Machine learning-based multi-method cross construction risk linkage early warning method
By calculating the disturbance contribution value and disturbance superposition coefficient of a single construction method and combining them with a machine learning model, the comprehensive problem of risk assessment in multi-method cross-construction was solved, and more accurate risk identification and control were achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SUBWAY ENG CO LTD OF CHINA RAILWAY 16TH CONSTR BUREAU
- Filing Date
- 2026-04-30
- Publication Date
- 2026-05-29
AI Technical Summary
Existing technologies are insufficient to comprehensively reflect the interaction between different construction methods under conditions of multiple construction methods being used simultaneously, making it difficult to accurately assess and control construction risks.
By acquiring construction parameters and environmental monitoring parameters, the disturbance contribution value and disturbance superposition coefficient of a single construction method are calculated. The risk level is then determined by combining the data with a machine learning model, and linkage early warning information is generated.
It enables more accurate identification of risk changes under conditions of multiple construction methods, provides a reliable basis for adjusting construction organization and controlling risks, and improves the level of construction safety management.
Smart Images
Figure CN122114664A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of predictive analysis technology based on data processing, specifically to a risk linkage early warning method for multi-method cross-construction based on machine learning. Background Technology
[0002] Under conditions of multiple construction methods operating simultaneously, different construction methods will cause varying degrees of disturbance to the surrounding strata, support structures, groundwater environment, and surrounding buildings. For example, excavation will cause a redistribution of stratum stress and produce an unloading effect; grouting will change the local stratum structure and pore pressure distribution; dewatering will lead to changes in groundwater level and may cause stratum settlement; and vibration operations may cause dynamic disturbance to surrounding structures and pipelines. When multiple construction methods are implemented simultaneously in the same construction area, they may interact with each other, causing the construction disturbances to tend to superimpose or amplify, thereby increasing construction risks.
[0003] In existing technologies, to reduce construction risks, various monitoring devices are typically deployed to monitor settlement, displacement, pore water pressure, and structural stress in the construction area in real time, and corresponding early warning thresholds are set based on the monitoring data. When the monitored parameters exceed the preset thresholds, an alarm is triggered or corresponding construction control measures are taken.
[0004] For example, a prior art patent, CN112031874A, discloses an automated monitoring and control method for tunnel engineering based on BIM technology, which includes the following steps: Step 1: Constructing tunnel surrounding rock structure components and sensor components based on IFC standard information extension, and constructing a tunnel BIM model including the tunnel surrounding rock structure components and sensor components. The tunnel BIM model is identical to the tunnel structure at the construction site; Step 2: Acquiring and storing monitoring data from sensors installed in the tunnel at the construction site; Step 3: Establishing a mapping relationship between the stored monitoring data and the sensor components in the tunnel BIM model, and adding the monitoring data to the tunnel BIM model based on the mapping relationship. This method establishes a connection between the monitoring data collected by on-site sensors and the sensor components in the BIM model, enabling the monitoring of multi-dimensional information during tunnel construction and ensuring construction safety.
[0005] However, most of the above-mentioned early warning methods are based on a single monitoring indicator or a single construction method to make risk judgments, which makes it difficult to comprehensively reflect the interaction between different construction methods under the condition of multiple construction methods being carried out simultaneously. Summary of the Invention
[0006] The purpose of this invention is to provide a risk linkage early warning method for multi-method cross-construction based on machine learning, so as to solve the problems mentioned in the background art.
[0007] To achieve the above objectives, the present invention provides the following technical solution: a multi-method cross-construction risk linkage early warning method based on machine learning, comprising the following steps: S1: Obtain the construction parameters corresponding to multiple construction methods in the cross-construction area and the environmental monitoring parameters corresponding to the cross-construction area, and preprocess the construction parameters and the environmental monitoring parameters to obtain a standardized multi-method feature dataset; S2: Based on the multi-method feature dataset, calculate the single-method disturbance contribution value corresponding to each construction method; S3: Based on the disturbance contribution value of each of the two construction methods and the influence correlation between the two construction methods, calculate the disturbance superposition coefficient between the two construction methods. S4: Based on the disturbance contribution value of each construction method and the disturbance superposition coefficient corresponding to the pairwise combination of all construction methods, calculate the linkage risk value of the cross-construction area. S5: The standardized multi-method feature dataset, the single-method disturbance contribution value, the disturbance superposition coefficient, and the linkage risk value are used as input features and input into a pre-trained machine learning model. The machine learning model outputs the risk level corresponding to the current construction status. S6: When the risk level meets the preset warning triggering conditions, generate and output the corresponding linkage warning information.
[0008] Furthermore, the construction parameters include at least one of grouting pressure, grouting volume, excavation depth, excavation speed, support loading parameters, dewatering intensity, vibration amplitude, and equipment operating parameters; the environmental monitoring parameters include at least one of surface settlement, displacement, pore water pressure, support axial force, structural strain, and groundwater level change.
[0009] As a further step, the preprocessing includes performing missing value completion, outlier removal, dimension unification, and normalization on the construction parameters and environmental monitoring parameters to obtain the standardized multi-method feature dataset.
[0010] As a further step, the first Single-method disturbance contribution of each construction method Calculate using the following formula: ,in, For the first The total number of parameters corresponding to each construction method For the normalized first Item parameter value, For the first The weighting coefficients of the item parameters.
[0011] As a further step, any two construction methods and The perturbation superposition coefficient between Calculate using the following formula: ,in Construction methods Construction methods The degree of correlation between the influences and Construction methods and construction methods The disturbance contribution value of the single-cell method, This is the positive correlation function used to quantify the amplification effect.
[0012] As a further step, the positive correlation function Specifically ,in, This is the preset superposition amplification factor. This is a correction term for the intensity of the interaction, and the correction term for the intensity of the interaction follows... and It increases as it grows.
[0013] As a further step, the combined effect strength correction term is calculated according to the following formula. ,in This is a preset positive correction constant used to avoid the denominator being zero and to adjust the sensitivity of the change in the joint effect strength correction term.
[0014] As a further step, the influence correlation... According to construction methods and construction methods The degree of overlap of the effects of at least one of the following objects—strata, support structures, groundwater-affected areas, surrounding buildings and underground pipelines, and common impacts on the same area—is pre-calibrated based on historical data on interconnected risks, and the following conditions are met: .
[0015] As a further step, the aforementioned linked risk value Calculate using the following formula: , where m is the total number of construction methods.
[0016] As a further step, the machine learning model is a classification model, and its input feature vector is composed of the standardized multi-method feature dataset, the disturbance contribution value of each of the single methods, the disturbance superposition coefficient, and the linkage risk value; the output of the machine learning model is the probability value of the current construction state belonging to multiple preset risk levels, and the linkage early warning information includes at least one of the following: risk level identifier, high-risk construction area location information, and construction method identification information with the highest contribution.
[0017] The beneficial effects of this invention are as follows: by comprehensively modeling the disturbance effects and their interactions generated by various construction methods, and combining them with machine learning models to determine risk levels, it is possible to more accurately identify risk changes under the condition of multiple construction methods being carried out simultaneously during actual construction. This provides a reliable basis for adjusting construction organization and controlling risks, and improves the level of construction safety management, as detailed below.
[0018] 1. By calculating the disturbance contribution value of each construction method individually and constructing a disturbance superposition coefficient between construction methods, it is possible to simultaneously reflect the disturbance impact generated independently by each construction method and the amplification effect produced when multiple construction methods work together. In actual construction, when multiple construction methods are implemented simultaneously in the same construction area, it can more realistically reflect the changes in construction disturbance, thereby helping to identify potential high-risk conditions under conditions of multiple construction methods working together in advance.
[0019] 2. By constructing a linkage risk value calculation model, the disturbance contribution value of a single construction method and the disturbance superposition coefficient between construction methods are comprehensively calculated. This ensures that the obtained linkage risk value reflects both the impact of a single construction method on the construction environment and the combined effect of multiple construction methods. In actual construction, the linkage risk value can be used to quantitatively assess the overall construction risk, thus providing a basis for construction managers to judge the current construction status.
[0020] 3. By inputting multi-method feature data, single-method disturbance contribution values, disturbance superposition coefficients, and linked risk values into a machine learning model to determine risk levels, and outputting linked early warning information when warning conditions are met, construction managers can promptly obtain the risk level and main sources of risk for the current construction status. During actual construction, the construction pace can be adjusted in a timely manner based on the early warning information, monitoring of key areas can be strengthened, or corresponding risk control measures can be taken, thereby reducing construction safety risks under conditions of multiple construction methods operating simultaneously. Attached Figure Description
[0021] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a schematic diagram of the calculation process for the disturbance contribution value of the single-cell method in this invention; Figure 3 This is a schematic diagram of the perturbation superposition coefficient calculation process of the present invention; Figure 4 This is a schematic diagram of the risk level determination process of the present invention; Figure 5 This is a schematic diagram of the linkage early warning output process of the present invention. Detailed Implementation
[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0023] Please see Figures 1-5 The present invention provides the following technical solution: S1. Obtain multi-method feature data and perform preprocessing. In this embodiment, the construction parameters corresponding to multiple construction methods in the cross-construction area and the environmental monitoring parameters corresponding to the cross-construction area are first obtained, and the construction parameters and environmental monitoring parameters are preprocessed to obtain a standardized multi-method feature dataset.
[0024] Among them, the cross-construction area is a construction area in which two or more construction methods are implemented simultaneously in the same construction stage or partially overlapped in the same influence range. The reason for limiting it to simultaneous implementation or partial overlap is that this embodiment needs to calculate the disturbance superposition coefficient under the combined effect of different construction methods. Construction situations that only have alternation and no common action section are not the focus of this embodiment.
[0025] In this embodiment, the collected construction parameters are not arbitrarily listed, but rather selected based on parameters that can characterize the disturbance intensity generated by each construction method in the overlapping construction area. Specifically, for grouting construction, grouting pressure and grouting volume are preferred because they can directly characterize the grout input intensity and its disturbance level to the surrounding strata; for excavation construction, excavation depth and excavation speed are preferred because they can reflect the degree of excavation unloading and the rate of disturbance propagation; for support construction, support loading parameters are preferred because these parameters can directly reflect the effect of the support structure on the surrounding soil and structural system; for dewatering construction, dewatering intensity and groundwater pumping flow rate are preferred because they directly correspond to changes in groundwater level and the resulting stratum response; for vibration operation construction, vibration amplitude and equipment operating parameters are preferred because they can reflect the intensity of vibration disturbance input. The above construction parameters are selectively collected according to the disturbance characteristics of each construction method and the on-site monitoring conditions to construct a disturbance contribution calculation model for the corresponding construction method.
[0026] At the same time, environmental monitoring parameters corresponding to the cross-construction areas are collected. In this embodiment, the environmental monitoring parameters preferably include surface settlement, displacement, pore water pressure, support axial force, structural strain, and groundwater level change. The reason for prioritizing these parameters is that they correspond to the stratum response, structural response, and water environment response, respectively, and can reflect the changing state of the construction environment under the combined action of multiple construction methods from different perspectives, thereby providing a response basis for subsequent calculation of the disturbance contribution value of a single construction method and linkage risk analysis.
[0027] To ensure that data from different sources, with different dimensions, and with different sampling frequencies can be included in a unified calculation framework, the construction parameters and environmental monitoring parameters are preprocessed after the parameter collection is completed. The preprocessing includes missing value completion, outlier handling, dimension unification, normalization, and sampling period alignment.
[0028] Regarding missing value completion, this embodiment adopts a targeted approach based on the changing characteristics of construction and monitoring parameters. For parameters with strong continuity and high correlation between adjacent sampled values, interpolation between adjacent acquisition times is preferred to complete the data and maintain the continuity of data changes. For parameters that are greatly affected by changes in working conditions and whose use of historical averages may introduce deviations after continuous missing values, such as grouting pressure and drainage flow rate, when short-term continuous missing values occur, reference values of similar working conditions are preferred to fill the gaps to avoid weakening the actual disturbance characteristics by directly using historical averages. Through the above methods, the original changing patterns of construction disturbances can be preserved as much as possible while ensuring data integrity.
[0029] Regarding outlier handling, this embodiment identifies sampled values that significantly exceed the allowable fluctuation range of parameters, as well as sampled values whose changes within adjacent acquisition cycles exceed a preset mutation threshold and do not match the actual working conditions on site, as outliers. For outliers, it is preferable to replace them with the nearest effective value or the moving average value. The reason for this processing is that data collected at cross-construction sites is easily affected by sensor jitter, short-term communication failures, and instantaneous equipment fluctuations. If outliers are directly incorporated into subsequent disturbance contribution calculations, the disturbance contribution value of a single construction method can be amplified by local noise, thereby affecting the stability of the disturbance superposition coefficient and the linkage risk value.
[0030] Regarding dimensional unification, pressure parameters, displacement parameters, flow parameters, and strain parameters are converted to a unified standard unit before being used in subsequent calculations. The purpose of dimensional unification is to eliminate the direct impact of differences in the units and orders of magnitude of different parameters on the weighted calculation results, so that the subsequent disturbance contribution value can reflect the actual disturbance degree corresponding to the change of the parameter itself, rather than just reflecting the difference in the size of the parameter units.
[0031] In terms of normalization, this embodiment adopts the minimum-maximum normalization method to map different parameters to a unified numerical range. This facilitates weighted analysis of parameters from different construction methods on the same computational scale. The reason for using minimum-maximum normalization instead of standard deviation normalization is that construction parameters and environmental monitoring parameters often do not follow a stable normal distribution in actual engineering projects, and boundary-shaped or skewed distributions are prone to occur when switching between different working conditions. Using minimum-maximum normalization is more conducive to maintaining the actual fluctuation range information of the parameters. The normalization formula is as follows: ,in, Indicates the first The original parameter value of the k-th term corresponding to each construction method. This represents the value of the k-th parameter after normalization. and These represent the minimum and maximum values of the k-th parameter within a preset statistical interval, respectively. = When the parameter is normalized, the result is recorded as 0, because the parameter does not change within the corresponding statistical interval, and its incremental impact on the disturbance contribution value can be regarded as zero.
[0032] Regarding the sampling period alignment, in order to ensure that different construction method parameters and environmental monitoring parameters correspond to the same construction state, this embodiment sets the unified acquisition period to 5 minutes. The reason for setting this unified acquisition period is that 5 minutes can cover the common data upload cycle of on-site monitoring equipment and reflect the changing trend of construction disturbances within a short time scale, thus balancing data real-time performance and computational stability. For high-frequency vibration data, it is preferable to extract the maximum value within each 5-minute period to retain the peak disturbance information. For slowly changing monitoring data such as displacement, settlement, and groundwater level, it is preferable to extract the last value at the end of each 5-minute period to reflect the current state. For parameters acquired at low frequency and without new values in the current period, the most recent valid acquisition value is used for alignment. Through the above methods, data from different sources can form a set of characteristic records corresponding to the same construction state under a unified period.
[0033] After completing the above preprocessing, the construction parameters corresponding to each construction method are organized into a subset of construction method parameters, and the environmental monitoring parameters are organized into a subset of environmental response parameters. This leads to the construction of a standardized multi-method feature dataset, which can be represented as follows: ,in These represent the standardized construction parameter subsets corresponding to the 1st, 2nd, up to the mth construction methods, respectively, where m represents the total number of construction methods; and M represents the standardized environmental monitoring parameter subsets corresponding to the cross-construction areas.
[0034] Through the above steps, the operating parameters and corresponding environmental response parameters of different construction methods can be converted into a data set with a unified format and scale that can reflect the current common construction status, providing a data foundation for subsequent calculation of the disturbance contribution value of each construction method.
[0035] S2: After completing the construction of the multi-method feature dataset in S1, based on the standardized multi-method feature dataset, calculate the single-method disturbance contribution value corresponding to each construction method to quantify the degree of independent disturbance generated by each construction method on the cross-construction area without considering the combined effects of other construction methods. Specifically, for the i-th construction method, its corresponding standardized construction parameter subset in S1 can be represented as: Based on this, the first Single-method disturbance contribution of each construction method Calculate using the following formula: ,in, For the first The total number of parameters corresponding to each construction method Represents the normalized th Item parameter value, For the first The weighting coefficients of the item parameters.
[0036] The reason for using the weighted summation method described above to calculate the disturbance contribution value of a single construction method is that although different construction methods will all cause disturbance to the overlapping construction area, the influence of different parameters within each method on the degree of disturbance is not the same. For example, for grouting construction, grouting pressure usually reflects the input intensity of grout to the surrounding strata more directly than grouting duration; for excavation construction, excavation depth and excavation speed usually characterize the degree of unloading and the speed of disturbance propagation better than advancement auxiliary parameters; for dewatering construction, dewatering intensity and pumping flow rate can more directly reflect the degree of change in the groundwater environment. Therefore, by assigning different weight coefficients to different parameters, the disturbance contribution value of a single construction method can more accurately reflect the actual independent disturbance level caused by the construction method to the construction environment, rather than simply reflecting the number or magnitude of parameters.
[0037] In this embodiment, the weighting coefficient Based on historical construction samples, monitoring response data, and parameter sensitivity analysis results for the corresponding construction method, the following steps are taken: First, construction parameters and environmental monitoring response data for the corresponding construction method during historical construction processes are collected; second, the impact of changes in each parameter on environmental monitoring responses such as surface settlement, displacement, pore water pressure, and structural strain is analyzed; finally, based on the contribution of each parameter to changes in environmental response, corresponding weighting coefficients are determined. The weighting coefficients determined in the above manner can provide a clear engineering basis for the disturbance contribution value of the single-work method, avoiding the deviation of results caused by relying solely on manual experience to assign values.
[0038] Furthermore, to ensure the comparability of disturbance contribution values for different construction methods, this embodiment normalizes the weight coefficients corresponding to each parameter within the same construction method to satisfy the following constraints: Through the above constraints, it is possible to make The numerical values reflect the combined effect of parameter changes rather than being directly affected by differences in the total weight scale, thus facilitating a horizontal comparison of the degree of disturbance between different construction methods.
[0039] In this embodiment, when the construction method is grouting, its single-method disturbance contribution value can be calculated by weighting parameters such as grouting pressure and grouting volume; when the construction method is excavation, its single-method disturbance contribution value can be calculated by weighting parameters such as excavation depth and excavation speed; when the construction method is dewatering, its single-method disturbance contribution value can be calculated by weighting parameters related to dewatering intensity, pumping flow rate, and groundwater level changes; when the construction method is support construction, its single-method disturbance contribution value can be calculated by weighting parameters related to support loading and structural stress. Thus, the single-method disturbance contribution value corresponding to each construction method can be obtained. , ,..., This provides a basis for further calculation of the disturbance superposition coefficients corresponding to the pairwise combinations of construction methods.
[0040] It should be noted that the disturbance contribution value of the single construction method calculated in S2 represents the basic disturbance level of the environment of the cross-construction area when each construction method acts alone. It does not yet reflect the amplification effect when multiple construction methods are constructed together. The combined effect of multiple construction methods will be further characterized in the subsequent S3 through the disturbance superposition coefficient.
[0041] S3: After calculating the disturbance contribution value of each construction method in S2, based on the disturbance contribution value of each of the two construction methods and the influence correlation between them, calculate the disturbance superposition coefficient between the two construction methods. This coefficient characterizes the amplification degree of the disturbance influence of different construction methods on the cross-construction area under the condition of joint action. Specifically, for any two construction methods... and The perturbation superposition coefficient between Calculate using the following formula: ,in Construction methods Construction methods The degree of correlation between the influences and Construction methods and construction methods The disturbance contribution value of the single-cell method, Here is the positive correlation function used to quantify the amplification effect, specifically: ,in, This is the preset superposition amplification factor. This is a correction term for the intensity of the interaction, and the correction term for the intensity of the interaction follows... and The intensity increases with increasing, and the correction term for the combined effect strength is calculated according to the following formula. ,in This is a preset positive correction constant used to avoid the denominator being zero and to adjust the sensitivity of the change in the joint effect strength correction term, and also affects the correlation. According to construction methods and construction methods The degree of overlap of the effects of at least one of the following objects—strata, support structures, groundwater-affected areas, surrounding buildings and underground pipelines, and common impacts on the same area—is pre-calibrated based on historical data on interconnected risks, and the following conditions are met: .
[0042] Among them, when two construction methods act on the same object and have a large common impact range, the corresponding impact correlation is... The value is relatively high; when the two construction methods affect different objects or have a low degree of common influence, the corresponding influence correlation is high. The value is relatively low. By introducing the influence correlation degree, the correlation between different construction methods at the engineering object level can be introduced into the calculation process of the disturbance superposition coefficient. Thus, the disturbance superposition coefficient not only reflects the magnitude of the disturbance intensity of the two construction methods, but also reflects the degree of mutual influence between the two construction methods.
[0043] in , , and They represent construction methods respectively. and construction methods The degree of overlap in effects, the degree of common impact on the same region, and the historical linkage correction coefficient obtained based on historical linkage risk data. , and These are the corresponding association degree combination weights, and they satisfy the condition that the sum of the three is 1 and all are non-negative, where: Indicates construction method and construction methods The area of the overlapping region that affects the target. Indicates construction method and construction methods The combined area affecting the object; when the areas of influence of the two construction methods on the object completely overlap. The value is close to 1; when the influence areas of the two construction methods on the object are basically non-overlapping, the value is close to 0. ,in This represents the total number of monitoring response indicators used to characterize the state of common impact. Indicates construction method For the Normalized impact values corresponding to the monitoring response indicators; Indicates construction method For the The normalized impact value corresponding to each monitoring response indicator. The method of averaging the minimum values is used because when both construction methods have a significant impact on the same monitoring response indicator, it indicates a strong synergistic relationship between them in the corresponding region; conversely, when either construction method has a weak impact on the indicator, it indicates a limited degree of synergistic influence. Furthermore, the historical linkage correction coefficient... in This indicates the construction methods in historical construction samples. and construction methods The number of samples that appear simultaneously and whose corresponding risk level reaches the preset medium-high risk level. This indicates the construction methods in historical construction samples. and construction methods The total number of samples that appear together This represents a preset positive smoothing constant used to avoid zero denominators and reduce the impact of small sample fluctuations on the calculation results. In this embodiment, to ensure that the value affecting the correlation degree remains stable within a preset range, it is preferable to use... , and Preprocessed to values within the range [0,1], thus allowing 0 to be... ≤1, therefore, construction method and construction methods When the areas of influence overlap significantly, the common impact on the same area is strong, and the corresponding high-risk linkage has occurred frequently in the past, the correlation of influence increases accordingly. Conversely, when the spatial overlap between the two construction methods is small, the common impact is weak, and the historical linkage risk is low, the correlation of influence decreases accordingly. Through the above methods, the correlation of influence can have a clear data source, calculation path, and engineering meaning, thereby providing a quantifiable basis for the subsequent calculation of the disturbance superposition coefficient.
[0044] Furthermore, when construction methods and construction methods When the single-cell disturbance contribution values are both large and the correlation between the two influences is high, the function... The calculated amplification term increases accordingly, thus increasing the perturbation superposition coefficient. Increase; when construction method and construction methods When the individual perturbation contributions are both small, or the correlation between the two is low, the amplification term decreases accordingly, and the perturbation superposition coefficient... This also decreases accordingly. As a result, the disturbance superposition coefficient can be dynamically changed according to the strength of the combined effect between different construction methods, rather than using a fixed constant to represent the superposition relationship between different construction methods.
[0045] In this embodiment, by calculating the corresponding disturbance superposition coefficients for each pair of construction methods, the set of disturbance superposition relationships between multiple construction methods within the overlapping construction area can be obtained, namely: ,in, This represents the total number of construction methods.
[0046] S4: After calculating the disturbance superposition coefficients for each pairwise combination of construction methods in S3, based on the disturbance contribution values of each individual construction method and the disturbance superposition coefficients for each pairwise combination of construction methods, calculate the linkage risk value of the intersecting construction area. This value characterizes the overall risk level under the combined influence of the independent disturbances of each construction method and the amplification effect of multiple construction methods. The specific linkage risk value is... Calculate using the following formula: Where m is the total number of construction methods, in the above formula, the first term This is used to characterize the total amount of foundation disturbance generated by each construction method individually in the overlapping construction area; the second item This is used to characterize the total amplified disturbance generated when different construction methods are combined in pairs. By including the total basic disturbance and the total amplified disturbance in the linkage risk value R, the linkage risk value can reflect not only the degree of impact of each construction method on the construction environment, but also the linkage amplification effect when multiple construction methods are implemented together.
[0047] Furthermore, when only a single construction method in the cross-construction area causes significant disturbance, while the disturbances of other construction methods are relatively small or have a low degree of mutual influence, the linkage risk value R is mainly formed by the sum of the individual disturbance contribution values of each construction method. When multiple construction methods in the cross-construction area have high disturbance contribution values and corresponding disturbance superposition coefficients, the proportion of the amplification term formed by the combined effect of two construction methods in the linkage risk value R increases, thus significantly increasing the linkage risk value. Therefore, the linkage risk value can distinguish between two different scenarios: high disturbance from a single construction method and high risk from the combined effect of multiple construction methods.
[0048] In this embodiment, assuming there are three construction methods within the overlapping construction area, namely construction method 1, construction method 2, and construction method 3, the corresponding linkage risk value can be written as: Under the combined effect of the three construction methods, the linked risk value is composed of three individual construction method disturbance contribution terms and three pairwise combination amplification terms. For scenarios with more construction methods acting in parallel, the above method can be further extended.
[0049] Through the above steps, the single-cell disturbance contribution value obtained in S2 and the disturbance superposition coefficient obtained in S3 can be further integrated into a unified linkage risk value R, thereby providing quantitative input for risk level judgment in the subsequent S5.
[0050] S5: After completing the calculation of the linkage risk value of the cross-construction area in S4, the standardized multi-method feature dataset, the disturbance contribution value of each construction method, the disturbance superposition coefficient corresponding to the pairwise combination of each construction method, and the linkage risk value are used as input features and input into the pre-trained machine learning model. The machine learning model outputs the risk level corresponding to the current construction state.
[0051] In this embodiment, the machine learning model is a classification model. Its input feature vector is composed of a standardized multi-method feature dataset, the disturbance contribution value of each single method, the superposition coefficient of each disturbance, and the linkage risk value. The output of the machine learning model is the probability value of the current construction state belonging to multiple preset risk levels.
[0052] Specifically, the standardized multi-method feature dataset obtained in S1 can be denoted as X, and the single-method disturbance contribution value corresponding to each construction method obtained in S2 can be denoted as... , ,..., The disturbance superposition coefficients corresponding to the pairwise combinations of each construction method obtained in S3 are denoted as... Let R be the linkage risk value obtained in S4. Then, the input feature vector input to the machine learning model can be expressed as: , where Z represents the comprehensive input feature vector used to characterize the current cross-construction status. By inputting the original standardized features, independent disturbance information of multiple construction methods, information on the combined effects of multiple construction methods, and overall linkage risk information into the machine learning model, the machine learning model can simultaneously consider the foundation disturbance, coupling amplification, and overall risk status under the multi-construction scenario in the process of risk level discrimination, rather than relying solely on a single type of feature for discrimination.
[0053] In this embodiment, the preset risk levels include at least low risk, medium risk, and high risk. After the machine learning model outputs the probability values of the current construction status belonging to low risk, medium risk, and high risk, the risk level with the highest probability value is taken as the target risk level corresponding to the current construction status. For example, when the machine learning model outputs the probability values of the current construction status belonging to low risk, medium risk, and high risk as P1, P2, and P3, respectively, if P3 is the largest, the current construction status is determined to be high risk; if P2 is the largest, the current construction status is determined to be medium risk; and if P1 is the largest, the current construction status is determined to be low risk.
[0054] Furthermore, the machine learning model can adopt a classification model trained based on historical construction samples. The historical construction samples include construction parameters, environmental monitoring parameters, corresponding calculated single-method disturbance contribution values, disturbance superposition coefficients, linkage risk values, and corresponding actual risk level labels collected during the historical construction process. By training on the above historical construction samples, the machine learning model can establish a mapping relationship between input features and risk levels, thereby judging the risk level of the current construction status in the actual construction process.
[0055] Through the above steps, the various intermediate quantities obtained in the previous steps can be further integrated to output the risk level corresponding to the current cross-construction status, providing a basis for judgment for the subsequent linkage early warning output in S6.
[0056] S6: After completing the risk level determination corresponding to the current construction status in S5, when the risk level meets the preset early warning triggering conditions, the corresponding linkage early warning information is generated and output. In this embodiment, the linkage early warning information includes at least one of the following: risk level identifier, high-risk construction area location information, and construction method identification information with the highest contribution.
[0057] Specifically, when the target risk level output by the machine learning model is high risk, a linked early warning is directly triggered; when the target risk level output by the machine learning model is medium risk, a prompt-level early warning can be triggered according to the pre-set early warning strategy on site; when the target risk level output by the machine learning model is low risk, no linked early warning is triggered, and only the current construction status record is retained. In this way, the risk level judgment result can be correlated with different levels of early warning response, thereby realizing graded early warning output.
[0058] Among them, the risk level identifier is used to intuitively represent the risk level of the current cross-construction status; the high-risk construction area location information is used to indicate the area of action, monitoring area or construction section with a higher risk value in the current cross-construction area; the construction method identification information with the highest contribution is used to represent the construction method that contributes the most to the linkage risk value among multiple construction methods, so as to help on-site management personnel identify the main sources of risk.
[0059] Furthermore, in this embodiment, the construction method identification information with the highest contribution can be determined based on the magnitude of the single-method disturbance contribution value corresponding to each construction method. Specifically, the single-method disturbance contribution values corresponding to each construction method are compared. The construction method with the largest value is identified as the construction method with the highest contribution under the current construction state. When there are two or more construction methods whose single method disturbance contribution values are close or at a high level at the same time, the corresponding multiple construction methods can also be marked as the main source of risk.
[0060] In this embodiment, the linkage early warning information can be output to the construction monitoring platform, the on-site management terminal, or the early warning display interface. The output method may include at least one of the following: graphical interface display, area highlighting, alarm prompt information push, and risk record writing. Through the above methods, the risk level judgment result can be fed back to the construction site in a timely manner, thereby providing a basis for construction organization adjustment, monitoring intensification, and risk control.
[0061] Through the above steps, corresponding linkage early warning information can be generated based on the risk level results output by the machine learning model, thereby completing the entire processing flow from multi-method feature data acquisition, single-method disturbance modeling, inter-method disturbance superposition analysis, linkage risk value calculation to risk level discrimination and linkage early warning output.
[0062] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A risk linkage early warning method for multi-method cross-construction based on machine learning, characterized by: It includes the following steps: S1: Obtain the construction parameters corresponding to multiple construction methods in the cross-construction area and the environmental monitoring parameters corresponding to the cross-construction area, and preprocess the construction parameters and the environmental monitoring parameters to obtain a standardized multi-method feature dataset; S2: Based on the multi-method feature dataset, calculate the single-method disturbance contribution value corresponding to each construction method; S3: Based on the disturbance contribution value of each of the two construction methods and the influence correlation between the two construction methods, calculate the disturbance superposition coefficient between the two construction methods. S4: Based on the disturbance contribution value of each construction method and the disturbance superposition coefficient corresponding to the pairwise combination of all construction methods, calculate the linkage risk value of the cross-construction area. S5: The standardized multi-method feature dataset, the single-method disturbance contribution value, the disturbance superposition coefficient, and the linkage risk value are used as input features and input into a pre-trained machine learning model. The machine learning model outputs the risk level corresponding to the current construction status. S6: When the risk level meets the preset warning triggering conditions, generate and output the corresponding linkage warning information.
2. The multi-method cross-construction risk linkage early warning method based on machine learning according to claim 1, characterized in that: The construction parameters include at least one of the following: grouting pressure, grouting volume, excavation depth, excavation speed, support loading parameters, precipitation intensity, vibration amplitude, and equipment operating parameters; the environmental monitoring parameters include at least one of the following: surface settlement, displacement, pore water pressure, support axial force, structural strain, and groundwater level change.
3. The multi-method cross-construction risk linkage early warning method based on machine learning according to claim 1, characterized in that: The preprocessing includes filling in missing values, removing outliers, unifying dimensions, and normalizing the construction parameters and environmental monitoring parameters to obtain the standardized multi-method feature dataset.
4. The multi-method cross-construction risk linkage early warning method based on machine learning according to claim 1, characterized in that: No. Single-method disturbance contribution of each construction method Calculate using the following formula: ,in, For the first The total number of parameters corresponding to each construction method For the normalized first Item parameter value, For the first The weighting coefficients of the item parameters.
5. The multi-method cross-construction risk linkage early warning method based on machine learning according to claim 1, characterized in that: Any two construction methods and The perturbation superposition coefficient between Calculate using the following formula: ,in Construction methods Construction methods The degree of correlation between the influences and Construction methods and construction methods The disturbance contribution value of the single-cell method, This is the positive correlation function used to quantify the amplification effect.
6. The multi-method cross-construction risk linkage early warning method based on machine learning according to claim 5, characterized in that: The positive correlation function Specifically ,in, This is the preset superposition amplification factor. This is a correction term for the intensity of the interaction, and the correction term for the intensity of the interaction follows... and It increases as it grows.
7. The multi-method cross-construction risk linkage early warning method based on machine learning according to claim 6, characterized in that: The interaction strength correction term is calculated according to the following formula. ,in This is a preset positive correction constant used to avoid the denominator being zero and to adjust the sensitivity of the change in the joint effect strength correction term.
8. The multi-method cross-construction risk linkage early warning method based on machine learning according to claim 5, characterized in that: The influence correlation According to construction methods and construction methods The degree of overlap of the effects of at least one of the following objects—strata, support structures, groundwater-affected areas, surrounding buildings and underground pipelines, and common impacts on the same area—is pre-calibrated based on historical data on interconnected risks, and the following conditions are met: .
9. The multi-method cross-construction risk linkage early warning method based on machine learning according to claim 1, characterized in that: The linked risk value Calculate using the following formula: , where m is the total number of construction methods.
10. The multi-method cross-construction risk linkage early warning method based on machine learning according to claim 1, characterized in that: The machine learning model is a classification model, and its input feature vector is composed of the standardized multi-method feature dataset, the disturbance contribution value of each of the single methods, the disturbance superposition coefficient, and the linkage risk value. The output of the machine learning model is the probability value of the current construction state belonging to multiple preset risk levels. The linkage early warning information includes at least one of the following: risk level identifier, high-risk construction area location information, and construction method identification information with the highest contribution.