CN-121997003-A - Intelligent slope health degree assessment system and method based on multi-source data fusion

CN121997003ACN 121997003 ACN121997003 ACN 121997003ACN-121997003-A

Abstract

The invention relates to the technical field of slope health evaluation, and particularly discloses a slope health intelligent evaluation system and method based on multi-source data fusion. The core technology is that each characteristic value is mapped through a Gaussian membership function and converted into initial credibility distribution under different risk grades to form a risk credibility vector. And carrying out differential weighting on the credibility vector to generate an evidence item representing the local risk contribution. Finally, all weighted evidence items are input into an improved evidence reasoning model, the contradiction of multi-source information is effectively solved, the global membership probability of the whole slope on each risk level is solved, a continuous integral health score which can dynamically reflect the stable state is generated in a fusion mode, the crossing from local heterogeneous perception to integral intelligent research and judgment is realized, and a core basis is provided for slope safety early warning and accurate management and control.

Inventors

Bai Dongxin
ZENG RUI
ZENG JINPENG
ZHOU ZHIWU
LIU YI
HUANG MIAO
WANG LICHANG
LU GUANGYIN
FANG JI
PENG ZHIFEI
CHEN JIA
HE HAIYU
TIAN FU
LI SHUAI
DENG SIWEI

Assignees

湖南致力工程科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260408

Claims (9)

1. The intelligent evaluation method for the health degree of the side slope based on the multi-source data fusion is characterized by comprising the following steps: Receiving slope multisource monitoring data collected by a plurality of monitoring terminals, wherein the slope multisource monitoring data comprises a displacement time sequence and rainfall environment data; The method comprises the steps that a sliding window with a preset step length is adopted to segment slope multi-source monitoring data, the change rate characteristics of a displacement time sequence are extracted through least square regression calculation, and a multi-dimensional characteristic matrix aiming at a single monitoring terminal is formed by combining rainfall environment data; Mapping each characteristic value in the multidimensional characteristic matrix into a preset Gaussian membership function, calculating to obtain initial reliability distribution of the monitoring terminal under different risk grades, and constructing and forming a risk reliability vector; The method comprises the steps of calling pre-stored weight configuration information, wherein the weight configuration information comprises relative weight values distributed to sensors in different monitoring dimensions and different positions by using a hierarchical analysis method; applying relative weight values to carry out weighted mapping on the risk reliability vectors to generate weighted evidence items representing local regional risk contributions; inputting weighted evidence items corresponding to a plurality of monitoring terminals into a weighted evidence reasoning model, executing evidence conflict recognition and joint probability synthesis processing in the weighted evidence reasoning model, eliminating data conflicts among different monitoring dimensions through an iterative synthesis algorithm, and calculating to obtain global membership probability of the whole slope on each risk evaluation level; And carrying out quantitative weighted summation on the global membership probability, generating a health score representing the overall stability degree of the slope, and sending the health score to the early warning terminal to trigger corresponding safety management and control logic.
2. The intelligent evaluation method for the health of the side slope based on multi-source data fusion according to claim 1, wherein the step of obtaining the multi-dimensional feature matrix for the single monitoring terminal is as follows: Moving a sliding window along a time axis according to a preset step length, and intercepting a displacement time sequence in slope multisource monitoring data by using the sliding window to obtain a displacement time data segment in a current window coverage time range; synchronously intercepting rainfall environment data in the slope multisource monitoring data by utilizing a sliding window to acquire a rainfall environment data segment consistent with the time point of the displacement time data segment; Taking each time point in the displacement time data period as an independent variable, and taking a displacement acquisition value corresponding to each time point as a dependent variable to establish a unitary linear fitting equation; Calculating a unitary linear fitting equation slope value which enables the residual square sum of all displacement acquisition values to be minimum through a least square regression algorithm, and setting the unitary linear fitting equation slope value as the change rate characteristic of the displacement time sequence; Extracting rainfall numerical values of each monitoring and collecting time in the rainfall environment data segment, and performing accumulation and summation operation on all the rainfall numerical values in the rainfall environment data segment to obtain accumulated rainfall characteristics corresponding to the sliding window; and carrying out dimension alignment and array splicing on the change rate characteristic and the accumulated rainfall characteristic according to the physical identification of the single monitoring terminal, and constructing and forming a multidimensional characteristic matrix aiming at the single monitoring terminal.
3. The intelligent assessment method for the health of a side slope based on multi-source data fusion according to claim 1, wherein the construction logic of the risk confidence vector comprises: Extracting change rate features and accumulated rainfall features from the multidimensional feature matrix, and obtaining a preset level center value corresponding to each risk level; Calculating deviation square values between the change rate characteristics, the accumulated rainfall characteristics and preset level center values corresponding to the risk levels respectively; Scaling the deviation square value by using a preset standard deviation, and performing negative operation by taking the scaled numerical value as a power exponent of a natural logarithm base to obtain membership scores of all features aiming at different risk grades; Carrying out saturation smoothing treatment on membership scores at the head end and the tail end of a risk level sequence according to the change rate characteristics and the numerical intervals in which the accumulated rainfall characteristics are positioned to obtain corrected membership scores of the characteristics under different risk levels; carrying out averaging treatment on the characteristic correction membership scores of the same monitoring terminal under the same risk level to generate initial credibility distribution of the monitoring terminal under different risk levels; and arranging the initial reliability distribution corresponding to each risk level according to a preset sequence from low risk to high risk, and constructing and forming a risk reliability vector.
4. The intelligent evaluation method for the health of a side slope based on multi-source data fusion according to claim 3, wherein the calculation process of the corrected membership scores of each feature under different risk levels comprises the following steps: Determining the lowest risk level at the initial position and the highest risk level at the final position in the risk level sequence; Comparing the change rate characteristic and the accumulated rainfall characteristic with a preset level central value corresponding to the lowest risk level respectively, and updating the membership score corresponding to the lowest risk level to a preset constant peak value under the condition that the characteristic value is smaller than the preset level central value corresponding to the lowest risk level; comparing the change rate characteristic and the accumulated rainfall characteristic with a preset level central value corresponding to the highest risk level respectively, and updating the membership score corresponding to the highest risk level to a preset constant peak value under the condition that the characteristic value is larger than the preset level central value corresponding to the highest risk level; And (3) keeping the membership scores of the intermediate risk levels except the lowest risk level and the highest risk level in the risk level sequence unchanged, and generating corrected membership scores of all the features under different risk levels by carrying out sequence recombination on the updated membership scores and the membership scores which are kept unchanged.
5. The intelligent evaluation method for the health of the side slope based on multi-source data fusion according to claim 1, wherein the preset logic of the weight configuration information is as follows: establishing a reference sample set containing a plurality of slope history destabilizing events, and extracting displacement time sequences, rainfall environment data and corresponding slope stability state labels in a preset time period before destabilization aiming at each history destabilizing event; Dividing a reference sample set into a deformation characteristic layer, a rainfall environment layer and a position distribution layer according to monitoring dimensions, and constructing an initial judgment matrix for pairwise comparison based on the associated contribution degree of each monitoring dimension to the slope stability state label; Setting the maximum eigenvalue and the consistency ratio of the matrix as iteration parameters, carrying out cyclic iterative computation on the initial judgment matrix by adopting a power method, and solving the eigenvector of the initial judgment matrix under the current iteration times; Weighting calculation is carried out on the reference sample set by utilizing each dimension component in the feature vector to obtain a sample risk evaluation value, and a mean square error value between the sample risk evaluation value and the slope stability state label is calculated; Gradient adjustment is carried out on elements in the initial judgment matrix according to the mean square error value, and the consistency proportion value in the current iteration parameter is synchronously calculated; Stopping iteration and extracting a current feature vector as a target weight vector under the condition that the consistency ratio value is smaller than a preset consistency threshold value and the mean square error value reaches global minimum distribution; and mapping and packaging the target weight vector according to physical indexes of sensors with different monitoring dimensions and different positions to generate weight configuration information.
6. The intelligent evaluation method for the health degree of the side slope based on the multi-source data fusion according to claim 1, wherein the calculation to obtain the global membership probability of the whole side slope on each risk evaluation level comprises the following steps: Extracting credibility distribution values corresponding to different risk evaluation grades in each weighted evidence item, and calculating to obtain basic credibility quality distributed to each risk evaluation grade and residual uncertainty quality not distributed to a clear grade according to the relative weight of each monitoring terminal; the mutual exclusion degree of different monitoring terminals under the same risk evaluation level is identified by performing cross product summation operation on the basic reliability quality among different monitoring terminals, and a conflict interference factor for representing the evidence contradiction degree is obtained by calculation; Adopting a recursive synthesis algorithm, taking the basic credibility quality of a first monitoring terminal as an initial fusion state, introducing the basic credibility quality of a subsequent monitoring terminal one by one, and carrying out gain correction on a cross product result by utilizing a conflict interference factor in each iteration to generate a joint probability distribution of an intermediate state; After the iterative synthesis of all monitoring terminals is completed, extracting the finally generated joint probability distribution and the synthesized residual uncertainty quality, and reallocating the residual uncertainty quality to each risk evaluation grade in proportion by executing normalized mapping processing; And defining the assigned numerical value as global membership probability of the whole side slope on each risk evaluation level.
7. The intelligent assessment method for the health of a side slope based on multi-source data fusion according to claim 6, wherein the calculation of the basic confidence quality assigned to each risk assessment level and the residual uncertainty quality not assigned to an explicit level comprises: For the weighted evidence item, respectively extracting corresponding credibility distribution values according to the preset five risk evaluation grades; Performing product operation on the reliability distribution values corresponding to the risk evaluation grades and the relative weights of the monitoring terminals respectively to generate basic reliability quality aiming at the risk evaluation grades; calculating the difference between the value 1 and the relative weight to obtain a weight discount mass component representing the dilution of the trust degree of the monitoring terminal by the weight; Performing accumulation summation operation on the basic credibility quality corresponding to all risk evaluation grades, and calculating a difference value between a numerical value 1 and an accumulation summation result to obtain an inherent uncertain quality component representing the ambiguity of the monitoring data; the weight discounted mass component is linearly summed with the inherent uncertainty mass component to generate a remaining uncertainty mass that is not assigned to the explicit level.
8. The intelligent evaluation method of the slope health based on multi-source data fusion according to claim 6, wherein the recursive synthesis algorithm is adopted, the basic reliability quality of the first monitoring terminal is used as an initial fusion state, the basic reliability quality of the subsequent monitoring terminals is introduced one by one, gain correction is performed on the cross product result by using a conflict interference factor in each iteration, and the generation of the joint probability distribution of the intermediate state comprises: assigning the basic credibility quality of the first monitoring terminal on each risk evaluation level to the current fusion probability vector to serve as an initial fusion state of the recursion calculation; Extracting the basic credibility quality of the components in the current fusion probability vector and the next monitoring terminal to be fused on the corresponding risk evaluation level, and calculating the value of the cross product of the components and the basic credibility quality; Calculating a difference value between the numerical value 1 and the conflict interference factor, and setting the reciprocal of the difference value as a gain correction coefficient; Multiplying the cross product value by using a gain correction coefficient to obtain an updated current fusion probability vector so as to complete evidence synthesis of the current round; And circularly executing the steps of extraction, calculation and amplification according to the arrangement sequence of the monitoring terminals until the basic credibility quality of all the monitoring terminals is subjected to fusion processing, and determining the finally obtained current fusion probability vector as the joint probability distribution of the intermediate state.
9. The intelligent evaluation system for the slope health degree based on the multi-source data fusion is characterized in that the intelligent evaluation system for the slope health degree based on the multi-source data fusion is realized based on the intelligent evaluation method for the slope health degree based on the multi-source data fusion as claimed in any one of claims 1 to 8, and comprises the following steps: the multi-source data acquisition module is used for receiving slope multi-source monitoring data acquired by the plurality of monitoring terminals, wherein the slope multi-source monitoring data comprise displacement time sequences and rainfall environment data; The risk vector construction module is used for carrying out segmentation processing on slope multisource monitoring data by adopting a sliding window with a preset step length, extracting the change rate characteristic of a displacement time sequence by least square regression calculation, and combining rainfall environment data to form a multidimensional characteristic matrix aiming at a single monitoring terminal; the risk weighting calculation module is used for calling pre-stored weight configuration information, wherein the weight configuration information comprises relative weight values distributed to different monitoring dimensions and sensors at different positions by using a hierarchical analysis method; The slope health evaluation module inputs weighted evidence items corresponding to the plurality of monitoring terminals into a weighted evidence reasoning model, performs evidence conflict recognition and joint probability synthesis processing in the weighted evidence reasoning model, eliminates data conflicts among different monitoring dimensions through an iterative synthesis algorithm, obtains global membership probabilities of the whole slope on each risk evaluation level through calculation, performs quantitative weighted summation on the global membership probabilities, generates health degree scores representing the whole stability degree of the slope, and sends the health degree scores to the early warning terminals to trigger corresponding safety management and control logic.

Description

Intelligent slope health degree assessment system and method based on multi-source data fusion Technical Field The invention belongs to the technical field of slope health evaluation, and relates to a slope health intelligent evaluation system and method based on multi-source data fusion. Background The side slope geological disaster is used as a main hidden danger threatening the safety of infrastructure and lives and properties of residents, the disaster mechanism is complex, the concealment and burst are strong, and the prevention and control difficulty is high. The multisource monitoring data is used as 'nerve endings' for sensing the slope state, and comprises key dimensions such as displacement, rainfall, deep displacement and the like, the accuracy of acquisition and the depth of analysis directly determine the timeliness and reliability of geological disaster early warning, and the multisource monitoring data is an indispensable decision basis in landslide control work. It is noted that the multisource monitoring data are not isolated, and together carry information reflecting the overall state of the system, but often have differences in dimension, accuracy and reliability and contain significant uncertainty. For example, abnormal changes in local displacement may be coupled with environmental factors, and a single sensor alarm may be insufficient to determine overall risk. However, most existing health assessment methods still have limitations in this regard. The traditional method often depends on threshold judgment of single or few key indexes, is difficult to comprehensively utilize complementation and verification value of multidimensional information, and is easy to generate false alarm or false omission due to false alarm or data fluctuation of individual sensors. More importantly, existing methods are weak in dealing with inherent ambiguity and uncertainty of the monitored data, and typically employ simple weighted averages or deterministic models, failing to explicitly characterize and infer differences in confidence levels from disparate sources and contradictions between them. This results in the final assessment result being probably a fragile "consensus" that fails to effectively distinguish the "deterministic" and "uncertain" portions of the information, making the health index insensitive to early, weak abnormal conditions, early warning delays, and difficult to interpret. Disclosure of Invention In view of the problems in the prior art, the invention provides a slope health intelligent evaluation system and a slope health intelligent evaluation method based on multi-source data fusion, which are used for solving the technical problems. In order to achieve the above and other objects, the present invention adopts the following technical scheme: the first aspect of the invention provides an intelligent evaluation method for the health degree of a side slope based on multi-source data fusion, which comprises the following steps: Receiving slope multisource monitoring data collected by a plurality of monitoring terminals, wherein the slope multisource monitoring data comprises a displacement time sequence and rainfall environment data; The method comprises the steps that a sliding window with a preset step length is adopted to segment slope multi-source monitoring data, the change rate characteristics of a displacement time sequence are extracted through least square regression calculation, and a multi-dimensional characteristic matrix aiming at a single monitoring terminal is formed by combining rainfall environment data; Mapping each characteristic value in the multidimensional characteristic matrix into a preset Gaussian membership function, calculating to obtain initial reliability distribution of the monitoring terminal under different risk grades, and constructing and forming a risk reliability vector; The method comprises the steps of calling pre-stored weight configuration information, wherein the weight configuration information comprises relative weight values distributed to sensors in different monitoring dimensions and different positions by using a hierarchical analysis method; applying relative weight values to carry out weighted mapping on the risk reliability vectors to generate weighted evidence items representing local regional risk contributions; inputting weighted evidence items corresponding to a plurality of monitoring terminals into a weighted evidence reasoning model, executing evidence conflict recognition and joint probability synthesis processing in the weighted evidence reasoning model, eliminating data conflicts among different monitoring dimensions through an iterative synthesis algorithm, and calculating to obtain global membership probability of the whole slope on each risk evaluation level; And carrying out quantitative weighted summation on the global membership probability, generating a health score representing the overall stability degree of the slop