of the working scenario of the method.

[0039] During the construction stage of a large commercial complex building, due to the complex building structure and frequent changes in construction loads, quality indicators such as foundation settlement and wall cracks need to be monitored in real time to ensure construction safety. The working process of the present invention is as follows: First, through the sensor network deployed at key parts of the building structure, dynamic information such as foundation settlement, wall crack width, and structural vibration is collected in real time to form an original detection data set containing timestamps, sensor positions, and measurement values. Then, noise reduction processing is performed on the original detection data set to identify and filter out outliers and invalid values caused by equipment failures or environmental interferences. For example, data with sudden jumps in crack width caused by instantaneous temperature changes are excluded; systematic errors such as sensor zero drift are corrected through a linear regression model to eliminate data biases; a moving average algorithm is used to smooth short-term fluctuations. For example, instantaneous displacement fluctuations caused by construction machinery vibration are suppressed; finally, the long-term trend term and random residual term are separated through time series decomposition, and the denoised data reflecting the true response of the structure is retained to generate a denoised monitoring data set.

[0040] After obtaining the monitoring data set, a clustering algorithm is used to divide the building structure into regions. Based on multi-dimensional feature matrices such as vibration frequency, displacement gradient, and stress distribution, the difference in different feature dimensions is eliminated through standardization processing to generate a standardized feature matrix; the K-means clustering algorithm is used to divide the building structure into several region categories. For example, the core tube, podium, and basement are divided into different monitoring regions; the data distribution center point and dispersion degree of each region category are calculated, and regions with a dispersion degree exceeding the preset risk threshold are marked as high-risk regions. For example, the podium part with a significantly higher foundation settlement rate than other regions is marked as a high-risk region.

[0041] For high-risk areas, time series data of key quality indicators such as foundation settlement rate and wall crack width are extracted. The time series data is segmented using the sliding window method, and the mean, variance, and extreme values of each time period are extracted as statistical features. The feature dimension is reduced through the principal component analysis method to generate a regional feature matrix. A time series prediction algorithm is used to calculate the change trend of the quality indicators in each area. For example, the foundation settlement rate and crack expansion trend in the podium area in the next month are predicted to generate a trend analysis result.

[0042] According to the trend analysis result, combined with the preset regional quality threshold, anomaly diagnosis is carried out. For example, when the foundation settlement rate in the podium area is greater than the preset threshold, the anomaly diagnosis module is triggered to generate an anomaly diagnosis result. The quality indicator data of different areas is extracted, and the correlation strength between the quality indicators of different areas is calculated through the association mining algorithm. For example, the correlation between the settlement in the podium area and the crack expansion in the adjacent area is analyzed to identify potential chain reaction relationships. The risk diffusion path is predicted in combination with historical monitoring data. For example, the risk of basement wall cracking caused by the settlement in the podium area is predicted to generate a potential chain reaction risk assessment report.

[0043] Finally, the potential chain reaction risk assessment report is input into the pre-trained building engineering safety monitoring model. The long short-term memory network model is used to predict the health trend of the building structure. For example, the impact of the settlement trend in the podium area on the overall building safety is predicted to generate a building structure health trend report including the distribution of risk hotspots, trend comparison curves, and key structure warning information, providing a scientific basis for construction safety management.

[0044] Through the above working process, the present invention realizes the dynamic global analysis of building quality and safety, can timely discover the global risks caused by local minor defects, and provides efficient and accurate technical support for the safety monitoring and risk management of building engineering.

[0045] In summary, the present invention discloses a building engineering safety monitoring method based on machine learning, including obtaining an original detection data set containing building dynamic information and engineering quality indicators; removing environmental noise and measurement errors from the original detection data set by using a noise reduction algorithm to obtain a denoised monitoring data set; dividing the building structure into several regions by using a clustering algorithm according to the monitoring data set, and performing risk analysis to obtain high-risk regions; extracting time series data of key quality indicators for building engineering safety according to the high-risk regions, and calculating the change trend of the quality indicators by using a time series analysis method to obtain a trend analysis result; diagnosing according to the trend analysis result in combination with a preset regional quality threshold, and analyzing the correlation of quality indicators between different regions by using an association mining algorithm to obtain a potential chain reaction risk assessment report; and inputting the potential chain reaction risk assessment report into a pre-trained building engineering safety monitoring model to obtain a building structure health trend report.

[0046] The present invention constructs an adaptive dynamic global analysis system for building quality and safety step by step through the following steps: First, obtain the original detection data set containing building dynamic information and engineering quality indicators to provide a comprehensive data basis for subsequent analysis; Second, adopt a noise reduction algorithm to remove environmental noise and measurement errors, eliminate invalid data caused by equipment failures or instantaneous interferences through outlier filtering, smooth short-term fluctuations by combining the moving average algorithm, and use time series decomposition technology to separate the long-term trend term and the random residual term, retaining the denoised data reflecting the true structural response to ensure the physical consistency and temporal continuity of the monitoring data set; Then, use a clustering algorithm to divide the building structure into several regions, perform clustering analysis based on the denoised multi-dimensional feature matrix, dynamically divide the monitoring regions according to the data space distribution density, break through the rigid limitations of traditional manually preset partitions, make the regional division results adapt to the structural characteristics of different buildings, and at the same time calculate the data dispersion within each clustering cluster, and mark the regions with dispersion exceeding the threshold as high-risk regions to achieve risk positioning; Subsequently, extract the time series data of the key quality indicators in the high-risk regions and perform trend analysis, construct sliding window statistical features for core indicators such as settlement rate and crack expansion amount, reduce the dimension by principal component analysis to eliminate feature redundancy, then use the ARIMA model to fit the index change trend, and dynamically adjust the prediction parameters by combining Bayesian optimization, so that the trend analysis results can capture short-term mutations and reflect long-term evolution laws. At the same time, introduce an elastic threshold mechanism to dynamically set the diagnostic threshold according to the historical data distribution and structural type, avoiding the adaptability defects of fixed thresholds for heterogeneous structures; After that, analyze the correlation of quality indicators between regions through an association mining algorithm, construct a regional quality correlation matrix, use the Apriori algorithm to mine high-frequency association rules, combine the Granger causality test to verify the statistical significance of the association relationship, identify potential chain reaction paths, and quantify the risk conduction probability based on the historical accident case database to construct a risk evolution map; Finally, input the chain reaction risk assessment into the LSTM model to generate a health trend report, train the long short-term memory network with historical monitoring data, embed engineering physical constraints in the model loss function, iterate and update the input sequence through the sliding time window mechanism, predict the trends of key indicators in the future time period, and generate a health trend report including risk hot spot distribution, trend comparison curves and maintenance suggestions by integrating the association mining results, forming a closed-loop decision-making link. Thus, the present invention realizes an adaptive dynamic global analysis of building quality and safety.

[0047] Referring to Figure 2 , the second embodiment of the present invention provides a building engineering safety monitoring system based on machine learning, including: A data acquisition module for acquiring an original detection data set containing building dynamic information and engineering quality indicators; A noise reduction processing module for removing environmental noise and measurement errors from the original detection data set by using a noise reduction algorithm to obtain a denoised monitoring data set; A risk division module, configured to divide a building structure into several regions by using a clustering algorithm according to the monitoring data set, and perform risk analysis to obtain high-risk regions; A trend analysis module, configured to extract time series data of key quality indicators for building engineering safety according to the high-risk regions, and calculate the change trend of the quality indicators by using a time series analysis method to obtain a trend analysis result; A chain diagnosis module, configured to perform diagnosis according to the trend analysis result in combination with a preset regional quality threshold, and analyze the correlation of quality indicators between different regions by using an association mining algorithm to obtain a potential chain reaction risk assessment report; A result output module, configured to input the potential chain reaction risk assessment report into a pre-trained building engineering safety monitoring model to obtain a building structure health trend report.

[0048] It should be noted that a building engineering safety monitoring system based on machine learning provided by an embodiment of the present invention is used to execute all process steps of a building engineering safety monitoring method based on machine learning in the above embodiment. The working principles and beneficial effects of the two correspond one by one, and thus will not be elaborated herein.

[0049] An embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a building engineering safety monitoring program based on machine learning. When the processor executes the computer program, the steps in the above embodiments of various building engineering safety monitoring methods based on machine learning are implemented, such as Figure 1 the step S11 shown. Alternatively, when the processor executes the computer program, the functions of each module / unit in the above device embodiments are implemented, such as the risk division module.

[0050] Exemplarily, the computer program may be divided into one or more modules / units. The one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device.

[0051] The electronic device may be a computing device such as a desktop computer, notebook, handheld computer, and smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the above components are only examples of the electronic device and do not constitute a limitation on the electronic device. It may include more or fewer components than the above, or combine some components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0052] The so-called processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the electronic device and connects various parts of the entire electronic device through various interfaces and lines.

[0053] The memory can be used to store the computer programs and / or modules. The processor realizes various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. Among them, the program storage area may store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area may store data created according to the use of the mobile phone (such as audio data, phone book, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one magnetic disk storage device, flash device, or other volatile solid-state storage devices.

[0054] Among them, if the modules / units integrated in the electronic device are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above-described embodiment methods of the present invention, it can also be completed by a computer program instructing relevant hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-described various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0055] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the attached drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which can be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art can understand and implement it without creative work.

[0056] The specific embodiments described above further elaborate on the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only the specific embodiments of the present invention and is not used to limit the protection scope of the present invention. It is particularly pointed out that for those skilled in the art, any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A construction engineering safety monitoring method based on machine learning, characterized in that: Executed by a computer, including: Obtain the original inspection data set containing building dynamic information and engineering quality indicators; According to the original detection data set, a noise reduction algorithm is used to remove environmental noise and measurement errors to obtain a denoised monitoring data set; Based on the monitoring data set, a clustering algorithm is used to divide the building structure into several areas, and risk analysis is performed to obtain high-risk areas; According to the high-risk areas, extract the time series data of key quality indicators of construction project safety, use the time series analysis method to calculate the change trend of the quality indicators, and obtain the trend analysis results; According to the trend analysis results, diagnosis is performed in combination with preset regional quality thresholds, and the correlation between quality indicators of different regions is analyzed through an association mining algorithm to obtain a potential chain reaction risk assessment report; The potential chain reaction risk assessment report is input into a pre-trained construction engineering safety monitoring model to obtain a building structure health trend report.

2. The construction engineering safety monitoring method based on machine learning according to claim 1 is characterized in that: The obtaining of the original detection data set containing building dynamic information and engineering quality indicators includes: Obtain dynamic information of building structures, foundation settlement data at preset acquisition frequencies, and wall crack monitoring data; Analyze and extract quality indicator data of construction project safety according to the dynamic information, the foundation settlement data and the wall crack monitoring data to obtain an initial monitoring data set; Performing integrity analysis on the initial monitoring data set, using an interpolation algorithm to fill in missing values, and obtaining a complete detection data set; According to the complete detection data set, data format standardization processing is performed to obtain an original detection data set containing building dynamic information and engineering quality indicators.

3. The construction engineering safety monitoring method based on machine learning according to claim 1 is characterized in that: The method of removing environmental noise and measurement errors by using a noise reduction algorithm according to the original detection data set to obtain a denoised monitoring data set includes: According to the original detection data set, identifying and filtering outliers and invalid values ​​in the original detection data set to obtain a preliminary screening data set; According to the preliminary screening data set, the statistical characteristics of data deviation are analyzed and a linear regression model is established, and a linear trend correction is performed on the preliminary screening data set to obtain a corrected data set; According to the correction data set, a sliding average algorithm is used to smooth the fluctuation of the correction data set, and a local average value is calculated through a sliding window to obtain a smoothed data set; According to the smoothed data set, time series decomposition is performed to separate trend terms and residual terms, and long-term trend data that conforms to the laws of engineering physics is retained to obtain a denoised monitoring data set.

4. The construction engineering safety monitoring method based on machine learning according to claim 1 is characterized in that: According to the monitoring data set, a clustering algorithm is used to divide the building structure into several areas, and risk analysis is performed to obtain high-risk areas, including: Extracting various features of the building structure according to the monitoring data set to generate a multi-dimensional feature matrix; According to the multidimensional feature matrix, standardization processing is performed on the differences in different feature dimensions to obtain a standardized feature matrix; According to the standardized feature matrix, a clustering algorithm is used to divide the area and obtain the building safety area category; According to the building safety area category, statistical analysis is performed to calculate the degree of dispersion of data distribution, and areas where the degree of dispersion exceeds a preset risk threshold are marked as high-risk areas.

5. The construction engineering safety monitoring method based on machine learning according to claim 1 is characterized in that: The method of extracting time series data of key quality indicators of construction project safety based on the high-risk areas, calculating the quality indicator change trend using a time series analysis method, and obtaining trend analysis results includes: Extracting time series data of quality indicators according to the high-risk areas to obtain a time series data set; According to the time series data set, a sliding window method is used to perform segmentation processing, extract the mean, variance and extreme value of each time period, and obtain the risk time series statistical characteristics; According to the risk time series statistical characteristics, the principal component analysis method is used to reduce the characteristic dimension and obtain the regional characteristic matrix; According to the regional feature matrix, a time series prediction algorithm is used to calculate the quality index change trend of each region to obtain a trend analysis result.

6. The construction engineering safety monitoring method based on machine learning according to claim 1 is characterized in that: According to the trend analysis results, the diagnosis is performed in combination with the preset regional quality threshold, and the correlation between the quality indicators of different regions is analyzed by the association mining algorithm to obtain a potential chain reaction risk assessment report, including: Comparing the trend analysis result with a preset quality trend threshold, and when the quality index of the trend analysis result is greater than the quality trend threshold, performing an abnormal diagnosis to obtain an abnormal diagnosis result; According to the abnormal diagnosis results, quality index data of different regions are extracted, feature analysis is performed, and a regional quality correlation matrix is ​​obtained; According to the regional quality association matrix, an association mining algorithm is used to calculate the association strength of quality indicators between different regions to obtain the potential chain reaction relationship between regions; Based on the potential chain reaction relationship and combined with historical monitoring data, the risk diffusion path is predicted to obtain a potential chain reaction risk assessment report.

7. The construction engineering safety monitoring method based on machine learning according to claim 1 is characterized in that: The potential chain reaction risk assessment report is input into a pre-trained construction engineering safety monitoring model to obtain a building structure health trend report, including: Taking the historical potential chain reaction risk assessment report as input and the historical building structure health trend report as output, an initial construction engineering safety monitoring model is constructed and trained. When the number of training times is greater than or equal to the preset number of training times, the training is determined to be completed, and a trained construction engineering safety monitoring model is obtained; The potential chain reaction risk assessment report is input into the trained construction engineering safety monitoring model to obtain a building structure health trend report.

8. A construction engineering safety monitoring system based on machine learning, characterized in that: include: A data acquisition module is used to obtain the original detection data set containing building dynamic information and engineering quality indicators; A noise reduction processing module is used to remove environmental noise and measurement errors according to the original detection data set by using a noise reduction algorithm to obtain a denoised monitoring data set; A risk division module, for dividing the building structure into several areas using a clustering algorithm according to the monitoring data set, and performing risk analysis to obtain high-risk areas; A trend analysis module is used to extract time series data of key quality indicators of construction project safety according to the high-risk areas, calculate the quality indicator change trend using a time series analysis method, and obtain trend analysis results; A chain diagnosis module is used to perform diagnosis based on the trend analysis results in combination with a preset regional quality threshold, and to analyze the correlation between quality indicators of different regions through an association mining algorithm to obtain a potential chain reaction risk assessment report; The result output module is used to input the potential chain reaction risk assessment report into the pre-trained construction engineering safety monitoring model to obtain a building structure health trend report.

9. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for monitoring construction project safety based on machine learning as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein, when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the construction project safety monitoring method based on machine learning as described in any one of claims 1 to 7.

Images