Bridge Slope Geological Disaster Early Warning System and Method Based on Multi-Source Data Fusion

Abstract

This invention discloses a bridge slope geological disaster early warning system and method based on multi-source data fusion, belonging to the field of bridge slope geological disaster early warning technology. The system includes: an acquisition module for acquiring monitoring data of the bridge slope to be monitored; a processing module for using abnormal data in the monitoring data as preliminary abnormal data; a prediction module for predicting a first probability of landslide on the bridge slope to be monitored based on the preliminary abnormal data, and a second probability of landslide based on the monitoring data, and determining the landslide risk level based on the first and second probabilities; and an early warning module for conducting landslide-debris flow-bridge damage disaster chain early warning analysis and issuing an early warning if the landslide risk level is greater than or equal to the first risk level. This system can predict whether a landslide disaster will occur on a bridge slope relatively accurately, enabling timely and accurate early warning of the landslide-debris flow-bridge damage disaster chain.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of bridge slope geological disaster early warning, and particularly relates to a bridge slope geological disaster early warning system and method based on multi-source data fusion. BACKGROUND

[0002] With the continuous advancement of infrastructure construction, highway bridge construction has been continuously extended from the original plain and hilly areas to the current hilly and mountainous areas. The terrain in hilly areas is rugged, and the slope distribution is relatively wide. The terrain and geological conditions are the main influencing factors of highway design in hilly and mountainous areas. Considering the drainage conditions and highway grade factors, bridge structures are often selected to cross complex terrain located in sliding zones, and therefore, large excavation slopes are present on both sides of the bridge. At this time, the stability of the bridge slope is closely related to the safety of the bridge. When a landslide occurs on the bridge slope, it is easy to trigger a debris flow, which in turn causes serious damage to the bridge, i.e., a landslide-debris flow-bridge damage disaster chain, thereby causing large disaster losses, such as bridge property and safety losses. Therefore, predicting in advance whether a landslide disaster will occur on the bridge slope is conducive to timely and accurate early warning of the landslide-debris flow-bridge damage disaster chain. However, the existing technology usually draws a historical landslide curve corresponding to each control factor in a historical case of an affected landslide, draws a current landslide curve corresponding to the control factor of the to-be-measured slope, and compares the current landslide curve with the historical landslide curve to predict the probability of a landslide occurring on the to-be-measured slope. However, this way of predicting whether a landslide disaster will occur on the bridge slope has low accuracy, which makes it difficult to timely and accurately perform early warning of the landslide-debris flow-bridge damage disaster chain.

[0003] It should be noted that the information disclosed in the above BACKGROUND section is only used to strengthen the understanding of the background of the present application, and therefore can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0004] In order to have a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. The summary is not a general review, nor is it intended to determine key / important components or delineate the scope of protection of these embodiments, but as a prelude to the detailed description below.

[0005] The bridge slope geological disaster early warning system and method based on multi-source data fusion provided by the embodiments of the present disclosure can more accurately predict whether a landslide disaster will occur on the bridge slope, so as to timely and accurately perform early warning of the landslide-debris flow-bridge damage disaster chain.

[0006] In some embodiments, the bridge slope geological disaster early warning system based on multi-source data fusion includes: an acquisition module for acquiring monitoring data of a bridge slope to be measured; wherein the monitoring data includes geological and topographical data, meteorological and hydrological data, engineering activity data, and displacement and deformation data; a processing module for determining whether there is abnormal data based on the monitoring data, and if so, treating the abnormal data as preliminary abnormal data; a prediction module for predicting a first probability of landslide on the bridge slope to be measured based on the preliminary abnormal data, and predicting a second probability of landslide on the bridge slope to be measured based on the monitoring data, and determining the landslide risk level of the bridge slope to be measured based on the first probability and the second probability; and an early warning module for performing a disaster chain early warning analysis of landslide-debris flow-bridge damage using a preset early warning model if the landslide risk level is greater than or equal to the first risk level, obtaining early warning analysis results, and issuing an early warning based on the early warning analysis results.

[0007] In some embodiments, the bridge slope geological disaster early warning method based on multi-source data fusion includes: acquiring monitoring data of the bridge slope to be measured; wherein the monitoring data includes geological and topographical data, meteorological and hydrological data, engineering activity data, and displacement and deformation data; determining whether there is abnormal data based on the monitoring data, and if so, using the abnormal data as preliminary abnormal data; predicting a first probability of landslide on the bridge slope to be measured based on the preliminary abnormal data, and predicting a second probability of landslide on the bridge slope to be measured based on the monitoring data, and determining the landslide risk level of the bridge slope to be measured based on the first probability and the second probability; if the landslide risk level is greater than or equal to the first risk level, performing a disaster chain early warning analysis of landslide-debris flow-bridge damage using a preset early warning model, obtaining early warning analysis results, and issuing an early warning based on the early warning analysis results.

[0008] The beneficial effects of this invention are as follows:

[0009] The acquisition module obtains monitoring data of the bridge slope under test. The processing module uses abnormal data from the monitoring data as preliminary abnormal data, reflecting local anomalies in the bridge slope, while the monitoring data reflects the overall state of the bridge slope. Thus, the prediction module uses the preliminary data to predict the first probability of landslides on the bridge slope under test, and uses the monitoring data to predict the second probability of landslides. Based on the first and second probabilities, the landslide risk level of the bridge slope under test is determined. This allows for a more accurate prediction of whether a landslide will occur based on both local anomalies and the overall state of the bridge slope. When the landslide risk level is greater than or equal to the first risk level, it indicates a higher probability of landslides. In this case, an early warning model is used to analyze the landslide-debris flow-bridge damage disaster chain, enabling timely and accurate early warning of the landslide-debris flow-bridge damage disaster chain.

[0010] The above general description and the description below are exemplary and illustrative only and are not intended to limit this application. Attached Figure Description

[0011] One or more embodiments are illustrated by way of example with reference to the accompanying drawings. These illustrations and drawings do not constitute a limitation on the embodiments. Elements having the same reference numerals in the drawings are shown as similar elements. The drawings are not to be scaled. And wherein:

[0012] Figure 1 This is a schematic diagram of a bridge slope geological disaster early warning system based on multi-source data fusion provided by the present invention;

[0013] Figure 2 This is a flowchart of a bridge slope geological disaster early warning method based on multi-source data fusion provided by the present invention. Detailed Implementation

[0014] To provide a more detailed understanding of the features and technical content of the embodiments of this disclosure, the implementation of the embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. The accompanying drawings are for illustrative purposes only and are not intended to limit the embodiments of this disclosure. In the following technical description, for ease of explanation, several details are used to provide a full understanding of the disclosed embodiments. However, one or more embodiments may still be implemented without these details. In other cases, well-known structures and devices may be simplified in their depiction to simplify the drawings.

[0015] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this disclosure described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion.

[0016] Unless otherwise stated, the term "multiple" means two or more.

[0017] In this embodiment of the disclosure, the character " / " indicates that the objects before and after it are in an "or" relationship. For example, A / B means: A or B.

[0018] The term "and / or" describes an association between objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or A and B.

[0019] The term "correspondence" can refer to an association or binding relationship. The correspondence between A and B means that there is an association or binding relationship between A and B.

[0020] Combination Figure 1 As shown in the embodiments of this disclosure, a bridge slope geological disaster early warning system based on multi-source data fusion is provided, including an acquisition module, a processing module, a prediction module, and an early warning module. The acquisition module is used to acquire monitoring data of the bridge slope to be measured; the monitoring data includes geological and topographical data, meteorological and hydrological data, engineering activity data, and displacement and deformation data. The processing module is used to determine whether there is abnormal data based on the monitoring data, and if so, to treat the abnormal data as preliminary abnormal data. The prediction module is used to predict a first probability of landslide on the bridge slope to be measured based on the preliminary abnormal data, and a second probability of landslide on the bridge slope to be measured based on the monitoring data, and to determine the landslide risk level of the bridge slope to be measured based on the first and second probabilities. The early warning module is used to perform a disaster chain early warning analysis of landslide-debris flow-bridge damage using a preset early warning model if the landslide risk level is greater than or equal to the first risk level, obtain the early warning analysis results, and issue an early warning based on the early warning analysis results.

[0021] The bridge slope geological disaster early warning system based on multi-source data fusion provided in this embodiment acquires monitoring data of the bridge slope to be tested through an acquisition module. A processing module uses abnormal data from the monitoring data as preliminary abnormal data, reflecting local anomalies in the bridge slope, while the monitoring data reflects the overall state of the bridge slope. The prediction module uses the preliminary data to predict the first probability of a landslide on the bridge slope and the monitoring data to predict the second probability. Based on the first and second probabilities, the landslide risk level of the bridge slope is determined. This allows for a more accurate prediction of whether a landslide will occur based on both local anomalies and the overall state of the bridge slope. When the landslide risk level is greater than or equal to the first risk level, it indicates a higher probability of a landslide. In this case, the early warning module performs an early warning analysis of the landslide-debris flow-bridge damage disaster chain, providing timely and accurate early warnings.

[0022] In some embodiments, geological and topographic data include soil and rock properties, geological structure, slope, slope length, vegetation cover, and soil moisture saturation. For example, the soil and rock properties and geological structure of the bridge slope under test can be obtained through geological surveys; the slope and slope length of the bridge slope under test can be obtained using topographic maps, digital elevation models (DEMs), or lidar; and the vegetation cover of the bridge slope under test can be monitored through remote sensing or field surveys. Meteorological and hydrological data include rainfall information (such as rainfall amount and intensity), groundwater information, and freeze-thaw cycle information. Engineering activities include construction activities within a first predetermined range centered on the bridge slope, which characterizes the range within which the construction activities have a significant impact on the stability of the bridge slope. It is understandable that the intensity of engineering construction activities affects the stability of bridge slopes to varying degrees. Therefore, two sub-regions can be set up, such as a first sub-region centered on the bridge slope and a second sub-region centered on the bridge slope (the second sub-region includes and is larger than the first sub-region). The intensity of engineering construction activities (e.g., high and low intensities) is then determined. All engineering construction activities within the first sub-region are considered to affect the stability of the bridge slope, and high-intensity engineering construction activities within the second sub-region are considered to affect the stability of the bridge slope. Displacement and deformation data include surface displacement (e.g., horizontal and vertical displacement), deep displacement (e.g., sliding surface displacement and stratified settlement), crack conditions (width, length, and orientation changes of cracks on the slope surface), and overall deformation (three-dimensional deformation field, such as composite deformations like tilting and torsion). Surface displacement can be obtained through a combination of GNSS and total station; deep displacement can be obtained through a combination of inclinometer and stratified settlement meter; crack conditions can be obtained through a combination of electronic crack gauge and photogrammetry; and overall deformation can be obtained through a combination of 3D laser scanning and InSAR (satellite remote sensing). That is, in this embodiment of the disclosure, the monitoring data is multi-source data, including geological and topographical data, meteorological and hydrological data, engineering activity data, and displacement and deformation data. For example, geological survey data, topographic map data, remote sensing data, meteorological and hydrological monitoring data, construction project information survey data, GNSS data, total station monitoring data, inclinometer monitoring data, stratified settlement meter monitoring data, three-dimensional laser scanning data, and InSAR (satellite remote sensing) data, etc.

[0023] In some embodiments, early anomaly data characterizes precursor monitoring data prior to a landslide on the slope of the bridge under test. The soil and rock properties and geological structures in geological and topographical data are relatively stable and generally do not change. However, engineering activities are dominated by human activity, making it difficult to quantify whether they are abnormal. Therefore, when determining whether abnormal data exists in the monitoring data, soil and rock properties and geological structures in the geological and topographical data, as well as engineering activity data, can be removed. For the remaining monitoring data, changes should have their own characteristics; changes that differ from these characteristics indicate a possible anomaly. For example, the processing module is specifically used to: use data other than soil and rock properties, geological structures, and engineering activity data as target monitoring data. The target monitoring data is divided into multiple continuous time windows. Dynamic changes in the bridge slope under test are calculated for each time window. These dynamic changes include: slope change rate, slope length change rate, vegetation cover rate, soil moisture saturation, rainfall change rate, groundwater level change, freezing depth and moisture content during freeze-thaw cycles, surface displacement change rate, deep displacement change rate, crack condition change rate, and overall deformation change rate. Threshold conditions are set for each type of dynamic change data. For each type of dynamic change data, if it meets the corresponding threshold condition within any time window, the target monitoring data corresponding to that type of dynamic change data is considered to be pre-abnormal data. The time window in which the threshold condition is met is defined as the duration of the pre-abnormal data. The total duration of the monitoring data is taken as the total monitoring duration, and the sum of the durations of the pre-abnormal data is taken as the cumulative abnormal duration.

[0024] Preferably, the prediction module includes a first submodule and a second submodule. The first submodule is used to calculate the subprobability of a landslide occurring on the slope of the bridge under test based on each type of early anomaly data. The second submodule is used to determine a first probability based on a comprehensive analysis of the various subprobabilities.

[0025] In this way, by calculating the sub-probabilities corresponding to each type of early abnormal data separately, and then comprehensively determining the first probability that the bridge slope under test will experience a landslide, the impact of various early abnormal data on the bridge slope under test can be determined more accurately.

[0026] Preferably, the first submodule includes: a first calculation unit and a second calculation unit. The first calculation unit is used to calculate the instability probability of the bridge slope under test undergoing natural instability, and to calculate the correlation between each type of prior anomaly data and the bridge slope under test, as well as the probability of each type of prior anomaly data occurring. The second calculation unit is used, for each type of prior anomaly data, based on the instability probability, correlation, and probability of the prior anomaly data occurring, to calculate the sub-probability of each prior anomaly data leading to a landslide on the bridge slope under test using Bayes' theorem, as follows:

[0027]

[0028] Wherein, P(A / B) i Let P(B) be the sub-probability that the i-th type of prior abnormal data leads to a landslide on the slope of the bridge under test. i / A) represents the correlation between the i-th type of early abnormal data and the slope of the bridge to be tested, P(A) is the instability probability, and P(B) is the instability probability. i ) represents the probability of the occurrence of the i-th type of early abnormal data. Among them, natural instability represents the occurrence of landslides under natural conditions.

[0029] In some embodiments, natural conditions can be considered as situations where there is no obvious cause for slope landslides. Obvious causes include extreme rainfall events and / or earthquake events. That is, in this embodiment of the disclosure, natural instability refers to landslides occurring in the absence of extreme rainfall events and earthquake events. Extreme rainfall events are characterized by rainfall amount ≥ extreme rainfall threshold and / or rainfall intensity ≥ extreme intensity threshold. For example, if slope A experiences a landslide under the influence of an extreme rainfall event or an earthquake event, then slope A is considered non-naturally unstable. If slope B experiences a landslide without the influence of extreme rainfall events and earthquake events, then slope B is considered naturally unstable. Since micro-earthquakes (such as a magnitude 1 earthquake) have a relatively small impact on bridge slopes and are unlikely to directly trigger landslides, earthquake magnitude conditions can also be set as obvious causes based on actual conditions. For example, earthquake events with a magnitude higher than a preset earthquake magnitude can be used as obvious causes.

[0030] In this way, by calculating the instability probability, the correlation degree corresponding to each type of early abnormal data, and the probability of occurrence of each type of early abnormal data, and then using Bayes' theorem for calculation, it is possible to obtain the sub-probability of each type of early abnormal data leading to a landslide on the bridge slope with relatively accurate results.

[0031] Preferably, the first calculation unit is specifically used for: obtaining the total number of slopes and the number of slopes that have experienced natural instability within the preset monitoring area; and, based on the total number of slopes and the number of slopes that have experienced natural instability within the preset monitoring area, calculating the probability of natural instability of the slopes within the preset monitoring area using the classical probability formula, as shown below: Wherein, P(C) is the probability of natural instability of the slope in the preset monitoring area, m is the total number of slopes in the preset monitoring area, and n is the number of slopes in the preset monitoring area that have experienced natural instability; the probability of natural instability of the slope in the preset monitoring area is taken as the instability probability of natural instability of the bridge slope to be tested.

[0032] In this way, by setting up a monitoring area and obtaining the total number of slopes and the number of slopes that have experienced natural instability within that monitoring area, and by using the classical probability formula to calculate the probability of natural instability of the slopes within the preset monitoring area, and using this probability as the instability probability of the bridge slope to be tested, the instability probability of the bridge slope to be tested can be obtained more accurately.

[0033] Understandably, the monitoring area is an area with a large number of slopes in order to obtain sufficient sample data (i.e., the total number of slopes and the number of slopes that have experienced natural instability), thereby increasing the probability of calculating the natural instability of slopes in this area.

[0034] In this embodiment, due to the different climatic conditions in various regions, the extreme rainfall threshold and extreme intensity threshold will also differ. For a preset monitoring area, multiple rainfall events (including rainfall amount and intensity) occurring in that area can be randomly selected and divided according to season or rainfall pattern characteristics to obtain different categories (e.g., seasons: spring, summer, autumn, and winter; rainfall pattern characteristics: periods with more rainfall, periods with less rainfall). Then, the extreme rainfall threshold and extreme intensity threshold corresponding to each category are determined separately. The extreme rainfall threshold and extreme intensity threshold corresponding to each category can be obtained in the following ways:

[0035] For each category of extreme rainfall threshold, multiple rainfall events are selected, and the variance of the total rainfall is determined using a T-distribution: in, X represents the average of the selected rainfall amounts. i Let be the i-th rainfall amount, n be the number of selected rainfall events, S be the sample standard deviation, t be the variance of the total rainfall, and μ be the mean of the total rainfall. Then, a confidence interval is constructed using the confidence level and the variance of the total rainfall. The upper limit of this confidence interval is used as the initial rainfall threshold. Where α is the significance level, then (1-α) is the pre-set reliability. Represents the area under the right tail in a t-distribution with n-1 degrees of freedom. The corresponding t value, Indicates the confidence width, which describes the distribution range of the acceptance region. This represents the lower limit of the confidence interval. The upper limit of the confidence interval is set at the initial rainfall threshold. The product of the initial rainfall threshold and the preset parameter is then used as the extreme rainfall threshold. This first preset parameter is greater than 1 (the extreme rainfall threshold should be significantly greater than the initial rainfall threshold), and its specific value can be determined based on actual conditions or expert experience. Similarly, the extreme intensity threshold is determined, which will not be elaborated upon here.

[0036] In some embodiments, when calculating the probability of natural instability of a slope within a preset monitoring area, slopes that have experienced landslides within the monitoring area are first designated as target slopes. The presence of extreme rainfall or earthquake events within a preset historical period (e.g., the previous 10 days) before a landslide on the target slope is determined. If such events exist, the target slope is classified as a non-naturally unstable slope; otherwise, it is classified as a naturally unstable slope. For example, for each target slope, rainfall and earthquake events within a preset historical period before a landslide are obtained. The category of the rainfall event is determined (e.g., spring), and then the extreme rainfall threshold and extreme intensity threshold of that category are used to determine whether the rainfall event is an extreme rainfall event. If it is an extreme rainfall event, the slope that triggered the landslide is classified as a non-naturally unstable slope, and the difference between the number of landslides and the number of non-naturally unstable slopes within the monitoring area is taken as the number of naturally unstable slopes (i.e., the number of slopes that have experienced natural instability). Similarly, based on earthquake events within a predetermined historical timeframe prior to a landslide, the target slope is determined to be either a non-naturally unstable slope or a naturally unstable slope. A non-naturally unstable slope is characterized by landslides triggered by extreme rainfall events and / or earthquake events. A naturally unstable slope is characterized by landslides occurring in the absence of extreme rainfall events and earthquake events. For example, if area A is a monitoring region with 100 slopes, and 8 slopes experienced landslides, of which 2 were non-naturally unstable slopes, then there are 6 naturally unstable slopes. The probability of a slope in monitoring area A experiencing natural instability is (8-2) / 100.

[0037] Preferably, a preset monitoring area is determined based on the characteristics of the bridge to be tested. These characteristics include at least one of administrative characteristics, geographical features, mountain range characteristics, and soil and rock mass characteristics. In some embodiments, the preset monitoring area is determined based on the administrative characteristics of the bridge to be tested: the area pointed to by the superior administrative unit to which the bridge slope belongs is used as the preset monitoring area. For example, if the bridge slope is located in town A, the preset area could be the area pointed to by the superior administrative unit to which town A belongs, i.e., province A, city A, or county A, etc. In other embodiments, the preset monitoring area is determined based on the geographical features of the bridge to be tested: the area within a second preset range centered on the bridge slope is used as the preset monitoring area, such as a range of several hundred kilometers. In other embodiments, the preset monitoring area is determined based on the mountain range characteristics of the bridge to be tested: the mountain range to which the bridge slope belongs is used as the preset monitoring area. For example, if the bridge slope is located at a certain location in mountain range A, then mountain range A is used as the preset monitoring area. In other embodiments, the preset monitoring area is determined based on the soil and rock mass characteristics of the bridge to be tested: areas with the same or similar soil and rock mass as the bridge slope are used as the preset monitoring area. Rock and soil mass includes rock strength, soil properties, and rock and soil structure.

[0038] Preferably, the correlation between the prior anomalous data and the bridge slope under test characterizes the degree of influence of the prior anomalous data on the occurrence of landslides on the bridge slope; that is, the greater the correlation, the more likely the prior anomalous data is to cause a landslide on the bridge slope. In some embodiments, the degree of influence of various prior anomalous data on the occurrence of landslides on the bridge slope can be determined by expert experience, and this degree of influence is taken as the corresponding correlation.

[0039] Preferably, the first calculation unit is used to calculate the probability of occurrence of each type of early abnormal data in the following manner: obtaining the total monitoring time and the cumulative abnormal time when the abnormality occurs for each type of early abnormal data; for each type of early abnormal data, based on the total monitoring time and the cumulative abnormal time, calculating the probability of occurrence of the early abnormal data using the following formula: Among them, P(B) i Let t be the probability of the i-th type of early outlier, and t be the probability of the i-th type of outlier. i T represents the cumulative duration of anomalies when the i-th type of early-stage anomalous data becomes abnormal. i denoted as the total monitoring duration for the i-th type of early abnormal data.

[0040] In this way, for each type of early abnormal data, the probability of its occurrence can be determined more reasonably and accurately by calculating the proportion of the cumulative abnormal duration to its total monitoring duration.

[0041] Preferably, the second submodule is specifically used to determine the first probability based on a combination of the various subprobabilities in the following manner: P z =1-(1-P1)·(1-P2)·(1-P3)...·(1-P i ), where P z For the first probability, P i Let be the sub-probability that the i-th type of early abnormal data leads to a landslide on the slope of the bridge under test.

[0042] In some embodiments, the prediction module can also utilize a combination of knowledge graphs and graph convolutional neural networks to predict a second probability of landslides occurring on the slope of the bridge under test based on monitoring data. For example, a knowledge graph ontology model is constructed based on the node information of the bridge slope under test. Then, a landslide disaster knowledge graph is constructed based on multiple historical landslide events and the ontology model. These historical landslide events include corresponding monitoring data, and the knowledge graph includes geological and topographical nodes, meteorological and hydrological nodes, engineering activity nodes, and displacement and deformation nodes. A graph convolutional neural network is then constructed. For each historical landslide event, the feature vector corresponding to the monitoring data is input into the graph convolutional neural network for training to obtain a corresponding landslide prediction model. Based on the feature vector corresponding to the monitoring data of the bridge slope under test, the landslide prediction model is used for prediction to obtain the second predicted probability. This feature vector can be static information (such as rainfall) or dynamic change information (such as displacement change rate) from the monitoring data. In this embodiment, if the monitoring time corresponding to the monitoring data is long, the latest monitoring data can be selected as the input data for the landslide prediction model. This latest time can be the latest data from the last day.

[0043] In other embodiments, the prediction module can also utilize a linear regression network to predict a second probability of landslides occurring on the slope of the bridge under test based on monitoring data. For example, a prediction function can be constructed and trained based on multiple historical landslide events to obtain the target prediction function: F(x)=α0+α1x1+α2x2+α3x3+...+α n x n Where x = {x1, x2, x3, ... x} n Let} be the feature vector corresponding to the monitoring data, α0 be the intercept term, and α1, α2, α3, ... α n The regression coefficients represent the degree of influence of each feature vector on the prediction result. The prediction function obtains the second probability of landslides occurring on the bridge slope under test by summing the products of each feature vector and its corresponding regression coefficient, and adding the intercept term. The function training process is a common technique and will not be elaborated here.

[0044] Preferably, the prediction module includes a third sub-module; the third sub-module is specifically used for: determining the first level corresponding to the first probability and the second level corresponding to the second probability respectively; if either the first level or the second level is greater than or equal to the second risk level, then the highest level among the first level and the second level is determined as the landslide risk level of the bridge slope to be tested; if both the first level and the second level are less than the second risk level, then the landslide risk level of the bridge slope to be tested is determined based on the first level, the first preset weight corresponding to the first level, and the second preset weight corresponding to the second level; wherein, the second risk level is greater than the first risk level.

[0045] In this way, by mapping the first probability and the second probability to corresponding levels, if the first level is greater than or equal to the second risk level, it indicates that based on the prior abnormal data, the slope of the bridge under test is very likely to experience a landslide. If the second level is greater than or equal to the second risk level, it indicates that based on the monitoring data, the slope of the bridge under test is very likely to experience a landslide. In this case, the highest level between the first and second levels is determined as the landslide risk level of the bridge under test, so as to accurately predict whether the bridge under test will experience a landslide. If both the first and second levels are less than the second risk level, it indicates that based solely on the prior abnormal data or the bridge under test, it is not possible to directly determine whether the bridge under test will experience a landslide. In this case, the landslide risk level of the bridge under test is determined by combining the first preset weight, the second preset weight, the first level, and the second level, so as to accurately determine whether the bridge under test will experience a landslide.

[0046] Preferably, the third submodule is specifically used to determine the first level corresponding to the first probability and the second level corresponding to the second probability in the following ways: performing a lookup operation in a preset data table using the first probability and the second probability respectively, finding the level corresponding to the first probability as the first level, and finding the level corresponding to the second probability as the second level. The preset data table stores the correspondence between the first probability, the second probability, and the levels.

[0047] Preferably, the third submodule is specifically used to determine the landslide risk level of the bridge slope under test based on the first level, the first preset weight corresponding to the first level, and the second preset weight corresponding to the second level if both the first level and the second level are less than the second risk level. Specifically, it performs table lookup operations based on the first and second levels respectively to find the mapping value corresponding to the first level as the first target value, and to find the mapping value corresponding to the second level as the second target value. The landslide risk value is calculated using the following formula: Landslide Risk Value = First Target Value * First Preset Weight + Second Target Value * Second Preset Weight. If the landslide risk value is greater than or equal to the preset risk value, the landslide risk level of the bridge slope under test is determined to be the first risk level.

[0048] In some embodiments, the first preset weight is 0.6, and the second preset weight is 0.4. Table 1 is an example table of preset data tables. The first risk level is level three, the second risk level is level four, and the preset risk value is 2a. Assuming P1 is 28% and P2 is 20%, then the first level is level four, and the landslide risk level is level four. Assuming P1 is 15% and P2 is 20%, then the first level is level two, with a corresponding first target value of 2a, and the second level is level three, with a corresponding second target value of 3a. The landslide risk value is then 2.4a (2a*0.6+3a*0.4). This landslide risk value is greater than the preset risk value, so the landslide risk level of the bridge slope to be tested is determined to be the first risk level, i.e., level three.

[0049] Table 1

[0050] first probability P1 first probability P2 Rank Mapping values [P1 < 15%] [P2 < 15%] First rank a 15% < P1 < 20% 15% < P2< 20% Second rank 2a 20% < P1 < 25% 20% < P2< 25% Third rank 3a 25% < P1 25% < P2 Fourth rank No

[0051] Preferably, a bridge slope geological disaster early warning system based on multi-source data fusion further includes a model building module. The model building module includes: a node submodule, a determination submodule, a case submodule, a calculation submodule, and a model submodule. The node submodule is used to acquire nodes related to the "landslide-debris flow-bridge damage" disaster chain. These nodes include three types: disaster-causing factors, disaster-inducing environments, and disaster events. Disaster-causing factor nodes include landslide characteristics and sliding paths; disaster-inducing environment nodes include rainfall intensity, valley gradient, debris flow fluid characteristics, the relative position of the bridge and the gully, and the bridge structure; and disaster event nodes include debris flow severity and bridge damage severity. The determination submodule is used to determine the influence relationships between nodes based on the disaster mechanism and chain transmission law of the "landslide-debris flow-bridge damage" disaster chain, and to determine the directed edge information between nodes based on these influence relationships. The case submodule is used to acquire multiple corresponding historical cases based on the "landslide-debris flow-bridge damage" disaster chain; historical cases include node variables corresponding to each node. The computation submodule is used to calculate the conditional probabilities between nodes using the node variables from multiple historical cases. The model submodule is used to construct a Bayesian network model based on the information of each node, the directed edges between nodes, and the conditional probabilities between nodes, and then uses this constructed Bayesian network model as the early warning model. Node variables represent the numerical information corresponding to nodes in historical cases.

[0052] In this way, by acquiring the nodes related to the "landslide-debris flow-bridge damage" disaster chain and determining the influence relationship between each node, and then using the node variables of each node in each historical case to calculate the conditional probability between each node, it is convenient to construct a Bayesian network model corresponding to the "landslide-debris flow-bridge damage" disaster chain, so as to provide more accurate early warning for disasters caused by bridge slopes.

[0053] In some embodiments, landslide characteristics characterize landslide intensity, including landslide volume, soil and rock type (such as cohesive soil, sand, gravel, and basement rock) and slope. Sliding paths include gully-constrained type (the landslide moves along an existing gully or river channel, constrained by the terrain on both sides, resulting in a relatively concentrated path), fan-shaped diffusion type (the landslide spreads in a fan shape in a flat area after rushing out of the gully, with a wide path), linear collapse type (the landslide slides down a single steep slope in a straight line without obvious gully guidance), and jump-collision type (the landslide bounces and collides on steep cliffs or stepped terrain, with a discontinuous path). Debris flow fluid characteristics characterize debris flow intensity, including volume, velocity, and solid particle composition.

[0054] Preferably, the submodule is specifically used for: if one node may affect the occurrence state of another node, then for these two nodes, the node that affects the occurrence state of the other node is designated as the parent node, and the node whose occurrence state is affected is designated as the child node. For example, given nodes A and B, if node A affects the occurrence state of node B, then node A is designated as the parent node of node B, and node B is designated as the child node of node A. The directed edge information (parent node pointing to child node) between nodes is determined based on the influence relationship (parent node and child node).

[0055] For example, landslide characteristics (volume, soil type) → sliding path; sliding path → debris flow fluid characteristics (capacity, velocity, solid particle composition); rainfall intensity → debris flow fluid characteristics; valley longitudinal gradient → debris flow velocity; debris flow fluid characteristics → debris flow grade; relative position of bridge and gully → bridge damage grade; debris flow grade + bridge structure → bridge damage grade; debris flow grade → bridge damage grade, etc. The arrow direction indicates that the parent node points to the child node.

[0056] Preferably, the calculation submodule is specifically used to: discretize the node variables in each historical case, and based on the discretized node variables, calculate the conditional probability between each node using the following formula to determine the conditional probability table:

[0057]

[0058]

[0059] Where l is the number of nodes, P(X1 / X2,X3,...X l Let X be the state of all parent nodes of a given node V1, where X1 is in the range X2, X3, ..., X4. l And the probability that the child node V1 is in state X1; P(X1,X2,...X l Let the states of all nodes be X1, X2, ..., X. l The joint probability at time, P(X)m =S a Let P(X) be the prior probability. i =S b / X m =S a ) represents the posterior probability, also known as the conditional probability, indicating that the state of all parent nodes is S. a The child node state is S b The probability of occurrence.

[0060] It is understandable that discretizing data is a common data preprocessing method that can effectively simplify the amount of computation, so it will not be elaborated on here.

[0061] Preferably, the model submodule is specifically used to construct a Bayesian network model BN = (N, E, P). Here, N is the set of all nodes in the network, E is the set of directed edges connecting the nodes in the network, i.e., the influence relationships (parent or child nodes) between nodes, and P is the set of conditional probability tables between nodes.

[0062] Preferably, the early warning module includes an acquisition submodule and an early warning submodule. The acquisition submodule is used to estimate the landslide characteristics, sliding path, and debris flow fluid characteristics of the bridge slope to be tested if the landslide risk level is greater than or equal to the first risk level, and to acquire the rainfall intensity, valley gradient, relative position of the bridge and the gully, and bridge structure corresponding to the bridge slope to be tested. The early warning submodule is used to input the landslide characteristics, sliding path, debris flow fluid characteristics, rainfall intensity, valley gradient, relative position of the bridge and the gully, and bridge structure corresponding to the bridge slope to be tested into the early warning model for analysis, obtain the corresponding early warning analysis results, and issue an early warning based on the early warning analysis results; the early warning analysis results include the catastrophic risk of debris flow and bridge damage.

[0063] Thus, if the landslide risk level is greater than or equal to the first risk level, it indicates that the probability of a landslide on the bridge slope is relatively high. In this case, an early warning model can be used to conduct an early warning analysis based on the characteristics of the landslide body, sliding path, debris flow characteristics, rainfall intensity, valley longitudinal gradient, relative position of the bridge and the gully, and bridge structure corresponding to the bridge slope under test. This allows for disaster early warning before the landslide disaster occurs, and timely and accurate early warning of the landslide-debris flow-bridge damage disaster chain, thereby reducing disaster losses.

[0064] In some embodiments, the prediction of landslide characteristics, sliding path, and debris flow fluid characteristics is a common technique and will not be elaborated upon here. The disaster risk situation specifically includes debris flow level information (each debris flow level and its corresponding probability of occurrence) and bridge damage level information (each bridge damage level and its corresponding probability of occurrence). The debris flow level with the highest probability of occurrence is selected as the target debris flow level, and the bridge damage level with the highest probability of occurrence is selected as the target bridge damage level.

[0065] Preferably, the early warning submodule is specifically used for: determining a fuzzy early warning strategy based on the landslide risk level of the bridge slope to be tested and the early warning analysis results; and issuing an early warning based on the fuzzy early warning strategy.

[0066] In this way, the analysis and early warning of the "landslide-debris flow-bridge damage" disaster chain is only carried out when the probability of landslide on the slope of the bridge under test is relatively high. Therefore, the early warning effect is better when the probability of landslide on the slope of the bridge under test (i.e., landslide risk level) is used together with the early warning analysis results.

[0067] In some embodiments, the early warning submodule determines the fuzzy early warning strategy in the following ways: The landslide risk level is divided into a first target level and a second target level, and mapped to the membership degree of a fuzzy set; the target debris flow level is divided into three fuzzy sets: low, moderate, and severe, and fuzzified using a trapezoidal membership function; the target bridge damage level is divided into three fuzzy sets: slight, moderate, and severe, and fuzzified using a trapezoidal membership function. A fuzzy rule base is constructed as follows: 1) When the landslide risk is at the second target level, and (the target debris flow disaster level and / or the target bridge damage level is severe), a prohibition order is triggered; 2) When the target debris flow level is severe and / or the target bridge damage level is severe, a prohibition order is triggered regardless of the landslide risk level; 3) When the target debris flow level is moderate and / or the target bridge damage level is moderate, a restricted order is triggered regardless of the landslide risk level; 4) When the landslide risk level is the first target level, and the target debris flow level is low and the target bridge damage level is slight, a permitted order is triggered. Based on three fuzzy sets, fuzzy reasoning and defuzzification are performed using a fuzzy rule base to determine the output value, and a fuzzy early warning strategy is determined based on the output value. The fuzzy early warning strategy includes: prohibition of passage, restriction of passage, and permission of passage. For example, when the output value is ≥ 0.7, a prohibition of passage is generated; when 0.4 ≤ output value < 0.7, a restriction of passage is generated; and when the output value < 0.4, a permission of passage is generated.

[0068] Preferably, combined with Figure 2 As shown, this disclosure provides a method for early warning of geological hazards on bridge slopes based on multi-source data fusion, including:

[0069] Step S101: Obtain monitoring data for the bridge slope to be measured. This monitoring data includes geological and topographical data, meteorological and hydrological data, engineering activity data, and displacement and deformation data.

[0070] Step S102: Determine whether there is abnormal data based on the monitoring data. If there is, treat the abnormal data as preliminary abnormal data.

[0071] Step S103: Based on the preliminary abnormal data, predict the first probability of landslide on the bridge slope to be tested, and based on the monitoring data, predict the second probability of landslide on the bridge slope to be tested, and determine the landslide risk level of the bridge slope to be tested based on the first probability and the second probability.

[0072] Step S104: If the landslide risk level is greater than or equal to the first risk level, a disaster chain early warning analysis of landslide-debris flow-bridge damage is performed using a preset early warning model to obtain the early warning analysis results, and an early warning is issued based on the early warning analysis results.

[0073] The bridge slope geological disaster early warning method based on multi-source data fusion provided in this disclosure acquires monitoring data of the bridge slope to be tested. Abnormal data in the monitoring data is used as preliminary abnormal data, reflecting local anomalies in the bridge slope, while the monitoring data reflects the overall state of the bridge slope. Thus, the preliminary data is used to predict the first probability of landslides on the bridge slope, and the monitoring data is used to predict the second probability of landslides. Based on the first and second probabilities, the landslide risk level of the bridge slope is determined. This allows for a more accurate prediction of whether a landslide will occur based on both local anomalies and the overall state of the bridge slope. When the landslide risk level is greater than or equal to the first risk level, it indicates a higher probability of landslides on the bridge slope. In this case, an early warning analysis of the landslide-debris flow-bridge damage disaster chain is performed, enabling timely and accurate early warning of the landslide-debris flow-bridge damage disaster chain.

[0074] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made to it in form and detail without departing from the scope defined by the claims of the present invention.

Claims

1. A bridge slope geological disaster early warning system based on multi-source data fusion, characterized in that, include: The acquisition module is used to acquire monitoring data of the bridge slope to be measured; the monitoring data includes geological and topographical data, meteorological and hydrological data, engineering activity data, and displacement and deformation data. The processing module is used to determine whether there is abnormal data based on the monitoring data, and if so, to treat the abnormal data as preliminary abnormal data. The prediction module is used to predict a first probability of landslide on the slope of the bridge under test based on the prior abnormal data, and to predict a second probability of landslide on the slope of the bridge under test based on the monitoring data, and to determine the landslide risk level of the slope of the bridge under test based on the first probability and the second probability. The early warning module is used to perform a disaster chain early warning analysis of landslide-debris flow-bridge damage using a preset early warning model if the landslide risk level is greater than or equal to the first risk level, obtain the early warning analysis results, and issue an early warning based on the early warning analysis results.

2. The system according to claim 1, characterized in that, The prediction module includes: The first submodule is used to calculate the subprobability of each type of early abnormal data causing a landslide on the slope of the bridge under test, based on each type of early abnormal data. The second submodule is used to determine the first probability based on the combination of various subprobabilities.

3. The system according to claim 2, characterized in that, The first submodule includes: The first calculation unit is used to calculate the probability of natural instability of the bridge slope under test, calculate the correlation between each type of early abnormal data and the bridge slope under test, and calculate the probability of each type of early abnormal data occurring. The second calculation unit is used to calculate, for each type of early abnormal data, the sub-probability of the landslide occurring on the slope of the bridge under test due to each early abnormal data, based on the instability probability, correlation degree, and probability of the occurrence of the early abnormal data, using Bayes' theorem, as follows: Wherein, P(A / B) i Let P(B) be the sub-probability that the i-th type of prior abnormal data leads to a landslide on the slope of the bridge under test. i / A) represents the correlation between the i-th type of early abnormal data and the slope of the bridge to be tested, P(A) is the instability probability, and P(B) is the instability probability. i ) represents the probability of the i-th type of early abnormal data appearing.

4. The system according to claim 3, characterized in that, The first computing unit is specifically used for: Obtain the total number of slopes and the number of slopes that have experienced natural instability within the preset monitoring area; Based on the total number of slopes within the preset monitoring area and the number of slopes that have experienced natural instability within the preset monitoring area, the probability of natural instability of the slopes within the preset monitoring area is calculated using the classical probability formula, as shown below: Where P(C) is the probability of natural instability of the slope in the preset monitoring area, m is the total number of slopes in the preset monitoring area, and n is the number of slopes that have natural instability in the preset monitoring area. The probability of natural instability of the slope within the preset monitoring area is taken as the instability probability of natural instability of the bridge slope under test.

5. The system according to claim 1, characterized in that, The prediction module includes a third sub-module; the third sub-module is specifically used for: Determine the first level corresponding to the first probability and the second level corresponding to the second probability, respectively; If either the first level or the second level is greater than or equal to the second risk level, then the highest level among the first level and the second level shall be determined as the landslide risk level of the bridge slope to be tested. If both the first level and the second level are less than the second risk level, then the landslide risk level of the bridge slope to be tested is determined based on the first level, the first preset weight corresponding to the first level, and the second preset weight corresponding to the second level; wherein, the second risk level is greater than the first risk level.

6. The system according to any one of claims 1 to 5, characterized in that, It also includes a model building module; the model building module includes: The node submodule is used to obtain nodes related to the "landslide-debris flow-bridge damage" disaster chain. Among them, the nodes include three types of nodes: disaster-causing factors, disaster-inducing environment, and disaster events. The disaster-causing factor nodes include landslide characteristics and sliding path. The disaster-inducing environment nodes include rainfall intensity, valley longitudinal gradient, debris flow fluid characteristics, relative position of bridge and gully, and bridge structure. The disaster event nodes include debris flow level and bridge damage level. The determination submodule is used to determine the influence relationship between each node based on the disaster mechanism and chain transmission law of the "landslide-debris flow-bridge damage" disaster chain, and to determine the directed edge information between each node based on the influence relationship; The case submodule is used to obtain multiple historical cases based on the "landslide-debris flow-bridge damage" disaster chain; the historical cases include node variables corresponding to each node; The calculation submodule is used to calculate the conditional probability between nodes using the node variables of multiple historical cases. The model submodule is used to construct a Bayesian network model based on the information of each node, the directed edges between each node, and the conditional probabilities between each node, and to determine the constructed Bayesian network model as the early warning model.

7. The system according to claim 6, characterized in that, The early warning module includes: The acquisition submodule is used to estimate the landslide characteristics, sliding path and debris flow characteristics of the bridge slope to be tested if the landslide risk level is greater than or equal to the first risk level, and to acquire the rainfall intensity, valley longitudinal gradient, relative position of the bridge and the gully and the bridge structure of the bridge slope to be tested. The early warning submodule is used to input the landslide characteristics, sliding path, debris flow fluid characteristics, rainfall intensity, valley longitudinal gradient, relative position of the bridge and the gully, and bridge structure corresponding to the slope of the bridge under test into the early warning model for analysis, obtain the corresponding early warning analysis results, and issue an early warning based on the early warning analysis results; the early warning analysis results include the disaster risk of debris flow and bridge damage.

8. The system according to claim 7, characterized in that, The early warning submodule is specifically used for: A fuzzy early warning strategy is determined based on the landslide risk level of the bridge slope under test and the early warning analysis results. Early warning is issued based on the aforementioned fuzzy warning strategy.

9. A method for early warning of geological hazards on bridge slopes based on multi-source data fusion, characterized in that, include: Acquire monitoring data of the bridge slope to be tested; the monitoring data includes geological and topographical data, meteorological and hydrological data, engineering activity data, and displacement and deformation data; Based on the monitoring data, determine whether there is any abnormal data; if so, treat the abnormal data as preliminary abnormal data. Based on the aforementioned early abnormal data, a first probability of landslide occurring on the slope of the bridge under test is predicted, and a second probability of landslide occurring on the slope of the bridge under test is predicted based on the monitoring data. Based on the first probability and the second probability, the landslide risk level of the slope of the bridge under test is determined. If the landslide risk level is greater than or equal to the first risk level, a disaster chain early warning analysis of landslide-debris flow-bridge damage is performed using a preset early warning model to obtain early warning analysis results, and an early warning is issued based on the early warning analysis results.

Images