Frame house protection structure optimization method under action of slope cutting house building induced landslide

Abstract

The invention discloses a frame house protection structure optimization method under the action of slope cutting house building induced landslide. The method comprises the steps of collecting and preprocessing multi-source spatio-temporal data of a back slope and a back wall of a house; performing sliding window division and feature extraction to obtain partition state parameters; establishing a Bayesian network model, and reasoning the risk probability of each partition structure; optimizing the layout of partitions, windows and reinforcing ribs according to the risk probability distribution, and forming and implementing a construction scheme; and continuously monitoring in an operation stage, and periodically inputting data to dynamically optimize structural parameters. According to the method, through fusion of multi-source spatio-temporal data analysis and Bayesian network reasoning, dynamic optimization and toughness self-adaptive regulation and control of the rear wall protection structure of the frame house under the influence of the landslide are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of civil engineering and structural safety protection, and particularly relates to a method for optimizing a frame house protection structure under the action of a landslide induced by a cut-slope building. BACKGROUND

[0002] With the advancement of mountain development and construction, cut-slope buildings are widely used in hilly and mountainous areas, and the cut-slope behavior significantly increases the risk of inducing landslides and other geological disasters. Existing house protection structures are mostly based on traditional static design and empirical reinforcement measures, and generally use uniform zoning and conventional reinforcement bar arrangement. In engineering practice, the back wall protection structure often fails to conduct dynamic analysis on the multi-source space-time monitoring data such as spatial zoning state parameters, soil deformation characteristics, rainfall and geological factors, and the structure design cannot take into account the differences in zoning risks and the complex stress changes under landslide induction. Traditional landslide risk assessment methods mostly rely on a single physical quantity or an empirical threshold, and cannot cover the synergistic effect of multiple parameters such as surface deformation, strain, stress, and soil moisture content, and it is also difficult to describe the probability distribution relationship between the local damage expansion and the overall failure in the spatial zoning.

[0003] The existing structure protection optimization method lacks a closed loop of real-time monitoring and dynamic reasoning of zoning risks, and the protection design cannot adjust key parameters such as spatial zoning division, window size and reinforcement bar arrangement in time according to the monitoring data during the operation stage of the house. Most protection structures are fixed after construction and cannot realize subsequent adaptive adjustment of zoning and optimization of structure toughness, and the house is prone to local damage spread or even overall failure under extreme rainfall or geological disaster induction. Due to the failure to effectively integrate multi-source monitoring information, the structure safety and disaster prevention capacity are limited. The existing technology has obvious deficiencies in comprehensive identification of zoning risks, dynamic fusion of multiple parameters and adaptive optimization of structure.

[0004] Therefore, how to provide a method for optimizing a frame house protection structure under the action of a landslide induced by a cut-slope building is a problem that needs to be solved by those skilled in the art. SUMMARY

[0005] One object of the present application is to provide a method for optimizing a frame house protection structure under the action of a landslide induced by a cut-slope building. The present application collects and processes multi-source space-time monitoring data, extracts state characteristic parameters of each spatial zoning of the back wall using a sliding window transformer, realizes intelligent reasoning and joint evaluation of the structure risk probability of the spatial zoning by combining a Bayesian network model, and then optimizes the structure parameters such as spatial zoning, window size and reinforcement bar arrangement. Through periodic updating of the monitoring data during the operation stage of the house, the present application realizes the toughness adaptive optimization of the structure protection parameters. The present application has the advantages of multi-source information fusion, accurate evaluation of zoning risks and adaptive optimization of structure, and improves the safety, durability and disaster prevention capacity of the house structure in the cut-slope building scenario.

[0006] The method for optimizing the protection structure of a frame house under the action of a landslide induced by a cutting slope building according to the embodiment of the application comprises: Collecting a multi-source spatio-temporal data set of a rear slope and a rear wall area of the house, preprocessing the multi-source spatio-temporal data set, and generating a standardized multi-source spatio-temporal data set; Dividing the standardized multi-source spatio-temporal data set according to a preset spatial step and a time step to obtain a sliding window, using a sliding window transformer to extract features of data in each window, and obtaining state feature parameters of each spatial partition; Taking the state feature parameters of each spatial partition as node variables, establishing a Bayesian network model, inputting the state feature parameters to perform conditional probability reasoning, and obtaining a structural risk probability distribution of each spatial partition of the rear wall; According to the structural risk probability distribution, optimizing the structural partition, window size, and reinforcing bar arrangement of the protection structure of the rear wall, setting up a cross-shaped embedded column-beam structure in the rear wall to realize spatial partition of the wall, setting small-size light windows in the upper two partitions of the cross-shaped structure, setting closed walls or reinforcing bars in the remaining partitions, performing structural construction according to the optimization design, and implementing the structural partition, window size, and reinforcing bar arrangement in the construction drawing and on-site implementation; During the operation stage of the house, continuously monitoring the state feature parameters of the rear wall partitions, periodically inputting new monitoring data into the sliding window transformer and the Bayesian network model, dynamically updating the parameters, and realizing adaptive optimization of the resilience of the protection structure of the house.

[0007] Optionally, the multi-source spatio-temporal data set specifically includes topographic and geological parameters, rainfall duration, soil moisture content, wall stress parameters, and structural layout information of the rear slope and the rear wall area of the house.

[0008] Optionally, the preprocessing of the multi-source spatio-temporal data set specifically includes format unification, missing value filling, abnormal value elimination, and data standardization processing of the multi-source spatio-temporal data set.

[0009] Optionally, the state feature parameters of each spatial partition are obtained by: For each spatial partition and each time interval in the standardized multi-source spatio-temporal data set, the change rates of four types of monitoring parameters, i.e., surface deformation, strain, stress, and soil moisture content, in a continuous time interval are calculated, the change rates of the four types of monitoring parameters are weighted and fused according to the importance weights of the surface deformation, strain, stress, and soil moisture content, a joint change perception index corresponding to each spatial partition and each time interval is obtained, it is determined whether there is a sudden extreme change in the four types of monitoring parameters, and a sudden change monitoring result corresponding to each spatial partition and each time interval is obtained. According to the joint change perception index and the sudden change monitoring result, each spatial partition and each time interval is adaptively divided by a sliding window, when the joint change perception index is higher than a preset threshold or the sudden change monitoring result is positive, a first spatial window size and a first time window step are used to divide the spatial partition and the time interval by the sliding window to obtain a first type of sliding window, when the joint change perception index is lower than or equal to the preset threshold and the sudden change monitoring result is negative, a second spatial window size and a second time window step are used to divide the spatial partition and the time interval by the sliding window to obtain a second type of sliding window, the first spatial window size and the first time window step are preset fine-grained parameters, and the second spatial window size and the second time window step are preset wide-scale parameters, according to the spatial position continuity and the time sequence continuity, all pairs of directly adjacent sliding windows in the spatial or time domain are identified, and a set of adjacent sliding windows is established; For the first type of sliding window and the second type of sliding window, global statistical features, local extreme value features, change rate features and segmented interval features in the sliding window are extracted according to spatial scales and time scales respectively to generate a multi-scale feature subset; The extracted global statistical features, local extreme value features, change rate features and segmented interval features are spliced and fused by weighting to generate a multi-scale feature vector of each sliding window; For each pair of directly adjacent sliding windows in the set of adjacent sliding windows, a sliding window overlap strategy is performed, the boundaries between the adjacent sliding windows are extended inward according to a preset overlap length on the basis of maintaining the original window boundary sequence, and a set of pairs of sliding windows with data overlapping regions in the spatial position or the time sequence is formed; For the overlapping region data corresponding to each pair of sliding windows in the set of pairs of sliding windows, a trend extrapolation is performed on the multi-scale feature vector of the sliding window boundary by using the change trend of the overlapping region data, the multi-scale feature vector generated by the trend extrapolation is fused with the multi-scale feature vector of the original overlapping region in a Bayesian smoothing weight manner to obtain a multi-scale feature vector after boundary continuity processing and abnormal correction, and if a feature mutation or an abnormal residual error occurs in the overlapping region data, a residual correction method is used to correct the multi-scale feature vector after the boundary continuity processing and the abnormal correction and the multi-scale feature vector of the corresponding adjacent sliding window; The multi-scale feature vectors are summarized with the multi-scale feature vectors after the boundary continuity processing and the abnormal correction to generate a multi-dimensional state feature parameter corresponding to each spatial partition and each time interval, and a state feature parameter of each spatial partition is obtained.

[0010] Optionally, the structure risk probability distribution of each spatial partition of the rear wall is obtained, including: The state characteristic parameters of each obtained spatial partition are taken as candidate node variables, and the spatial partition node structure is dynamically divided according to the spatial distribution characteristics and variation trend of the state characteristic parameters, so as to determine a spatial partition node set; For the determined spatial partition node set, the spatial partition node with state mutation is subdivided into a plurality of spatial partition sub-nodes according to the abnormal variation of the state characteristic parameters and the risk distribution difference, the spatial partition node with high state correlation or balanced risk is merged into a single spatial partition node, and a spatial partition node set is obtained; Each spatial partition node in the spatial partition node set is taken as an input node, four types of monitoring parameters of ground deformation, strain, stress and soil moisture content are taken as external disaster factor nodes, and a preset overall structure influence node in a multilayer Bayesian network model structure is taken as an output node of the overall structure failure probability, so as to construct a multilayer Bayesian network model structure comprising the spatial partition node, the external disaster factor node and the overall structure influence node; For each node in the multilayer Bayesian network model structure, the conditional probability distribution of each node is regularly updated based on historical monitoring data, real-time acquisition data and external disaster environment data, so as to form a multilayer Bayesian network model structure driven by a plurality of sources of data; The conditional probability table set and the state characteristic parameters of each spatial partition node in the spatial partition node set are input, and the Bayesian inference method is used to find the corresponding state parameter combination in the conditional probability table, read the conditional probability of the local damage of the spatial partition node under the state parameter combination, and if there are a plurality of parent nodes or a plurality of observation information, the probability propagation mechanism is used to sequentially perform probability weighting and normalization on the state of all parent nodes, so as to obtain the local damage probability of each spatial partition node under the current observation information; The local damage probability of each spatial partition node in the spatial partition node set is calculated by using a probability joint inference function, so as to obtain the overall structure failure probability of the overall structure influence node; Based on the multilayer Bayesian network model structure and the local damage probability of each spatial partition node in the spatial partition node set, the correlation between the spatial partition nodes is analyzed, the network edge connection between the spatial partition nodes with significant correlation is established, the network edge connection between the spatial partition nodes with low correlation is deleted, and a dynamically adjusted Bayesian network model structure is obtained; The local damage probability of each spatial partition node in the spatial partition node set, the overall structure failure probability of the overall structure influence node, and the structure risk probability distribution corresponding to the dynamically adjusted Bayesian network model structure are output, so as to obtain the structure risk probability distribution of each spatial partition of the rear wall.

[0011] Optionally, the structural partition, window size and reinforcement arrangement of the back wall protective structure are optimized according to the structural risk probability distribution, a cross-shaped embedded column beam structure is set up in the back wall to realize spatial partition of the wall, small-size light windows are arranged in the upper two partitions of the cross-shaped structure, and the remaining partitions are provided with closed walls or reinforcements, the structural partition, window size and reinforcement arrangement are implemented in the construction drawing and on-site implementation, including: According to the structural risk probability distribution of each spatial partition, the spatial partition parameters, window size parameters and reinforcement arrangement parameters of the back wall structure are combined to establish a structural protection optimization objective function; According to the structural protection optimization objective function, the spatial partition parameters, window size parameters and reinforcement arrangement parameters are constrained and optimized, and in the optimization process, the allowable range and design constraints of the number and size of spatial partitions, window size and position, reinforcement arrangement mode are limited according to the structural design specification and safety standard, and the optimal spatial partition scheme, optimal window size scheme and optimal reinforcement arrangement scheme with the lowest structural risk probability and reasonable structural protection cost are determined under the premise that all the constraints are met; The optimal spatial partition scheme is applied to the back wall structure, a cross-shaped embedded column beam structure is set up in the back wall structure to partition the space of the back wall, and the number and size of spatial partitions are determined according to the optimal spatial partition scheme; The optimal window size scheme is applied to the spatial partitions of the back wall structure, light windows are arranged in the upper spatial partitions of the cross-shaped embedded column beam structure, the size of the light windows is determined according to the optimal window size scheme, and the remaining spatial partitions are provided with closed walls or reinforcements; The optimal reinforcement arrangement scheme is applied to each spatial partition of the back wall structure, and the reinforcement is arranged in the specified spatial partition according to the reinforcement arrangement parameters; The comprehensive optimization design results form a complete optimization scheme of the back wall protective structure, including the optimal configuration of spatial partition, window size and reinforcement arrangement; According to the complete optimization scheme, a structural construction drawing is prepared to clearly define the size of each spatial partition, the size and position of each light window, and the arrangement requirements of each reinforcement, and a design file for engineering construction is generated; The prepared structural construction drawing and design file are applied to on-site structural construction to implement the optimization design of spatial partition, window size and reinforcement arrangement, and complete the engineering implementation of the back wall protective structure.

[0012] Optionally, the state characteristic parameters of the back wall partitions are continuously monitored during the operation stage of the house, the new monitoring data is periodically input into the sliding window transformer and the Bayesian network model to dynamically update the parameters, and the resilience adaptive optimization of the house protective structure is realized, including: In the house operation stage, the monitoring data of each space partition of the back wall is continuously collected to form a new round of space partition state characteristic parameter data set; The collected back wall space partition state characteristic parameter data set is standardized and preprocessed; The newly collected and preprocessed space partition state characteristic parameter data set is periodically input into the sliding window transformer for feature extraction and multi-scale feature updating; The space partition state characteristic parameters output by the sliding window transformer are periodically input into the Bayesian network model for dynamic updating of network node parameters and conditional probability distribution; Based on the dynamically updated Bayesian network model, the structural risk probability distribution and the overall structural resilience state of each space partition of the back wall are inferred and evaluated in real time; According to the inference and evaluation results, the space partition design, window size and reinforcement arrangement of the house protective structure are adaptively adjusted and optimized to improve the resilience and adaptability of the house protective structure, and the optimized protective scheme is implemented in the structure maintenance and operation management.

[0013] The beneficial effects of the present application are: The present application introduces the deep fusion of sliding window transformer and Bayesian network model to extract the state characteristics and infer the structural risk probability of each space partition of the back wall, realizes the dynamic integration of multi-source monitoring information and the quantitative evaluation of space partition risk. The structural optimization objective function is used to jointly optimize the space partition division, window size and reinforcement arrangement, which improves the scientificity of structural design and enhances the ability of protective structure to cope with complex disaster-causing environment and heterogeneous risk of partition. In the operation stage of the house, the monitoring data is periodically collected and processed, and continuously input into the sliding window transformer and the Bayesian network model to dynamically update the structural risk parameters, realize the resilience adaptive adjustment of the structural protection design, and effectively inhibit the expansion of local damage to overall failure. Compared with the existing structural protection method relying on single index or static design, the present application realizes the fine identification of partition-level risk and the dynamic optimization of structural protection parameters, improves the safety and reliability of house structure in landslide-induced environment, and has significant engineering application value. BRIEF DESCRIPTION OF DRAWINGS

[0014] The accompanying drawings are used to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation on the present application. In the drawings: Figure 1 The flow chart of the frame house protective structure optimization method under the action of cutting slope building induced landslide proposed by the present application; Figure 2A processing flow diagram of a sliding window transformer extracting spatial partition state characteristic parameters according to the slope cutting house landslide induced frame house protection structure optimization method. DETAILED DESCRIPTION

[0015] The application will now be described in further detail with reference to the drawings. These drawings are simplified schematic diagrams which only show the basic structure of the application in a schematic manner, and thus only show the components relevant to the application.

[0016] REFERENCE Figure 1 and Figure 2 The slope cutting house landslide induced frame house protection structure optimization method comprises the following steps: Collecting a multi-source spatio-temporal data set of a house rear slope and rear wall area, preprocessing the multi-source spatio-temporal data set, and generating a standardized multi-source spatio-temporal data set; Dividing the standardized multi-source spatio-temporal data set according to a preset spatial step and a time step to obtain a sliding window, using a sliding window transformer to extract features of data in each window, and obtaining state characteristic parameters of each spatial partition; Taking the state characteristic parameters of each spatial partition as node variables, establishing a Bayesian network model, inputting the state characteristic parameters for conditional probability reasoning, and obtaining structural risk probability distribution of each spatial partition of the rear wall; According to the structural risk probability distribution, optimizing the design of the structural partition, window size and reinforcement bar arrangement of the rear wall protection structure, setting up a cross-shaped embedded column beam structure in the rear wall to realize wall spatial partition, setting small size light windows in the upper two partitions of the cross-shaped structure, and setting closed walls or reinforcement bars in the remaining partitions, carrying out structural construction according to the optimization design, and implementing the structural partition, window size and reinforcement bar arrangement in the construction drawing and on-site implementation; During the operation stage of the house, the state characteristic parameters of the rear wall partition are continuously monitored, the new monitoring data is periodically input into the sliding window transformer and the Bayesian network model, the parameters are dynamically updated, and the resilience adaptive optimization of the house protection structure is realized.

[0017] The application realizes comprehensive perception of dynamic environment and structural state of the rear slope and rear wall area of the house by collecting and standardizing the multi-source spatio-temporal data set. The sliding window transformer is used to extract the partition characteristics of the standardized data, which improves the sensitivity to the state change of the spatial partition and the data utilization efficiency. The conditional probability inference is combined with the Bayesian network model to accurately evaluate the structural risk probability of each spatial partition of the rear wall, realize the quantification and partition identification of the risk. Based on the structural risk probability distribution, the protection structure parameters of the rear wall are optimized and designed, considering the reasonable configuration of the spatial partition, window size and reinforcement arrangement, which improves the structural safety and disaster prevention ability. During the operation stage of the house, the sliding window transformer and the Bayesian network model are dynamically adjusted through continuous monitoring and periodic data updating, so that the structural protection scheme can be automatically optimized according to the environmental change and actual risk, and the toughness and long-term reliability of the house structure are improved.

[0018] In the embodiment, the multi-source spatio-temporal data set specifically includes topographic and geological parameters of the rear slope and rear wall area of the house, rainfall duration, soil moisture content, wall stress parameters and structural layout information.

[0019] In the embodiment, the preprocessing of the multi-source spatio-temporal data set specifically includes format unification, missing value filling, outlier removal and data standardization processing of the multi-source spatio-temporal data set.

[0020] In the embodiment, the state characteristic parameters of each spatial partition are obtained, including: For each spatial partition and each time interval in the standardized multi-source spatio-temporal data set, the change rates of four types of monitoring parameters, i.e. ground deformation, strain, stress and soil moisture content, in the continuous time interval are calculated respectively, and the change rates of the four types of monitoring parameters are weighted and fused according to the importance weight of ground deformation, the importance weight of strain, the importance weight of stress and the importance weight of soil moisture content, to obtain the corresponding joint change perception index of each spatial partition and each time interval, to determine whether there is a sudden extreme change in the four types of monitoring parameters, i.e. ground deformation, strain, stress and soil moisture content, and to obtain the corresponding sudden change monitoring result of each spatial partition and each time interval. According to the joint change perception index and the sudden change monitoring result, each spatial partition and each time interval is adaptively divided by a sliding window, when the joint change perception index is higher than a preset threshold or the sudden change monitoring result is positive, a first spatial window size and a first time window step are used to divide the spatial partition and the time interval by the sliding window to obtain a first type of sliding window, when the joint change perception index is lower than or equal to the preset threshold and the sudden change monitoring result is negative, a second spatial window size and a second time window step are used to divide the spatial partition and the time interval by the sliding window to obtain a second type of sliding window, the first spatial window size and the first time window step are preset fine-grained parameters, and the second spatial window size and the second time window step are preset wide-scale parameters, according to the spatial position continuity and the time sequence continuity principle, all pairs of directly adjacent sliding windows in the spatial or time domain are identified, and a set of adjacent sliding windows is established; For the first type of sliding window and the second type of sliding window, global statistical features, local extreme value features, change rate features and segmented interval features in the sliding window are extracted according to spatial scales and time scales respectively to generate a multi-scale feature subset; The extracted global statistical features, local extreme value features, change rate features and segmented interval features are spliced and fused by weighting to generate a multi-scale feature vector of each sliding window; For each pair of directly adjacent sliding windows in the set of adjacent sliding windows, a sliding window overlap strategy is performed, on the basis of maintaining the original window boundary sequence, the boundaries between the adjacent sliding windows are extended inward according to a set overlap length to form a set of pairs of sliding windows with data overlap regions in the spatial position or the time sequence; For the overlapping region data corresponding to each pair of sliding windows in the set of pairs of sliding windows, the multi-scale feature vectors of the sliding window boundaries are trend extrapolated by using the change trend of the overlapping region data, the multi-scale feature vectors generated by the trend extrapolation are fused with the multi-scale feature vectors of the original overlapping region in a Bayesian smoothing weight manner to obtain multi-scale feature vectors after boundary continuity processing and abnormal correction, and if a feature mutation or an abnormal residual error occurs in the overlapping region data, the multi-scale feature vectors after the boundary continuity processing and the abnormal correction and the multi-scale feature vectors of the corresponding adjacent sliding windows are modified by using a residual error correction method; The multi-scale feature vectors and the multi-scale feature vectors after the boundary continuity processing and the abnormal correction are summarized to generate multi-dimensional state feature parameters corresponding to each spatial partition and each time interval, and state feature parameters of each spatial partition are obtained.

[0021] The application comprehensively reflects the actual deformation and risk state of each spatial partition and time interval by calculating and weighting the dynamic change rate of multi-dimensional monitoring parameters of the standardized multi-source spatio-temporal data set, and improves the sensitivity and accuracy of abnormal change identification. The adaptive sliding window division and window category distinction are adopted to realize fine analysis of high-risk sections and efficient processing of low-risk sections, and the data utilization efficiency and analysis granularity are improved. The multi-scale extraction and fusion of global statistical features, local extreme value features, change rate features and segmented interval features enhance the richness and comprehensiveness of feature expression. The sliding window overlap, trend extrapolation, Bayesian smoothing and residual correction mechanism improve the window boundary continuity and feature stability under abnormal interference. The multi-scale features and the corrected feature parameters ensure that the state feature parameters of each spatial partition are accurate, comprehensive and robust.

[0022] In the embodiment, the structure risk probability distribution of each spatial partition of the rear wall is obtained, including: Taking the obtained state feature parameters of each spatial partition as candidate node variables, the spatial partition node structure is dynamically divided according to the spatial distribution characteristics and change trend of the state feature parameters, and a spatial partition node set is determined; According to the abnormal change and risk distribution difference of the state feature parameters, the spatial partition node set is determined, the spatial partition node with state mutation is subdivided into multiple spatial partition sub-nodes, the spatial partition node with high state correlation or risk balance is merged into a single spatial partition node, and a spatial partition node set is obtained; Taking each spatial partition node in the spatial partition node set as an input node, taking four types of monitoring parameters of ground deformation, strain, stress and soil moisture content as external disaster-causing factor nodes, and presetting an overall structure influence node in a multi-layer Bayesian network model structure as an output node of the overall structure failure probability, a multi-layer Bayesian network model structure containing spatial partition nodes, external disaster-causing factor nodes and overall structure influence nodes is constructed; For each node in the multi-layer Bayesian network model structure, based on historical monitoring data, real-time acquisition data and external disaster-causing environmental data, the conditional probability distribution of each node is regularly updated to form a multi-source data-driven conditional probability table set; The conditional probability table set and the state feature parameters of each spatial partition node in the spatial partition node set are input, and the Bayesian inference method is used to find the corresponding state parameter combination in the conditional probability table, read the conditional probability of local damage of the spatial partition node under the state parameter combination, and if there are multiple parent nodes or multiple observation information, the probability weighted sum and normalization of all parent node states are sequentially performed according to the probability propagation mechanism, to obtain the local damage probability of each spatial partition node under the current observation information; The local damage probability of each spatial partition node in the spatial partition node set is calculated by using a probability joint inference function to obtain the overall structure failure probability of the overall structure influence node: ; wherein, is the overall structure failure probability, is the total number of spatial partition nodes, is the local damage probability of the i-th spatial partition node, is the local damage probability of the j-th spatial partition node, is the synergistic coupling influence factor between the i-th and j-th spatial partition nodes, is the influence weight of the i-th spatial partition node; The overall structure failure probability comprehensively considers the influence of the local damage risk of each spatial partition on the global safety of the structure. The safety of the overall structure depends not only on the probability of individual destruction of each partition, but also on the importance of the role of each partition in the structure and the synergistic destruction effect between partitions. By introducing the influence weight, the dominant role of the key partition on the structure safety is highlighted. The synergistic coupling factor considers the superposition effect of the structure as a whole when multiple partitions are damaged at the same time, breaking through the simple set calculation of traditional independent partition probability, and more truly reflecting the overall failure mechanism of complex structure systems under the conditions of local damage transmission and interaction between partitions, thereby providing a scientific basis for structure safety evaluation and protection optimization.

[0023] Based on the multi-layer Bayesian network model structure and the local damage probability of each spatial partition node in the spatial partition node set, the correlation between the spatial partition nodes is analyzed. Network edges are established between spatial partition nodes with significant correlation, and network edges between spatial partition nodes with low correlation are deleted to obtain a dynamically adjusted Bayesian network model structure. The local damage probability of each spatial partition node in the spatial partition node set, the overall structure failure probability of the overall structure influence node, and the structure risk probability distribution corresponding to the dynamically adjusted Bayesian network model structure are output to obtain the structure risk probability distribution of each spatial partition of the back wall.

[0024] The application accurately reflects the change trend and risk distribution of the structure partition state by dynamic division and adaptive adjustment of the spatial partition node structure, and realizes fine identification and modeling of partition risk. In combination with multiple source monitoring parameters such as ground deformation, strain, stress, soil moisture content and the like, a multi-layer Bayesian network model is established, the spatial partition node, external disaster-causing factor node and overall structure influence node are organically integrated, and the systematicness and scientificity of structure risk reasoning are improved. Through dynamic updating of the conditional probability table driven by multiple source data, the synchronization of the model parameters and the actual engineering state is ensured, and the precise response capability for different risk scenarios is provided. By using Bayesian inference and probability propagation mechanism, the local damage probability of each spatial partition node can be calculated in real time, and the overall structure failure probability can be comprehensively reflected based on the probability joint inference function. The introduction of the coupling factor and the partition weight effectively strengthens the comprehensive influence expression of the multi-partition linkage damage on the overall structure safety. The correlation self-learning optimization of the network structure improves the accuracy of the dependence relationship description between partitions. The application provides high-quality partition risk quantitative evaluation and overall safety judgment for structure protection, enhances the intelligence and adaptability of structure protection design, and improves the safety guarantee level of house structure in landslide and other disaster-induced scenarios.

[0025] In the embodiment, according to the structure risk probability distribution, the structure partition, window size and reinforcement arrangement of the back wall protection structure are optimized and designed, the inner column beam structure of a field shape is set in the back wall, the wall space partition is realized, the small size light window is arranged in the upper two partitions of the field structure, the remaining partitions are provided with closed walls or reinforcement, the structure partition, window size and reinforcement arrangement are implemented in the construction drawing and the field implementation according to the optimization design, and the structure construction is carried out, including: According to the structure risk probability distribution of each spatial partition, in combination with the spatial partition parameters, window size parameters and reinforcement arrangement parameters of the back wall structure, a structure protection optimization objective function is established: ; Among them, represents the spatial partition parameter, represents the window size parameter, represents the reinforcement arrangement parameter, is the structure risk probability of the first spatial partition, is a structure protection cost function, and are weight coefficients, is the total number of spatial partitions, and min is a minimum value operation; The structural protection optimization objective function reflects the goal of simultaneously considering structural risk control and engineering cost optimization in structural protection design. By optimizing the combination of spatial zoning parameters, window size parameters, and stiffener arrangement parameters, the structural risk probability of each spatial zone is effectively reduced, and the investment cost of structural protection is reasonably controlled. Weighting coefficients are used to balance the relative importance of structural safety and economy. The optimization result satisfies both safety performance requirements and economic feasibility. The optimization objective is to find the optimal solution among all feasible parameter configurations, minimizing the weighted sum of structural risk and engineering cost. This breaks away from the traditional single-index optimization model and provides a scientific, systematic, and engineering-guiding parameter configuration scheme for structural protection.

[0026] Based on the structural protection optimization objective function, the spatial partition parameters, window size parameters, and stiffener arrangement parameters are constrained and optimized. During the optimization process, the allowable range and design constraints of the number and size of spatial partitions, window size and location, and stiffener arrangement are limited according to structural design specifications and safety standards. Under the premise that all constraints are met, the optimal spatial partition scheme, optimal window size scheme, and optimal stiffener arrangement scheme with the lowest structural risk probability and reasonable structural protection cost are determined. The optimal space zoning scheme is applied to the rear wall structure. A grid-shaped embedded column-beam structure is set up in the rear wall structure to divide the rear wall space into zones. The number and size of the space zones are determined according to the optimal space zoning scheme. The optimal window size scheme is applied to the spatial partitioning of the rear wall structure. Lighting windows are set in the upper spatial partition of the grid-shaped embedded column-beam structure. The size of the lighting windows is determined according to the optimal window size scheme. The remaining spatial partitions are arranged with closed walls or reinforcing ribs. The optimal stiffener arrangement scheme is applied to each spatial zone of the rear wall structure, and stiffeners are set in the specified spatial zones according to the stiffener arrangement parameters. Based on the comprehensive optimization design results, a complete optimized scheme for the rear wall protection structure was formed, including the optimal configuration of space zoning, window size, and reinforcement arrangement; Based on the complete optimization plan, structural construction drawings are prepared, specifying the dimensions of each spatial partition, the dimensions and location of each skylight, and the arrangement requirements of each reinforcing rib, generating design documents for engineering construction; The prepared structural construction drawings and design documents were applied to the on-site structural construction to implement the optimized design of space zoning, window size and reinforcement arrangement, and then the engineering implementation of the wall protection structure was completed.

[0027] The application realizes systematic joint design of structural parameters such as spatial partition, window size and reinforcement arrangement by establishing an optimal objective function through structural protection and multi-parameter constraint optimization. The structural risk probability distribution is organically combined with the protection cost to effectively control the engineering cost while ensuring the safety of the structure, thereby improving the economic rationality of the protection design. According to the risk difference of spatial partition, the rear wall structure adopts a cross-shaped embedded column beam for spatial partition, thereby realizing fine control of the partition size and layout and enhancing the overall bearing capacity and stability of the structure. The optimal window size scheme and the coordinated design of reinforcement arrangement improve the stress uniformity and lighting performance of the structure. The construction drawings and design documents are strictly prepared according to the optimization results throughout the whole process, thereby ensuring efficient implementation of the design scheme in the field.

[0028] In the present embodiment, the state characteristic parameters of the rear wall partitions are continuously monitored during the operation stage of the house, and the new monitoring data is periodically input into the sliding window transformer and the Bayesian network model to dynamically update the parameters, thereby realizing the resilience adaptive optimization of the house protection structure, including: During the operation stage of the house, the monitoring data of each spatial partition of the rear wall is continuously collected to form a new round of spatial partition state characteristic parameter data set; The collected rear wall spatial partition state characteristic parameter data set is standardized and preprocessed; The newly collected and preprocessed spatial partition state characteristic parameter data set is periodically input into the sliding window transformer for feature extraction and multi-scale feature updating; The spatial partition state characteristic parameters output by the sliding window transformer are periodically input into the Bayesian network model to dynamically update the network node parameters and conditional probability distribution; Based on the dynamically updated Bayesian network model, the structural risk probability distribution of each spatial partition of the rear wall and the overall structural resilience state are inferred and evaluated in real time; According to the inference and evaluation results, the protection scheme of the spatial partition design, window size and reinforcement arrangement of the house protection structure is adaptively adjusted and optimized to improve the resilience and adaptability of the house protection structure, and the optimized protection scheme is implemented in the structure maintenance and operation management.

[0029] The application realizes real-time perception and dynamic data accumulation of the structural health state by continuously collecting and preprocessing the monitoring data of each spatial partition of the rear wall in the operation stage of the house. The sliding window transformer is used to extract and update the multi-scale features of the new data, improving the response ability of the partition state feature parameters to environmental and working condition changes. The latest feature parameters are periodically input into the Bayesian network model, and the model node parameters and conditional probability distribution are dynamically adjusted to reflect the latest state of the structure stress and environmental risk. Based on the dynamically updated Bayesian network model, the structural risk probability distribution and overall structural resilience of each spatial partition of the rear wall are inferred and evaluated in real time, providing data support for the adaptive optimization of the structure protection scheme. According to the inference results, the key parameters such as spatial partition design, window size and reinforcement arrangement of the house protection structure can be automatically adjusted and applied to maintenance and operation management, enhancing the resilience and environmental adaptability of the structure protection. Embodiment

[0030] In order to verify the feasibility of the application in implementation, the application is applied to a mountain residential area, and the developer plans to build multiple three-story houses on a south-facing steep slope. The site has undulating terrain, with a 30-degree rear slope. The geology is mainly silty clay with pebbles, and it is affected by heavy rain all year round. Due to the problems of traditional experience-based rear wall design in similar geological disaster-prone areas, such as rear wall deformation, water seepage and even local structure damage, the construction unit hopes to solve the problems of rear wall safety and resilience through advanced structure optimization methods.

[0031] During the implementation of this project, the team selected one of the C2 residential buildings and used the method of the application to optimize the rear wall protection structure throughout the cycle. At the beginning of the project, technicians first laid out multi-source monitoring sensors along the rear slope and rear wall of the house, including 6 ground deformation monitoring points, 8 wall strain points, 8 wall stress points and 4 soil moisture monitoring points, which automatically collected data every 30 minutes. The project period lasted from May 2024 to April 2025, spanning a complete flood season.

[0032] After the collected monitoring data is standardized and preprocessed, it is input into the sliding window transformer, which automatically divides it into adaptive sliding windows for different spatial partitions and time intervals. The system automatically extracts the ground deformation rate, strain rate, stress rate and moisture content dynamic characteristics of each partition at different times. In mid-June 2022, Chongqing experienced consecutive heavy rains, and the ground deformation rate of the rear wall B zone reached 0.78 mm / d on a certain day, and the strain rate increased significantly. The system monitored that the joint change perception index of B zone and C zone was higher than the preset threshold, triggering the division of fine-grained sliding windows.

[0033] The engineering team constructed a Bayesian network model using the multi-scale state feature parameters extracted by the sliding window transformer, and included surface deformation, strain, stress, and water content as external disaster-causing nodes in the reasoning system. The Bayesian model calculates the local failure probability and overall structural failure probability of each partition in real time based on the latest data. For example, on the day of the storm on June 16, 2024, the local failure probability of B zone under the original structural plan was as high as 0.28, and that of C zone was 0.22, with an overall structural failure probability of 0.17.

[0034] According to the structural optimization objective function, the system automatically iterates the partition scheme of the rear wall space, the size of the window, and the layout scheme of the reinforcing rib multiple times. Finally, the number of partitions is adjusted from the original 4 to 6, the upper part uses two small windows of 0.7m x 0.6m, the lower part and both sides are all closed walls, and all partitions are increased with double-layer reinforcing ribs. The optimized protective structure scheme ensures safety while only increasing the engineering cost by 6.4%.

[0035] After the implementation of the optimized design, continuous monitoring from July 2024 to April 2025 shows that the maximum partition failure probability of the rear wall has decreased to 0.13, and the overall structural failure probability has decreased to 0.06. During the same period, the maximum deformation rate has decreased from 0.69 mm / d in the original scheme to 0.31 mm / d, and the cumulative number of stress overruns during the rainy season has decreased from 14 days to 2 days. After the construction is completed, residents have provided positive feedback, and there have been no further instances of rear wall seepage and cracking on site.

[0036] Table 1 Comparison of structural safety performance before and after optimization

[0037] The above Table 1 shows the comparison of main technical indexes and engineering effects before and after the optimization of the structure protection of the application. After optimization, the number of rear wall space partitions is increased from 4 to 6, realizing more detailed partition management and risk sharing. The size of the upper light window is reduced from 1.2x0.8 meters to 0.7x0.6 meters, effectively limiting the weak interval of the window opening part and improving the overall strength of the wall. The number of reinforcing ribs is increased from 1 per partition to 2, enhancing the local deformation resistance. The structural optimization reduces the risk-related indicators, with the maximum partition damage probability reduced from 0.28 to 0.13, the overall structural failure probability reduced from 0.17 to 0.06, and the safety margin of the structure greatly improved. The maximum deformation rate is also greatly reduced, indicating that the controllability of wall deformation under extreme rainfall and other disaster environments has been improved. The cumulative number of stress overruns in the rainy season is reduced from 14 days to 2 days, and the long-term reliability of the structure is enhanced. Although the cost of the protective structure increases slightly after optimization, only increasing 0.45 thousand yuan, but in return, the number of reports of rear wall water seepage and cracking is reduced from 6 to 0, and the residents' satisfaction is increased from 78% to 94%. The optimization measures of the application realize the coordination and unity of structural safety, functionality and economy, and improve the practicality of the housing protection structure and the residents' living confidence under the environment of building houses on mountain cutting slopes.

[0038] The above is only the preferred specific embodiment of the application, but the protection scope of the application is not limited thereto, and any person skilled in the art can make equivalent replacements or changes to the technical range disclosed in the application according to the technical solution and inventive concept of the application, which should be covered within the protection scope of the application.

Claims

1. A method for optimizing the protective structure of a frame house under the action of a landslide induced by building on a slope, characterized in that, The method comprises the following steps: Collecting a multi-source spatio-temporal data set of the rear slope and rear wall area of a house, preprocessing the multi-source spatio-temporal data set, and generating a standardized multi-source spatio-temporal data set; Dividing the standardized multi-source spatio-temporal data set into sliding windows according to a preset spatial step and a time step, using a sliding window transformer to extract features of data in each window, and obtaining state feature parameters of each spatial partition; Taking the state feature parameters of each spatial partition as node variables, establishing a Bayesian network model, inputting the state feature parameters for conditional probability reasoning, and obtaining a structural risk probability distribution of each spatial partition of the rear wall; According to the structural risk probability distribution, optimizing the structural partition, window size and reinforcement bar arrangement of the rear wall protective structure, setting up a cross-shaped embedded column-beam structure in the rear wall to realize spatial partition of the wall, setting small-size light windows in the upper two partitions of the cross-shaped structure, and setting closed walls or reinforcement bars in the remaining partitions, implementing structural construction according to the optimization design, and implementing the structural partition, window size and reinforcement bar arrangement in the construction drawing and on-site implementation; During the operation stage of the house, the state feature parameters of the rear wall partition are continuously monitored, the new monitoring data is periodically input into the sliding window transformer and the Bayesian network model, the parameters are dynamically updated, and the resilience adaptive optimization of the house protective structure is realized.

2. The method according to claim 1, wherein, The multi-source spatio-temporal data set specifically includes topographic and geological parameters, rainfall duration, soil moisture content, wall stress parameters and structural layout information of the rear slope and rear wall area of the house.

3. The method according to claim 1, wherein the method is characterized by, The preprocessing of the multi-source spatio-temporal data set specifically includes format unification, missing value filling, outlier removal and data standardization processing of the multi-source spatio-temporal data set.

4. The method of claim 1, wherein the method is characterized by: The state feature parameters of each spatial partition are obtained, including: For each spatial partition and each time interval in the standardized multi-source spatio-temporal data set, the change rates of four types of monitoring parameters, i.e., ground deformation, strain, stress and soil moisture content, in the continuous time interval are calculated, and the change rates of the four types of monitoring parameters are weighted and fused according to the importance weight of ground deformation, the importance weight of strain, the importance weight of stress and the importance weight of soil moisture content, to obtain a joint change perception index corresponding to each spatial partition and each time interval, to determine whether there is a sudden extreme change in the four types of monitoring parameters, and to obtain a sudden change monitoring result corresponding to each spatial partition and each time interval; According to the joint change perception index and the sudden change monitoring result, each spatial partition and each time interval is adaptively divided by a sliding window, when the joint change perception index is higher than a preset threshold or the sudden change monitoring result is positive, a first spatial window size and a first time window step are used to divide the spatial partition and the time interval by the sliding window to obtain a first type of sliding window, when the joint change perception index is lower than or equal to the preset threshold and the sudden change monitoring result is negative, a second spatial window size and a second time window step are used to divide the spatial partition and the time interval by the sliding window to obtain a second type of sliding window, the first spatial window size and the first time window step are preset fine-grained parameters, and the second spatial window size and the second time window step are preset wide-scale parameters, according to the spatial position continuity and the time sequence continuity principle, all pairs of directly adjacent sliding windows in the spatial or time domain are identified, and a set of adjacent sliding windows is established; For the first type of sliding window and the second type of sliding window, global statistical features, local extreme value features, change rate features and segmented interval features in the sliding window are extracted according to spatial scales and time scales to generate a multi-scale feature subset; The extracted global statistical features, local extreme value features, change rate features and segmented interval features are spliced and fused by weighting to generate a multi-scale feature vector of each sliding window; For each pair of directly adjacent sliding windows in the set of adjacent sliding windows, a sliding window overlap strategy is performed, the boundaries between the adjacent sliding windows are extended inward according to a preset overlap length on the basis of maintaining the original window boundary sequence, and a set of pairs of sliding windows with data overlap regions in the spatial position or the time sequence is formed; For the overlapping region data corresponding to each pair of sliding windows in the set of pairs of sliding windows, the multi-scale feature vectors of the sliding window boundaries are trend extrapolated by using the change trend of the overlapping region data, the multi-scale feature vectors generated by the trend extrapolation are fused with the multi-scale feature vectors of the original overlapping region in a Bayesian smoothing weight manner, and the multi-scale feature vectors after boundary continuity processing and abnormal correction are obtained, if a feature mutation or an abnormal residual error occurs in the overlapping region data, the multi-scale feature vectors after the boundary continuity processing and the abnormal correction and the multi-scale feature vectors of the corresponding adjacent sliding windows are modified by using a residual error correction method; The multi-scale feature vectors and the multi-scale feature vectors after the boundary continuity processing and the abnormal correction are summarized to generate multi-dimensional state feature parameters corresponding to each spatial partition and each time interval, and state feature parameters of each spatial partition are obtained.

5. The method of claim 1, wherein the method is characterized by: The structural risk probability distribution of each spatial partition of the back wall is obtained, including: The state feature parameters of each spatial partition obtained are used as candidate node variables, and the spatial partition node structure is dynamically divided according to the spatial distribution characteristics and the change trend of the state feature parameters to determine a set of spatial partition nodes. According to the abnormal change of the state characteristic parameters and the risk distribution difference, the space partition node set is divided into a plurality of space partition sub-nodes, and the space partition node set is obtained by combining the space partition nodes with high state correlation or balanced risk into a single space partition node; Taking each space partition node in the space partition node set as an input node, taking four types of monitoring parameters of ground deformation, strain, stress and soil moisture content as external disaster-causing factor nodes, and presetting an overall structure influence node in the multi-layer Bayesian network model structure as an output node of the overall structure failure probability, a multi-layer Bayesian network model structure including the space partition node, the external disaster-causing factor node and the overall structure influence node is constructed; For each node in the multi-layer Bayesian network model structure, the conditional probability distribution of each node is regularly updated based on historical monitoring data, real-time acquisition data and external disaster environment data, and a multi-source data driven conditional probability table set is formed; The conditional probability table set and the state characteristic parameters of each space partition node in the space partition node set are input, and the corresponding state parameter combination in the conditional probability table is searched by using the Bayesian inference method, the conditional probability of the local damage of the space partition node under the state parameter combination is read, if there are multiple parent nodes or multiple observation information, the probability weighted sum and normalization of all parent node states are sequentially performed according to the probability propagation mechanism, and the local damage probability of each space partition node under the current observation information is obtained; The overall structure failure probability of the overall structure influence node is calculated by using a probability joint inference function for the local damage probability of each space partition node in the space partition node set; Based on the multi-layer Bayesian network model structure and the local damage probability of each space partition node in the space partition node set, the correlation between the space partition nodes is analyzed, the network edge connection between the space partition nodes with significant correlation is established, the network edge connection between the space partition nodes with low correlation is deleted, and a dynamically adjusted Bayesian network model structure is obtained; The local damage probability of each space partition node in the space partition node set, the overall structure failure probability of the overall structure influence node, and the structure risk probability distribution corresponding to the dynamically adjusted Bayesian network model structure are output, and the structure risk probability distribution of each space partition of the back wall is obtained.

6. The method of claim 1, wherein the method is characterized by: According to the structure risk probability distribution, the structure partition, window size and reinforcement arrangement of the back wall protective structure are optimized and designed, the inner column beam structure of the field-shaped structure is set up in the back wall, the wall space partition is realized, the small size lighting window is arranged in the upper two partitions of the field-shaped structure, the remaining partitions are arranged with closed walls or reinforcement, the structure partition, window size and reinforcement arrangement are implemented in the construction drawing and on-site implementation, including: According to the structure risk probability distribution of each space partition, the space partition parameters, window size parameters and reinforcement arrangement parameters of the back wall structure are combined to establish a structure protection optimization objective function; Based on the structural protection optimization objective function, the spatial partition parameters, window size parameters, and stiffener arrangement parameters are constrained and optimized. During the optimization process, the allowable range and design constraints of the number and size of spatial partitions, window size and location, and stiffener arrangement are limited according to structural design specifications and safety standards. Under the premise that all constraints are met, the optimal spatial partition scheme, optimal window size scheme, and optimal stiffener arrangement scheme with the lowest structural risk probability and reasonable structural protection cost are determined. The optimal space zoning scheme is applied to the rear wall structure. A grid-shaped embedded column-beam structure is set up in the rear wall structure to divide the rear wall space into zones. The number and size of the space zones are determined according to the optimal space zoning scheme. The optimal window size scheme is applied to the spatial partitioning of the rear wall structure. Lighting windows are set in the upper spatial partition of the grid-shaped embedded column-beam structure. The size of the lighting windows is determined according to the optimal window size scheme. The remaining spatial partitions are arranged with closed walls or reinforcing ribs. The optimal stiffener arrangement scheme is applied to each spatial zone of the rear wall structure, and stiffeners are set in the specified spatial zones according to the stiffener arrangement parameters. Based on the comprehensive optimization design results, a complete optimized scheme for the rear wall protection structure was formed, including the optimal configuration of space zoning, window size, and reinforcement arrangement; Based on the complete optimization plan, structural construction drawings are prepared, specifying the dimensions of each spatial partition, the dimensions and location of each skylight, and the arrangement requirements of each reinforcing rib, generating design documents for engineering construction; The prepared structural construction drawings and design documents were applied to the on-site structural construction to implement the optimized design of space zoning, window size and reinforcement arrangement, and then the engineering implementation of the wall protection structure was completed.

7. The method of claim 1, wherein the method is characterized by: During the building's operation phase, the state characteristic parameters of the rear wall zones are continuously monitored. New monitoring data are periodically input into a sliding window transformer and a Bayesian network model to dynamically update the parameters, achieving adaptive optimization of the building's protective structure's resilience, including: During the building's operation phase, monitoring data from each spatial zone of the rear wall are continuously collected to form a new set of spatial zone status characteristic parameters. The collected dataset of spatial zoning characteristic parameters of the rear wall is standardized and preprocessed. The newly acquired and preprocessed spatial partition state feature parameter dataset is periodically input into the sliding window transformer for feature extraction and multi-scale feature update. The spatial partition state feature parameters output by the sliding window transformer are periodically input into the Bayesian network model to dynamically update the network node parameters and conditional probability distribution. Based on the dynamically updated Bayesian network model, the structural risk probability distribution of each spatial partition of the rear wall and the overall structural toughness status are inferred and evaluated in real time. Based on the reasoning and evaluation results, the spatial zoning design, window size, and reinforcement arrangement of the building's protective structure are adaptively adjusted and optimized to improve the resilience and adaptability of the building's protective structure. The optimized protective scheme is then implemented in structural maintenance and operation management.

Images