Building vulnerability assessment method under action of debris flow

Abstract

This invention provides a method for assessing building vulnerability under debris flow, belonging to the field of geological hazard risk assessment technology. First, based on a database of debris flow-damaged buildings, this invention constructs a building vulnerability function and plots vulnerability curves to clarify the correlation between hazard intensity and building damage probability. Second, it establishes a system of building vulnerability offset influencing factors and uses a random forest algorithm to construct a calculation model of the offset between actual and theoretical building vulnerability. Finally, by combining the vulnerability function and the offset calculation model, it achieves accurate prediction of building vulnerability. Traditional research using vulnerability functions struggles to accurately quantify the offset characteristics between damaged building data points and vulnerability curves, leading to insufficient reliability of assessment results. This invention overcomes this deficiency.

Description

Technical Field

[0001] This invention relates to the field of geological hazard assessment technology, and in particular to a method for assessing the vulnerability of buildings under debris flow. Background Technology

[0002] Debris flows are characterized by their wide distribution and sudden occurrence. They often destroy towns and rural settlements, damage roads, bridges and engineering facilities, block river channels and reservoirs, and bury farmland and forests through impact, erosion, and sedimentation. They cause enormous casualties, property losses, and ecological damage, seriously threatening the lives and property of people in mountainous areas and hindering the development of mountain resources and the economy. They are one of the most frequent and devastating geological disasters globally. According to the first national comprehensive natural disaster risk survey bulletin, natural disaster risk assessment follows the three-dimensional criteria of "hazard-exposure-vulnerability." Vulnerability assessment, as an indispensable key component of the risk assessment system, analyzes and evaluates the physical structure and functional characteristics of buildings, as well as potential surrounding hazards, to reveal the extent and scope of damage that may occur when threatened by a disaster. This is crucial for achieving accurate risk assessment.

[0003] Currently, while the theoretical and methodological framework for vulnerability assessment is relatively complete, certain limitations still exist. The vulnerability function and vulnerability curve are refined but affected by sample quality and scenario differences. Furthermore, existing research has not sufficiently explored the sources of uncertainty, quantification methods, and correction approaches in the curve construction process.

[0004] Against the backdrop of a complex coupling between significant climate change and intensified human activities, debris flows are becoming increasingly frequent and have a more severe impact on residents' lives and livelihoods. There is an urgent need for accurate vulnerability assessment of buildings, and for machine learning methods to quantify the degree of uncertainty, analyze the discrete characteristics and offset patterns of data points of damaged buildings, reveal the influence mechanism of factors such as the characteristics of the buildings themselves and the differences in the channels on the vulnerability assessment results, and achieve accurate prediction and analysis of the offset, thereby completing the correction of the vulnerability assessment results. Summary of the Invention

[0005] To address the aforementioned shortcomings of existing technologies, this invention provides a method for assessing building vulnerability under debris flow conditions, enabling accurate prediction of building vulnerability. This method aims to overcome the uncertainties inherent in traditional research relying on a single vulnerability curve—the difficulty in accurately quantifying the deviation between the actual vulnerability of a damaged building and the vulnerability curve—thereby improving the reliability of assessment results caused by the aforementioned problems.

[0006] To achieve the above objectives, the technical solution adopted by this invention is: a method for assessing the vulnerability of buildings under debris flow, comprising the following steps: S1. Debris Flow Damaged Building Database Construction: Obtain debris flow characteristic parameters and building damage status of buildings damaged by debris flows, and construct a debris flow damaged building database. S2. Construction of Building Vulnerability Function: Based on the constructed database of debris flow-damaged buildings, using debris flow characteristic parameters and building damage level as independent variables, a building vulnerability function is established based on the Logistic function. V ; S3. Construction of the building vulnerability offset influencing factor system: Select multiple influencing factors, conduct factor correlation analysis, eliminate strongly correlated factors, and establish the building vulnerability offset influencing factor system. S4. Building Vulnerability Offset Modeling: Based on Building Vulnerability Function V The theoretical vulnerability value of the damaged building is calculated, and based on the building vulnerability offset influence factor system, a random forest algorithm is used to obtain the building vulnerability offset calculation model. S5. Accurate prediction of building vulnerability: Input new building sample data and use the vulnerability offset model to output the building vulnerability prediction, thus completing the assessment of building vulnerability.

[0007] The beneficial effects of this invention are as follows: This invention improves the accuracy and completeness of basic data by constructing a database of debris flow-damaged buildings, and on this basis, establishes a building vulnerability function and plots vulnerability curves. It establishes a system of factors influencing building vulnerability shift, and uses a random forest algorithm to analyze and suggest vulnerability shifts (the difference between actual and theoretical vulnerability), constructing a building vulnerability shift calculation model. By integrating the vulnerability function and the shift prediction model, accurate prediction of building vulnerability is achieved. This invention overcomes the problem that traditional studies often use single methods to characterize uncertainty, making it difficult to accurately quantify the shift characteristics of damaged building data points and vulnerability curves, resulting in insufficient reliability of assessment results. This technical solution, based on the establishment of damaged building vulnerability, further conducts quantitative analysis on uncertainty issues, thereby providing a scientific basis for accurate prediction of building vulnerability, debris flow risk assessment, and disaster prevention and mitigation.

[0008] Further, S1 includes: The mud depth of the damaged building was obtained by combining on-site measurement and 3D laser modeling. Calculate the flow velocity of the debris flow and the impact pressure of the debris flow; Debris flow characteristic parameters are obtained based on mud depth, debris flow velocity, and debris flow impact pressure. The damage level of a building is classified to obtain information about the extent of the damage. A database of buildings damaged by debris flows was constructed based on debris flow characteristic parameters and building damage conditions.

[0009] The beneficial effects of the above-mentioned further solutions are: the present invention improves the accuracy and completeness of basic data by constructing a database through on-site investigation combined with measurement and multi-source data integration; and standardizes the calculation of flow velocity, impact pressure and building damage level, providing reliable and standardized basic parameters for subsequent assessment.

[0010] Furthermore, S2 includes: Based on the debris flow-damaged building database, data on damaged buildings within a confidence interval of a preset threshold are sampled. Using debris flow characteristic parameters and building damage level as independent variables, a vulnerability curve for buildings was established by fitting an S-curve using the Logistic function. V :

[0011] in, a and b Both represent constants. x These represent the characteristic parameters of debris flows.

[0012] The beneficial effects of the above-mentioned further solutions are: by screening reliable samples to fit vulnerability curves, this invention adapts to different building structural characteristics and initially establishes a quantitative analysis of the influence of debris flow characteristic parameters on building vulnerability.

[0013] Furthermore, S3 includes: Multiple dimensions of building vulnerability offset influencing factors were selected, including the building's own physical properties, geographical features, spatial location, and disaster characteristics. Extract the building vulnerability offset influence factor values ​​corresponding to each building sample, and normalize them to obtain standardized factor data; The correlation between factors was analyzed by statistically analyzing the normalized factor data using the Pearson correlation coefficient. Based on the correlation analysis results, strongly correlated factors were eliminated, and a system of factors affecting building vulnerability shift was established.

[0014] The beneficial effects of the above-mentioned further solutions are: the present invention selects multiple influencing factors (such as 18) from four dimensions, eliminates strongly correlated factors through Pearson correlation coefficient analysis, and optimizes to obtain a reasonable building vulnerability offset influencing factor system.

[0015] Furthermore, S4 includes: Based on the vulnerability function of buildings V Calculate the theoretical vulnerability value of the damaged building; The difference between the theoretical vulnerability of the damaged building and the actual vulnerability defined on-site is calculated to obtain the building vulnerability offset:

[0016] in, Indicates the first Building vulnerability offset for each sample Indicates the first The actual building vulnerability value for each sample. Indicates based on the first The theoretical building vulnerability value for each sample; Based on building vulnerability offset And the vector of building vulnerability offset influence factors Construct a modeling sample set; Using the modeling sample set, the building vulnerability offset is... As the dependent variable, the vulnerability offset factors of each building are used as independent variables to train the random forest regression model; During the sample sampling phase, a bootstrap sampling method is used to randomly sample the sample set containing the building vulnerability offset and the building vulnerability offset influence factor value with replacement, generating a training subset for each decision tree that is consistent with the original damaged building sample dataset and has sample overlap. Decision tree models are trained using each training subset. In the process of constructing a single decision tree, some features are randomly selected from the building vulnerability offset influence factor system as candidate features. The values ​​of each candidate feature are used as the splitting threshold, and the criterion for determining the optimal offset model is the minimum sum of the mean square errors of the child nodes after splitting.

[0017] in, Indicates the first Building vulnerability offset for each sample An index representing the building vulnerability offset influence factor. Indicating the influence factor of building vulnerability offset The splitting threshold, and These represent the impact factors respectively. and threshold The divided left and right child node regions, and This represents the mean value of the building vulnerability offset within the corresponding child node. Indicates the first The feature vector of building vulnerability offset influence factor corresponding to each sample Indicates the first The first sample One influencing factor characteristic value; By selecting the optimal features and corresponding splitting thresholds as node splitting rules, and combining them with pre-set hyperparameters to form a random forest, a model for calculating building vulnerability offset is constructed.

[0018] The beneficial effects of the above-mentioned further solutions are: the present invention integrates the offset between the actual vulnerability and theoretical vulnerability of buildings, as well as the influencing factors of building vulnerability offset, to construct a sample set of damaged buildings; and uses the random forest algorithm to establish an evaluation model for building vulnerability offset.

[0019] Furthermore, the expression for the building vulnerability offset calculation model is as follows:

[0020] in, Indicates the first Predicted vulnerability offset values ​​for each sample. This represents the number of decision trees in a random forest. Indicates the first Decision tree function, Indicates the first The vulnerability offset influence factor vector for each sample.

[0021] Furthermore, the objective function of the building vulnerability offset calculation model... The expression is as follows:

[0022] in, Indicates the first Building vulnerability offset for each sample This represents the building vulnerability offset predicted by the building vulnerability offset calculation model. This indicates the total number of samples of the buildings.

[0023] Furthermore, S5 includes: In the prediction phase, new building sample data is input, and the building's vulnerability function is used. V Calculate the theoretical vulnerability of a building. ; Using the building vulnerability offset model, output the predicted value of building vulnerability offset. ; Theoretical fragility value And predicted values ​​of building vulnerability offset The results are added together to obtain the final predicted building vulnerability offset, thus completing the building vulnerability prediction. : .

[0024] The beneficial effects of the above-mentioned further solutions are: the present invention integrates the vulnerability function and the vulnerability offset prediction model, corrects the evaluation bias, and achieves accurate prediction of building vulnerability. Attached Figure Description

[0025] Figure 1 This is a flowchart of the method of the present invention.

[0026] Figure 2 This is a schematic diagram of the method framework of the present invention.

[0027] Figure 3 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0028] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0029] Example 1 like Figure 1 and Figure 2 As shown, this invention provides a method for assessing the vulnerability of buildings under debris flow, the implementation of which is as follows: S1. Debris Flow Damaged Building Database Construction: Obtain debris flow characteristic parameters and building damage status of buildings damaged by debris flows, and construct a debris flow damaged building database. The implementation method is as follows: A combination of on-site measurement and 3D laser modeling was used to obtain the mud depth of damaged buildings; the flow velocity of debris flows was calculated, and the impact pressure of debris flows was also calculated; based on the mud depth, flow velocity, and impact pressure of debris flows, characteristic parameters of debris flows were obtained; the damage levels of buildings were classified to obtain the extent of building damage; and based on the characteristic parameters of debris flows and the extent of building damage, a database of buildings damaged by debris flows was constructed.

[0030] In this embodiment, the debris flow characteristic parameters (mud depth, flow velocity, impact pressure) and building damage status (structure type, damage level) of the buildings damaged by debris flow are obtained. The method for achieving this is as follows: The mud depth of the damaged buildings was obtained by combining on-site measurement and 3D modeling measurement. Calculate the velocity of debris flow :

[0031] in, Represents the roughness coefficient. Indicates flow depth. Indicates the slope.

[0032] Considering hydrostatic pressure and hydrodynamic pressure, the impact force of debris flow is calculated. Calculation:

[0033] in, This represents the average density of a debris flow. Indicates flow rate, Indicates the flow depth. Represents gravitational acceleration. Represents hydrostatic pressure. Indicates dynamic water pressure; The building structure type was recorded using data from both on-site surveys and 3D models; Building damage was categorized into four levels, from least to most severe: minor, moderate, severe, and collapse. Damage levels were determined using both on-site survey data and 3D model data: a vulnerability score of 0.25 for minor damage, 0.5 for moderate damage, 0.75 for severe damage, and 1 for collapse.

[0034] S2. Construction of Building Vulnerability Function: Based on the constructed database of debris flow-damaged buildings, using debris flow characteristic parameters (flow velocity, mud depth, impact force, etc.) and building damage level as independent variables, a building vulnerability function is established based on the Logistic function. V The implementation method is as follows: Based on the constructed database of debris flow-damaged buildings, data on damaged buildings within a confidence interval of a preset threshold (95%) were sampled. Using debris flow characteristic parameters and building damage level as independent variables, a vulnerability curve for buildings was established by fitting an S-curve using the Logistic function. V :

[0035] in, a and b Both represent constants. x The constant represents characteristic parameters of debris flows such as flow velocity, mud depth, or impact force. a and b Obtained through statistical regression.

[0036] S3. Construction of the Building Vulnerability Offset Influence Factor System: Multiple influencing factors are selected, factor correlation analysis is performed, strongly correlated factors are eliminated, and factor contribution rates are evaluated to establish the building vulnerability offset influence factor system. The implementation method is as follows: Multiple influencing factors were selected, including the building's physical properties, geographical features, spatial location, and disaster characteristics. The building vulnerability shift influencing factor values ​​for each building sample were extracted and normalized to obtain standardized factor data. The Pearson correlation coefficient was used to statistically analyze the correlation between the factors. Based on the correlation analysis results, strongly correlated factors were removed, and a building vulnerability shift influencing factor system was established.

[0037] In this embodiment, 18 influencing factors are selected from four aspects: the building's physical properties, geographical features, spatial location, and disaster characteristics. Factor correlation analysis is performed, and strongly correlated factors are eliminated to establish a building vulnerability shift influencing factor system. The implementation method is as follows: Eighteen influencing factors were selected from four aspects: the physical properties of the building itself, its geographical features, its spatial location, and its disaster characteristics. The physical properties of the building were characterized by eight factors: structural type, floor area, number of floors, building height, number of windows, window height from the ground, window area, and door area. The geographical features were characterized by three factors: slope, aspect, and vegetation. The spatial location was characterized by six factors: row, row spacing, angle of debris flow impact on the building, distance from the gully, distance from the exit of the mountain, and angle. For the vulnerability of buildings under three different disaster characteristics, one of the core disaster parameters, such as flow velocity, mud depth, or impact pressure, was selected as the influencing factor to ensure the suitability of the factors and the relevance of the analysis.

[0038] The correlation analysis of the unit attribute values ​​obtained from the statistical analysis of 18 selected influencing factors using the Pearson correlation coefficient was conducted. This coefficient is an important indicator for measuring the degree of linear correlation between two continuous variables, and is usually expressed as a symbol. r It is represented as follows. Its core calculation formula is as follows:

[0039] in, Representing two factors X and Y The corresponding number i Data from a sample point, and Representing samples respectively and The mean, Indicates the number of samples. The range of values ​​is The sign indicates the direction of the correlation, and the absolute value indicates the strength of the correlation. When there is a completely linear correspondence between variables, strongly correlated factors are eliminated; after eliminating factors with relatively strong correlations, a system of factors affecting building vulnerability shift is formed.

[0040] S4. Building Vulnerability Offset Modeling: Based on Building Vulnerability Function V The theoretical vulnerability value of the damaged building is calculated, and based on the building vulnerability offset influencing factor system, a random forest algorithm is used to obtain the building vulnerability offset calculation model. The implementation method is as follows: Based on the vulnerability function of buildings V Calculate the theoretical vulnerability value of the damaged building; The difference between the theoretical vulnerability of the damaged building and the actual vulnerability defined on-site is calculated to obtain the building vulnerability offset:

[0041] in, Indicates the first Building vulnerability offset for each sample Indicates the first The actual vulnerability value of each sample Indicates based on the first The theoretical vulnerability value of each sample; Based on building vulnerability offset Construct a modeling sample set based on the building vulnerability offset influencing factors; Using the modeling sample set, the building vulnerability offset is... As the dependent variable, the vulnerability offset factors of each building are used as independent variables to train the random forest regression model; During the sample sampling phase, a bootstrap sampling method is used to randomly sample the sample set containing the building vulnerability offset and the building vulnerability offset influence factor value with replacement, generating a training subset for each decision tree that is consistent with the original damaged building sample dataset and has sample overlap. Decision tree models are trained using each training subset. In the process of constructing a single decision tree, some influencing factors are randomly selected as candidates from the building vulnerability offset influencing factor system. The values ​​of each candidate feature are used as the splitting threshold, with the minimum sum of the mean square errors of the child nodes after splitting being the criterion for judgment.

[0042] in, Indicates the first Building vulnerability offset for each sample An index representing the building vulnerability offset influence factor. Indicating the influence factor of building vulnerability offset The splitting threshold, and These represent the impact factors respectively. and threshold The divided left and right child node regions, and This represents the mean value of the building vulnerability offset within the corresponding child node. Indicates the first The feature vector of building vulnerability offset influence factor corresponding to each sample Indicates the first The first sample One influencing factor characteristic value; By selecting the optimal features and corresponding splitting thresholds as node splitting rules, and combining them with pre-set hyperparameters to form a random forest, a model for calculating building vulnerability offset is constructed.

[0043] In this embodiment, the expression for the building vulnerability offset calculation model is as follows:

[0044] in, Indicates the first Predicted vulnerability offset values ​​for each sample. This represents the number of decision trees in a random forest. Indicates the first Decision tree function, Indicates the first The vulnerability offset influence factor vector for each sample.

[0045] In this embodiment, the training process of the building vulnerability offset calculation model aims to minimize the mean square error (MSE) between the predicted offset and the actual offset:

[0046] in, Describe the objective function. Indicates the first Building vulnerability offset for each sample This represents the building vulnerability offset predicted by the building vulnerability offset calculation model. This represents the total number of building samples. The smaller the objective function L value, the higher the model's prediction accuracy.

[0047] In this embodiment, the Random Forest algorithm is used for offset analysis to achieve accurate prediction of building vulnerability. Random Forest (RF) is a supervised machine learning algorithm based on ensemble learning. Its core principle is to improve the generalization ability and accuracy of the model by constructing multiple decision trees and integrating their prediction results. In vulnerability offset analysis, RF achieves efficient modeling through the following mechanism: In vulnerability offset analysis, the RF algorithm achieves efficient and accurate modeling through a three-dimensional mechanism of "random sample sampling + random feature selection + result ensemble fusion". The specific process and principle are as follows: First, in the sample sampling stage, the Bagging (Bootstrap Aggregating) strategy is adopted to repeatedly draw multiple independent sub-training sets from the original damaged building sample dataset by random sampling with replacement. The sample size of each sub-training set is consistent with that of the original damaged building sample dataset, and sample overlap is allowed between different sub-training sets. Subsequently, a decision tree is trained independently based on each sub-training set. Due to the randomness of each sub-training set, the model's dependence on a single data subset is effectively reduced, significantly decreasing the risk of overfitting and improving model stability. Secondly, during the node splitting process of each decision tree, the RF algorithm introduces feature randomness constraints: instead of using all features to select the optimal split point at each node of each decision tree, a subset of features is randomly selected from all input features as candidate features. The optimal splitting feature and splitting threshold are then selected from these candidate features to complete the node split. This random feature selection mechanism further enhances the diversity of each decision tree, avoids the dominant influence of a single feature on the model, and ensures that different decision trees can focus on different dimensions of data features. This allows for a more comprehensive and accurate capture of the complex nonlinear relationships between multiple driving factors of vulnerability offset, improving the predictive ability of the building vulnerability offset calculation model for offsets. S5. Accurate Prediction of Building Vulnerability: Input new building sample data and use the vulnerability offset model to output a prediction of building vulnerability, thus completing the assessment of building vulnerability. The implementation method is as follows: In the prediction phase, new building sample data is input, and the building's vulnerability function is used. V Calculate the theoretical vulnerability of a building. ; Using the building vulnerability offset model, output the predicted value of building vulnerability offset. ; Theoretical fragility value And predicted values ​​of building vulnerability offset The results are added together to obtain the final predicted building vulnerability offset, thus completing the building vulnerability prediction. : .

[0048] In this embodiment, the vulnerability function of the building is... V Input new data on damaged buildings and calculate their theoretical vulnerability values ​​based on the disaster characteristics such as debris flow velocity, mud depth, or impact force at the location of the building. For a building vulnerability offset calculation model based on the random forest algorithm, new damaged building data is input. This sample is fed into each decision tree in the model, and each decision tree outputs an independent vulnerability offset prediction value. The final vulnerability offset prediction result is obtained by averaging the prediction outputs of all decision trees.

[0049] In summary, this invention improves the accuracy and completeness of basic data by constructing a database of debris flow-damaged buildings. Based on this, it constructs a building vulnerability function and plots vulnerability curves, calculates the offset between actual and theoretical vulnerability, analyzes and obtains the system of factors influencing building vulnerability offset, and constructs a vulnerability offset prediction model through a random forest model. Finally, by integrating the building vulnerability function and the vulnerability offset prediction calculation model, it achieves accurate prediction of building vulnerability.

[0050] Example 2 like Figure 3 As shown, this invention provides a building vulnerability assessment system under debris flow conditions, used to perform the building vulnerability assessment method under debris flow conditions described in Example 1, comprising: The first processing module is used to obtain the debris flow characteristic parameters (mud depth, flow velocity, impact force) and the damage status (structure type, damage level) of the buildings damaged by the debris flow. The second processing module is used to establish the vulnerability function and vulnerability curve of buildings based on the debris flow damaged building database, with debris flow characteristic parameters and building damage level as independent variables. The third processing module is used to select 18 influencing factors from four dimensions: the physical properties of the building itself, its geographical features, its spatial location, and its disaster characteristics. It then conducts factor correlation analysis, eliminates strongly correlated factors, and establishes a building vulnerability shift influencing factor system. The fourth processing module is used to perform offset analysis using a random forest model based on the offset between the actual and theoretical vulnerability of the building and the building vulnerability offset influencing factors, and to construct a vulnerability offset prediction model. The fifth processing module integrates the building vulnerability function and the vulnerability offset calculation model to complete the accurate prediction of building vulnerability.

[0051] In this embodiment, the building vulnerability assessment system under debris flow conditions, in order to realize the principle and beneficial effects of the building vulnerability assessment method under debris flow conditions, includes hardware structures and / or software modules corresponding to the execution of various functions. Those skilled in the art should readily recognize that, in conjunction with the illustrative units and algorithm steps described in the embodiments disclosed in this invention, the present invention can be implemented in hardware and / or a combination of hardware and computer software. Whether a function is executed by hardware or computer software depends on the specific application and design constraints of the technical solution. Different methods can be used to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

Claims

1. A method for assessing the vulnerability of buildings under debris flow, characterized in that, Includes the following steps: S1. Debris Flow Damaged Building Database Construction: Obtain debris flow characteristic parameters and building damage status of buildings damaged by debris flows, and construct a debris flow damaged building database. S2. Construction of Building Vulnerability Function: Based on the constructed database of debris flow-damaged buildings, using debris flow characteristic parameters and building damage level as independent variables, a building vulnerability function is established based on the Logistic function. V ; S3. Construction of the building vulnerability offset influencing factor system: Select multiple influencing factors, conduct factor correlation analysis, eliminate strongly correlated factors, and establish the building vulnerability offset influencing factor system. S4. Building Vulnerability Offset Modeling: Based on Building Vulnerability Function V The theoretical vulnerability value of the damaged building is calculated, and based on the building vulnerability offset influence factor system, a random forest algorithm is used to obtain the building vulnerability offset calculation model. S5. Accurate prediction of building vulnerability: Input new building sample data and use the vulnerability offset model to output the building vulnerability prediction, thus completing the assessment of building vulnerability.

2. The method for assessing building vulnerability under debris flow conditions according to claim 1, characterized in that, S1 includes: The mud depth of the damaged building was obtained by combining on-site measurement and 3D laser modeling. Calculate the flow velocity of the debris flow and the impact pressure of the debris flow; Debris flow characteristic parameters are obtained based on mud depth, debris flow velocity, and debris flow impact pressure. The damage level of a building is classified to obtain information about the extent of the damage. A database of buildings damaged by debris flows was constructed based on debris flow characteristic parameters and building damage conditions.

3. The method for assessing building vulnerability under debris flow conditions according to claim 1, characterized in that, S2 includes: Based on the debris flow-damaged building database, data on damaged buildings within a confidence interval of a preset threshold are sampled. Using debris flow characteristic parameters and building damage level as independent variables, a vulnerability curve for buildings was established by fitting an S-curve using the Logistic function. V : in, a and b Both represent constants. x These represent the characteristic parameters of debris flows.

4. The method for assessing building vulnerability under debris flow conditions according to claim 1, characterized in that, S3 includes: Multiple dimensions of building vulnerability offset influencing factors were selected, including the building's own physical properties, geographical features, spatial location, and disaster characteristics. Extract the building vulnerability offset influence factor values ​​corresponding to each building sample, and normalize them to obtain standardized factor data; The correlation between factors was analyzed by statistically analyzing the normalized factor data using the Pearson correlation coefficient. Based on the correlation analysis results, strongly correlated factors were eliminated, and a system of factors affecting building vulnerability shift was established.

5. The method for assessing building vulnerability under debris flow conditions according to claim 1, characterized in that, S4 includes: Based on the vulnerability function of buildings V Calculate the theoretical vulnerability value of the damaged building; The difference between the theoretical vulnerability of the damaged building and the actual vulnerability defined on-site is calculated to obtain the building vulnerability offset: in, Indicates the first Building vulnerability offset for each sample Indicates the first The actual building vulnerability value for each sample. Indicates based on the first The theoretical building vulnerability value for each sample; Based on building vulnerability offset Construct a modeling sample set based on the building vulnerability offset influencing factors; Using the modeling sample set, the building vulnerability offset is... As the dependent variable, the vulnerability offset factors of each building are used as independent variables to train the random forest regression model; During the sample sampling phase, a bootstrap sampling method is used to randomly sample the sample set containing the building vulnerability offset and the building vulnerability offset influence factor value with replacement, generating a training subset for each decision tree that is consistent with the original damaged building sample dataset and has sample overlap. Decision tree models are trained using each training subset. In the process of constructing a single decision tree, some features are randomly selected from the building vulnerability offset influence factor system as candidate features. The values ​​of each candidate feature are used as the splitting threshold, and the criterion for determining the optimal offset model is the minimum sum of the mean square errors of the child nodes after splitting. in, Indicates the first Building vulnerability offset for each sample An index representing the building vulnerability offset influence factor. Indicating the influence factor of building vulnerability offset The splitting threshold, and These represent the impact factors respectively. and threshold The divided left and right child node regions, and This represents the mean value of the building vulnerability offset within the corresponding child node. Indicates the first The feature vector of building vulnerability offset influence factor corresponding to each sample Indicates the first The first sample One influencing factor characteristic value; By selecting the optimal features and corresponding splitting thresholds as node splitting rules, and combining them with pre-set hyperparameters to form a random forest, a model for calculating building vulnerability offset is constructed.

6. The method for assessing building vulnerability under debris flow according to claim 5, characterized in that, The expression for the building vulnerability offset calculation model is as follows: in, Indicates the first Predicted vulnerability offset values ​​for each sample. This represents the number of decision trees in a random forest. Indicates the first Decision tree function, Indicates the first The vulnerability offset influence factor vector for each sample.

7. The method for assessing building vulnerability under debris flow according to claim 5, characterized in that, The objective function of the building vulnerability offset calculation model The expression is as follows: in, Indicates the first The vulnerability offset of each sample, This represents the building vulnerability offset predicted by the building vulnerability offset calculation model. This indicates the total number of samples of the buildings.

8. The method for assessing building vulnerability under debris flow according to claim 1, characterized in that, S5 includes: In the prediction phase, new building sample data is input, and the building's vulnerability function is used. V Calculate the theoretical vulnerability of a building. ; Using the building vulnerability offset model, output the predicted value of building vulnerability offset. ; Theoretical fragility value And predicted values ​​of building vulnerability offset The results are added together to obtain the final predicted building vulnerability offset, thus completing the building vulnerability prediction. : 。

Images