CN116090696A - Landslide geological disaster risk classification prediction method suitable for mountain railway line - Google Patents
Landslide geological disaster risk classification prediction method suitable for mountain railway line Download PDFInfo
- Publication number
- CN116090696A CN116090696A CN202211535856.XA CN202211535856A CN116090696A CN 116090696 A CN116090696 A CN 116090696A CN 202211535856 A CN202211535856 A CN 202211535856A CN 116090696 A CN116090696 A CN 116090696A
- Authority
- CN
- China
- Prior art keywords
- landslide
- data
- prediction
- elm
- geological disaster
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 55
- 238000012544 monitoring process Methods 0.000 claims abstract description 67
- 239000002245 particle Substances 0.000 claims abstract description 35
- 239000002689 soil Substances 0.000 claims abstract description 34
- 239000011435 rock Substances 0.000 claims abstract description 21
- 238000013058 risk prediction model Methods 0.000 claims abstract description 20
- 238000005457 optimization Methods 0.000 claims abstract description 13
- 230000006870 function Effects 0.000 claims description 22
- 238000005065 mining Methods 0.000 claims description 18
- 238000012549 training Methods 0.000 claims description 17
- 210000002569 neuron Anatomy 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 11
- 230000005284 excitation Effects 0.000 claims description 10
- 238000012360 testing method Methods 0.000 claims description 9
- 238000007418 data mining Methods 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000010801 machine learning Methods 0.000 abstract description 5
- 238000009412 basement excavation Methods 0.000 abstract 1
- 210000004027 cell Anatomy 0.000 description 10
- 238000012876 topography Methods 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 4
- 230000015572 biosynthetic process Effects 0.000 description 4
- 239000004927 clay Substances 0.000 description 4
- 238000007477 logistic regression Methods 0.000 description 4
- 238000012706 support-vector machine Methods 0.000 description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 238000013138 pruning Methods 0.000 description 1
- 238000012502 risk assessment Methods 0.000 description 1
- 238000010845 search algorithm Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
- G06Q50/265—Personal security, identity or safety
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Tourism & Hospitality (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Economics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Primary Health Care (AREA)
- Computer Security & Cryptography (AREA)
- Game Theory and Decision Science (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Testing Or Calibration Of Command Recording Devices (AREA)
Abstract
The invention belongs to the technical field of landslide geological disaster risk prediction based on machine learning, and particularly relates to a landslide geological disaster risk classification prediction method suitable for mountain railway lines. According to the method, landslide is divided according to the rock and soil types, main influencing factors of each type of landslide are excavated by using an association rule excavation algorithm, and the main influencing factors are fused with the earth surface deformation to generate a sample data set of each type of landslide. And respectively establishing an extreme learning machine algorithm landslide geological disaster risk prediction model after optimization of a particle swarm algorithm aiming at a sample data set of each type of landslide. Then dividing the railway line into different monitoring areas according to the type of the rock-soil landslide, and further dividing each monitoring area into smaller grid units by adopting a grid dividing method. And inputting the monitoring data of each grid cell into a corresponding prediction model to obtain a prediction result of each grid cell, and summarizing the prediction results of each grid cell to obtain a landslide geological disaster risk prediction result of the monitoring area, thereby realizing accurate prediction.
Description
Technical Field
The invention belongs to the technical field of landslide geological disaster risk prediction based on machine learning, and particularly relates to a landslide geological disaster risk classification prediction method suitable for mountain railway lines.
Background
When landslide geological disasters occur, huge economic losses and even casualties can be caused for people. In order to reduce the loss caused by landslide geological disasters, the probability of occurrence of landslide geological disasters needs to be predicted by adopting a technical means, and corresponding preventive measures are taken in advance. Therefore, many researchers aim at researching a landslide hazard risk prediction model method, so that the method can help related departments to take corresponding measures in advance, and the loss caused by landslide hazards is reduced.
There are two main ways to predict landslide geological disaster risk: 1) Landslide geological disaster risk prediction based on single factors, such as landslide geological disaster risk prediction based on rainfall factors or based on surface deformation amounts. The method for predicting based on the single factor data has the advantages of simple calculation process, low prediction accuracy and poor applicability. 2) The landslide geological disaster risk prediction method based on more relevant factor data has the advantages of higher prediction precision and low calculation processing cost and universality, and is based on geological condition factor data, meteorological factor data, surface deformation factor data and the like.
Because the railway passes through a plurality of regions, the geological condition of each region along the line has large difference, the topography and the meteorological changes are complex, and particularly the topography, the geological structure and the meteorological environment of the region are extremely complex along the railway in the mountain area, so that a plurality of difficulties are brought to landslide geological disaster risk prediction. The conventional landslide geological disaster risk prediction method is difficult to effectively solve the landslide geological disaster risk prediction problem along the railway in the mountain area. Therefore, researchers propose to add various sensor devices into the traditional method to analyze and process monitoring data online in real time, but because a prediction model is built in a specific area, the model has weak applicability in different areas.
In recent years, with the vigorous development of artificial intelligence and big data, machine learning algorithms such as logistic regression, support vector machines, BP neural networks, extreme Learning Machines (ELM) and other model algorithms are widely applied in landslide geological disaster risk prediction gradually, and good effects are obtained. However, when a logistic regression algorithm is used for establishing a model, the nonlinear problem cannot be solved, the prediction accuracy is not high enough, and the logistic regression algorithm is not suitable for landslide geological disaster risk prediction in a railway line area; when a support vector machine algorithm is used for establishing a model, compared with a logistic regression algorithm, the support vector machine algorithm can predict the nonlinear problem, but the algorithm is not suitable for training a large-scale sample set, and the areas spanned by the railways along the lines are large, and the sample data scale is large, so that the algorithm is not suitable for learning a landslide geological disaster risk prediction model of the areas along the railways; the BP neural network and the ELM extreme learning algorithm have no defects of the two algorithms, and can be used for learning landslide geological disaster prediction models in areas along the railway. BP neural network prediction accuracy is high, but the time consumption is long, and the condition of fitting easily appears. Compared with BP neural network, ELM extreme learning machine has the characteristics of faster learning rate, stronger generalization capability and the like, and is more suitable for landslide geological disaster risk prediction along mountain railway lines. However, in practical application, the sample points of landslide occurrence in the railway line area are far less than the sample points of landslide non-occurrence, and the unbalance of the sample data easily causes the model prediction result learned by the ELM algorithm to be more biased to the result of landslide non-occurrence, thereby seriously affecting the prediction accuracy.
Therefore, the landslide geological disaster risk classification prediction method suitable for the railway line in the mountain area is researched, and has important significance for safe operation of the railway.
Disclosure of Invention
The invention aims to provide a landslide geological disaster risk classification prediction method suitable for mountain railway lines, which utilizes an association rule mining algorithm to mine regional geological structures, geological conditions, topography, meteorological data and the like of different rock-soil type landslide, finds out main influencing factors of four types of typical rock-soil type landslide, fuses the main influencing factor data with surface deformation monitoring data, adopts a machine learning method to perform modeling learning on landslide geological disaster risk prediction models of the four types of typical rock-soil regions, constructs landslide geological disaster risk prediction models based on the different rock-soil type regions, and realizes landslide geological disaster risk prediction of different rock-soil type monitoring regions along the mountain railway lines.
In order to achieve the above purpose, the invention adopts the following technical scheme:
a landslide geological disaster risk classification prediction method suitable for mountain railway lines comprises the following steps:
s1, acquiring sample data and earth surface deformation data of a rock-soil landslide type, wherein the rock-soil type comprises: viscous soil landslide, loess landslide, clastic rock landslide and rock landslide;
s2, performing data mining on main influence factors of each type of landslide hazard sample data by using an association rule mining algorithm (Apriori), and finding out the main influence factors associated with landslide risks;
s3, fusing the main influencing factors of each type of excavated landslide with the surface deformation data to generate four landslide type data sets, and randomly dividing each data set into a training set and a testing set;
s4, inputting a training set of each type of landslide into an ELM (electronic learning machine), optimizing parameter input weights and hidden layer deviation values in the ELM by adopting a Particle Swarm Optimization (PSO), and optimizing the embedding dimension and hidden layer node number in the ELM by adopting a traversal search method to obtain a landslide geological disaster risk prediction model optimized by the PSO;
s5, inputting the test set into a landslide geological disaster risk prediction model optimized by a PSO algorithm for testing, and finally obtaining landslide prediction models of different types;
s6, dividing the railway line area into a viscous soil landslide monitoring area, a loess landslide monitoring area, a clastic rock landslide monitoring area and a rock landslide monitoring area according to the rock and soil type;
s7, collecting landslide monitoring data of each monitoring area in real time, and then respectively inputting the landslide monitoring data of each monitoring area into a corresponding landslide prediction model for prediction.
Further, the detailed process of S2 includes:
s2.1, preprocessing the acquired sample data to obtain a standard sample data set;
s2.2, carrying out de-duplication weighting treatment on the standard sample data set by adopting a HashMap function method to obtain a weighted sample set without repeated data;
s2.3, carrying out association rule mining on the weighted sample set by using an Apriori algorithm, and finding out major influence factors which are associated with landslide risks.
Furthermore, in the process of mining data by using the association rule mining Apriori algorithm, the S2.3 adopts the multithread programming of distributed parallel computing to realize the program execution of the Apriori algorithm, so as to accelerate the association rule mining computation processing speed of landslide factors and disaster risks.
Further, the step S4 includes the following steps:
s4.1, the ELM extreme learning machine model consists of an input layer, an hidden layer and an output layer, wherein an excitation function of the hidden layer is set to be g (omega, X, b), and then the ELM extreme learning machine expression is as follows:
in which y i Representing ELM desired output; beta i The connection weight of the i-th neuron of the hidden layer and the neuron of the output layer is obtained; ωw i Representing the connection weight of the input layer neuron to the ith neuron of the hidden layer; b i Is the threshold of the hidden layer neuron, g (ωw i *x j +b i ) Is an excitation function; the Hardlim function is adopted as an excitation function of the ELM extreme learning model, and the expression is as follows:
s4.2, randomly dividing an input training set into an A part and a B part, wherein the A part adopts a Particle Swarm Optimization (PSO) to optimize a parameter input weight and an hidden layer deviation value in the ELM extreme learning machine, and the B part adopts a traversal search method to optimize an embedding dimension and a hidden layer node number in the ELM extreme learning machine;
the optimization process of the parameter input weight and the hidden layer deviation value in the ELM extreme learning machine by adopting a Particle Swarm Optimization (PSO) is as follows:
input weight w in ELM i Hidden layer node bias b i As particles of a Particle Swarm Optimization (PSO), taking the root mean square difference of an ELM training sample as a fitness function, calculating the fitness value of each particle, comparing and continuously iterating to update the speed and the position of the particle until the error rate is minimum or the maximum iteration number, and finally obtaining ELM parameters after PSO optimization;
the update formula of the particle swarm algorithm is as follows:
wherein ωw is a weight, t is the number of iterations, V in For the speed of the particles c 1 、c 2 X is the position of the particle, r 1 、r 2 Is a random number in the range of [0,1 ]]Between them.
And optimizing the embedding dimension and the hidden layer node number in the ELM extreme learning machine by using a traversing search method, namely obtaining the optimal combination by continuously changing different embedding dimensions and hidden layer node numbers.
Further, in order to improve the positioning accuracy of the landslide occurrence location, the step S7 is to obtain the risk prediction result of each area by adopting the following steps:
s7.1, dividing grids of each monitoring area, and respectively collecting real-time monitoring data of sub-grids of each monitoring area;
s7.2, in each monitoring area, monitoring data of each sub-grid are input into a corresponding landslide geological disaster risk prediction model optimized by a PSO algorithm, and landslide prediction risk results of the sub-grids are obtained;
and S7.3, summarizing and calculating landslide prediction risk result data of each regional unit grid to obtain a final risk prediction result of the region.
Furthermore, the S7.2 calculates and processes the sub-grid data of each region of the data by adopting a multithreading parallel method.
According to the invention, the Apriori algorithm is utilized to carry out association rule mining on sample data of different types of rock-soil landslide types, so that main influencing factors which are associated with landslide risks are found out, and the main influencing factors are fused with ground surface deformation data to obtain a sample data set. And (3) taking the sample data as input, and learning the sample data set by using an ELM algorithm in machine learning to obtain a geological disaster risk prediction model of the rock-soil landslide. On the basis, the railway is divided into a plurality of monitoring areas of different types according to the rock-soil type, and different landslide geological disaster risk prediction models are respectively built based on the different types of monitoring areas. And then, respectively inputting landslide monitoring data of each monitoring area acquired in real time into a corresponding landslide geological disaster risk prediction model to perform risk prediction, thereby realizing landslide geological disaster risk classification prediction along the railway. The invention is characterized in that in the prediction process: aiming at the problem that the ELM extreme learning machine easily falls into a local optimal solution in landslide geological disaster risk prediction, a particle swarm algorithm is adopted to optimize the ELM extreme learning machine, and a landslide geological disaster risk prediction model optimized by the particle swarm algorithm is constructed to improve the prediction accuracy of the original ELM extreme learning machine; aiming at the problems that a specific landslide occurrence place cannot be positioned due to the fact that a monitoring area is too large, the problem is solved by adopting a method of dividing and shrinking a prediction area through grid cells. Aiming at the problems that the monitoring areas along the railway are numerous, disaster risk prediction processing is required to be carried out on each monitoring area at the same time, a landslide geological disaster risk value processing scheme for calculating each monitoring area in parallel is adopted, so that the risk prediction value result of each monitoring area is predicted at the same time, the on-line landslide risk prediction processing requirement along the railway in the mountain area is met, and the prediction accuracy is high.
Drawings
FIG. 1 is a flow chart of a landslide geological disaster risk prediction model learning based on particle swarm improvement ELM;
FIG. 2 is a flow chart for classification and prediction of landslide geological disaster risk along a mountain railway;
FIG. 3 is a flow chart of a main influencing factor mining algorithm for a typical geotechnical landslide;
FIG. 4 is a diagram of an Apriori distributed parallel computing process;
FIG. 5 is a PSO-ELM network model training flow chart for landslide hazard risk prediction;
FIG. 6 is a classification prediction scheme for landslide along a mountain railway;
FIG. 7 is a grid division of landslide hazard monitoring areas;
FIG. 8 is a landslide geological disaster risk value prediction for a range of grid cells of a monitored area;
fig. 9 is a landslide geological disaster risk prediction calculation for a monitored area range.
Detailed Description
The invention will now be described in detail with reference to the drawings and examples.
Referring to fig. 1 and 2, the landslide geological disaster risk classification prediction method suitable for the mountain railway line provided by the embodiment includes the following steps:
s1, landslide can be divided into four types of cohesive soil landslide, loess landslide, broken rock landslide and rock landslide according to rock-soil types. Many factors affect landslide, such as rainfall, air temperature, vegetation cover, geological structure, geological conditions, topography, etc., but each type of landslide has different major factors. The method and the device collect sample data of four types of typical landslide geological disasters, and collect surface deformation data of areas along the railway at regular time;
s2, carrying out data mining on main influence factors of each type of landslide sample data by utilizing an association rule mining Apriori algorithm, and finding out main influence factors which are associated with landslide risks. Referring to fig. 3:
s2.1, preprocessing the collected data of different rock and soil types, namely formatting the collected data to obtain a standard sample data set.
S2.2, iteratively scanning the standard sample data set, and filtering repeated sample data through a HashMap function method to improve the data processing speed of the association mining algorithm. Specific: when the sample data is read, firstly, judging whether the sample data is repeated sample data or not through HashMap function calculation. If the sample data is repeated, the sample data is filtered out, and then the weight value of the same sample data which is stored is added with 1. Otherwise, the sample data is put into the standard data set and the weight value is set to 1. When the scanning is finished, the obtained sample data set is a weighted sample set without repeated data.
S2.3, performing association rule mining on the sample data set by using an Apriori algorithm, referring to FIG. 4:
setting the minimum support degree and the minimum confidence degree, firstly finding out frequent 1-item sets meeting the minimum support degree, and then obtaining frequent 2-item sets according to the frequent 1-item sets until a final mining result is obtained. Because the sample data size of landslide geological disaster risk analysis is relatively large, the data mining speed is relatively low, and therefore, distributed parallel calculation is needed to solve the problem, as shown in fig. 4 below, when the support degree of each item set is calculated iteratively, distributed parallel calculation is adopted, the support degree of a plurality of item sets is calculated simultaneously to speed up the processing speed, and the calculation results are summarized to obtain the final result of the calculation. And repeating the step of the previous step after connecting and pruning the obtained result until the n-frequent item set is empty.
The types of influencing factors involved in landslide geological disasters are shown in table 1.
TABLE 1 influence factors of landslide geological disasters
Because non-numeric data cannot be processed directly in data mining, it is necessary to discretize it, such as where formation lithology uses a single continuous numeric value to represent formation firmness. Geological structure, topography and topography use one-hot coding: fracture= (1, 0); fold= (0, 1, 0); level= (0, 1, 0); tilting= (0, 1); plateau= (1, 0); mountain area= (0, 1, 0); basin= (0, 1, 0); hills= (0, 1). Further, the continuous numerical data is subjected to discretization processing using a k-means clustering algorithm, such as dividing the gradient into [40.86,46.26], [46.26,57.66], [57.66,69.06], and [69.06,82.46].
Discretization results of the clay landslide sample data are shown in table 2.
Table 2 discretized results table for clay landslide sample data
And carrying out association rule mining on the data in the table 2 through an Apriori algorithm, setting the association rule with the support degree larger than 20% and the confidence degree larger than 70%, and finally obtaining the association rule as shown in the table 3.
TABLE 3 correlation rules for clay landslide samples
As is clear from table 3, the vegetation coverage, the soil water content, the current rainfall, the surface deformation amount, and the surface temperature have a relatively high influence on the viscous soil landslide. Thus, it can be determined that the main influencing factors of the cohesive soil landslide are vegetation coverage, soil moisture content, current rainfall, and surface temperature. Item 1 of data option D in table 2, indicated by D1 in table 3, [0.10,0.25], D2 indicates the second item of data of D [0.25,0.55], and so on, wherein the target parameter is whether landslide would occur.
The main influencing factors of loess landslide can be obtained by the same method, such as the current geological structure, the front rainfall, the surface temperature, the gradient and the elevation. The main influencing factors of the clastic rock landslide are temperature, soil water content, gradient and formation lithology. The main influencing factors of rock landslide are temperature, soil water content, elevation, gradient and formation lithology.
S3, a major influence factor which is related to each type of landslide risk is found out and fused with the ground surface deformation data to generate sample data sets of landslide of different types, and the sample data sets are randomly divided into a training set and a testing set.
S4, inputting the training set into an ELM extreme learning machine model, optimizing parameter input weights and hidden layer deviation values in the ELM extreme learning machine by adopting a Particle Swarm Optimization (PSO), and optimizing the embedded dimension and hidden layer node number in the ELM extreme learning machine by adopting a traversal search method to obtain a landslide geological disaster risk prediction model optimized by the PSO algorithm.
Although the extreme learning machine model (ELM) has the characteristics of high learning rate, strong generalization capability and the like, the extreme learning machine model is used for landslide geological disaster risk prediction and is required to solve the problem of unbalance of positive and negative samples. The ELM consists of an input layer, an hidden layer and an output layer, and an excitation function of the hidden layer is set to be g (omega, X, b), and then the ELM extreme learning machine expression is as follows:
in which y i Representing ELM desired output; beta i The connection weight of the i-th neuron of the hidden layer and the neuron of the output layer is obtained; w (w) i Representing the connection weight of the input layer neuron to the ith neuron of the hidden layer; b i Is the threshold of the hidden layer neuron, g (w i *x j +b i ) Is an excitation function. There are three excitation functions of ELM, namely the Sine function, the Radbas function and the Hardlim function. The experiment will use Hardlim as the excitation function, expressed as follows:
FIG. 5 is a PSO-ELM network model training flowchart for landslide hazard risk prediction, as shown in FIG. 5:
a. the input training set is randomly divided into two parts A and B.
b. Because landslide geological disaster negative sample data along the railway in the mountain area is far larger than positive samples, the negative samples refer to samples without landslide. The problem of locally optimal solutions is easily caused when ELM extreme learning machines are used. To solve this problem, the present embodiment uses a particle swarm algorithm (PSO) for inputting a weight W in an ELM extreme learning machine based on the A-part data i And hidden layer bias value b i Optimizing to obtain a global optimal solution; and (3) optimizing the embedding dimension m and the hidden layer node L number in the ELM extreme learning machine by adopting a traversal search method based on the part B data. PSO is a population intelligent global search algorithm, and is basically characterized in that d particles are arranged in an N-dimensional search space. And assuming that a certain particle alone obtains an optimal solution, the optimal solution is a local optimal solution, and meanwhile, the optimal solution is shared with the particles in the group to obtain a global optimal solution. And adjusting the speed and the position of all particles of the particle swarm according to the local optimal solution and the global optimal solution, and finally obtaining the optimal solution. Specific:
input weight w in ELM i Hidden layer node bias b i As particles of a Particle Swarm Optimization (PSO), taking the root mean square difference of an ELM training sample as a fitness function, calculating the fitness value of each particle, comparing and continuously iterating to update the speed and the position of the particle until the error rate is minimum or the maximum iteration number, and finally obtaining ELM parameters after PSO optimization;
the update formula of the particle swarm algorithm is as follows:
wherein w is a weight, t is the number of iterations, V in For the speed of the particles c 1 、c 2 X is the position of the particle, r 1 、r 2 Is a random number in the range of [0,1 ]]Between them.
And optimizing the embedding dimension and the hidden layer node number in the ELM extreme learning machine by using a traversing search method, namely obtaining the optimal combination by continuously changing different embedding dimensions and hidden layer node numbers.
c. Inputting the optimized parameters into a PSO-ELM algorithm for training iteration.
d. Performing accuracy judgment training on the iteration result, if the iteration result meets the preset accuracy requirement, performing step e, and otherwise returning to step b;
e. and outputting the constructed PSO-ELM landslide geological disaster prediction model.
S5, testing the PSO-ELM landslide geological disaster risk prediction model based on the test set.
S6, building different landslide risk prediction models aiming at different rock-soil type landslide. The railway in the complicated mountain area is divided into different monitoring areas such as a clay landslide monitoring area, a loess landslide monitoring area, a broken rock landslide monitoring area and a rock landslide monitoring area according to different rock and soil types.
S7, establishing different PSO-ELM landslide geological disaster risk prediction models based on different types of monitoring areas:
and collecting monitoring data in each monitoring area along the railway of the mountain area, selecting a corresponding PSO-ELM landslide geological disaster prediction model for calculation processing according to the rock-soil type, and then respectively outputting respective risk prediction results. See fig. 6 for a specific embodiment.
In practical application, because the areas along the railway are numerous, even if the areas are divided, the problem that a specific landslide occurrence place cannot be positioned due to the fact that a single monitoring area is too large still exists. In order to solve the problem, the risk prediction method adopts a method of dividing and shrinking prediction areas by grid cells to predict risks of all areas. Referring to fig. 7:
s7.1, assuming that the area of the original monitoring area of each type is 1km×1km, dividing the original monitoring area into a plurality of sub-grids, and then dividing the area of each grid unit into 100m×100m.
S7.2, on-line monitoring the flow data of each grid unit, and inputting the monitoring data of each sub-grid into a corresponding PSO-ELM landslide geological disaster prediction model to obtain a landslide geological disaster risk prediction value of the grid. The specific flow is shown in fig. 8.
And S7.3, summarizing and calculating the data of each grid cell of the monitoring area to obtain a final risk prediction result of the area, wherein the result comprises all grid cells which are fed back as safe and all grid cells which are fed back as dangerous in the monitoring area. The monitoring areas are numerous, and in order to update the information of each monitoring area synchronously, the embodiment starts multithreading to process the grid cell data of each monitoring area in parallel. As shown in fig. 9:
and after the data of the monitoring areas are input, starting a plurality of threads to process the grid cell data of each monitoring area, and summarizing and calculating to finally obtain the final prediction result of each monitoring area. The following formula is used for summary calculation of a certain monitored area:
wherein D represents a risk prediction value, ni is the number of risk level grid cells, the risk levels are classified into red, orange, yellow and blue, pi is a weight map table of each risk level as follows, and n is the total number of cell grids
Table 4 weight type mapping table
Type(s) | Weighting of |
Red color | 0.8 |
Orange color | 0.6 |
Yellow colour | 0.4 |
Blue color | 0.2 |
Green colour | 0.0 |
As can be seen from the description of the above embodiments, the PSO-ELM predictive model of the present invention has a better fitting effect than a simple ELM extreme learning machine. In the prediction experiments of the viscous soil landslide, the loess landslide, the clastic rock landslide and the rock landslide, the average error rate of the prediction results of the PSO-ELM extreme learning machine of the four landslide types is 0.67%, 0.82%, 5.3% and 6.1%. The average error rate of the prediction results of the pure ELM extreme learning machine of the four types of landslide is 4.6%, 3.8%, 6.4% and 6.8%. The result shows that the PSO-ELM prediction model provided by the invention can obviously reduce the prediction error. In the face of the problem that the accuracy of prediction results is not high in geological disaster risk prediction of the railway along the complicated mountain area, the monitoring areas are divided according to four typical rock-soil types, and after each type of landslide monitoring area is processed by adopting a prediction model of each type, the accuracy rate of the landslide monitoring area reaches more than 89%, so that the accuracy of landslide geological disaster prediction is remarkably improved.
Claims (6)
1. The landslide geological disaster risk classification prediction method suitable for the mountain railway line is characterized by comprising the following steps of:
s1, acquiring sample data and ground deformation data of different rock-soil landslide types, wherein the rock-soil landslide types comprise: viscous soil landslide, loess landslide, clastic rock landslide and rock landslide;
s2, carrying out data mining on main influence factors of each type of landslide hazard sample data by using an association rule mining algorithm, and finding out the main influence factors associated with landslide risks;
s3, fusing the main influencing factors of each type of excavated landslide with the surface deformation data to generate four landslide type data sets, and randomly dividing each data set into a training set and a testing set;
s4, inputting a training set of each type of landslide into an ELM (electronic learning machine), optimizing parameter input weights and hidden layer deviation values in the ELM by adopting a PSO (particle swarm optimization) algorithm, and optimizing the embedded dimension and hidden layer node number in the ELM by adopting a traversal search method to obtain a landslide geological disaster risk prediction model optimized by the PSO algorithm;
s5, inputting each type of landslide test set into a landslide geological disaster risk prediction model optimized by a PSO algorithm for testing, and finally obtaining landslide prediction models of different types;
s6, dividing the railway line area into a viscous soil landslide monitoring area, a loess landslide monitoring area, a clastic rock landslide monitoring area and a rock landslide monitoring area according to different rock and soil types;
s7, collecting landslide monitoring data of each monitoring area in real time, then respectively inputting the landslide monitoring data of each monitoring area into a corresponding landslide prediction model for prediction, and outputting after obtaining risk prediction results of each area.
2. The landslide geological disaster risk classification prediction method suitable for mountain railway lines according to claim 1, wherein the method comprises the following steps of: the detailed process of the S2 comprises the following steps:
s2.1, preprocessing the acquired sample data to obtain a standard sample data set;
s2.2, carrying out de-duplication weighting treatment on the standard sample data set by adopting a HashMap function method to obtain a weighted sample set without repeated data;
s2.3, carrying out association rule mining on the weighted sample set by using an Apriori algorithm, and finding out major influence factors which are associated with landslide risks.
3. The landslide geological disaster risk classification prediction method suitable for use along mountain railways of claim 2, wherein: and S2.3, in the process of mining data by using the association rule mining Apriori algorithm, the program execution of the Apriori algorithm is realized by adopting the multithread programming of distributed parallel calculation, so that the association rule mining calculation processing speed of landslide factors and disaster risks is increased.
4. The landslide geological disaster risk classification prediction method for mountain railways as recited in claim 1, wherein S4 comprises the steps of:
s4.1, the ELM extreme learning machine model consists of an input layer, an hidden layer and an output layer, wherein an excitation function of the hidden layer is set to be g (omega, X, b), and then the ELM extreme learning machine expression is as follows:
in which y i Representing ELM desired output; beta i The connection weight of the i-th neuron of the hidden layer and the neuron of the output layer is obtained; omega i Representing the connection weight of the input layer neuron to the ith neuron of the hidden layer; b i Is the threshold of the hidden layer neuron, g (omega i *x j +b i ) Is an excitation function; the Hardlim function is adopted as an excitation function of the ELM extreme learning model, and the expression is as follows:
s4.2, randomly dividing an input training set into an A part and a B part, wherein the A part adopts a PSO algorithm to optimize a parameter input weight and an hidden layer deviation value in the ELM extreme learning machine, and the B part adopts a traversal search method to optimize an embedding dimension and a hidden layer node number in the ELM extreme learning machine;
the optimization process of parameter input weight and hidden layer deviation value in the ELM extreme learning machine by adopting PSO algorithm () is as follows:
input weight w in ELM i Hidden layer node bias b i As particles of PSO algorithm, taking root mean square difference of ELM training samples as fitness function, calculating fitness value of each particle, comparing and continuously iterating to update speed and position of the particles until error rate is minimum or maximum iteration times, and finally obtaining ELM parameters after PSO optimization;
the update formula of the particle swarm algorithm is as follows:
wherein ω is a weight, t is the number of iterations, V in For the speed of the particles c 1 、c 2 X is the position of the particle, r 1 、r 2 Is a random number in the range of [0,1 ]]Between them.
5. The landslide geological disaster risk classification prediction method suitable for mountain railway lines according to claim 1, wherein the method comprises the following steps of: in order to improve the positioning accuracy of the landslide occurrence place, the step S7 is to obtain risk prediction results of each area by adopting the following steps:
s7.1, dividing grids of each monitoring area, and respectively collecting real-time monitoring data of sub-grids of each monitoring area;
s7.2, in each monitoring area, monitoring data of each sub-grid are input into a corresponding landslide geological disaster risk prediction model optimized by a PSO algorithm, and landslide prediction risk results of the sub-grids are obtained;
and S7.3, summarizing and calculating landslide prediction risk result data of each regional unit grid to obtain a final risk prediction result of the region.
6. The landslide geological disaster risk classification prediction method suitable for use along mountain railways of claim 5, wherein: and S7.2, calculating and processing the sub-grid data of each region of the data by adopting a multithreading parallel method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211535856.XA CN116090696A (en) | 2022-12-01 | 2022-12-01 | Landslide geological disaster risk classification prediction method suitable for mountain railway line |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211535856.XA CN116090696A (en) | 2022-12-01 | 2022-12-01 | Landslide geological disaster risk classification prediction method suitable for mountain railway line |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116090696A true CN116090696A (en) | 2023-05-09 |
Family
ID=86185818
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211535856.XA Pending CN116090696A (en) | 2022-12-01 | 2022-12-01 | Landslide geological disaster risk classification prediction method suitable for mountain railway line |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116090696A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117037432A (en) * | 2023-10-08 | 2023-11-10 | 四川省公路规划勘察设计研究院有限公司 | Risk evaluation geological disaster early warning method based on multi-method cooperation |
-
2022
- 2022-12-01 CN CN202211535856.XA patent/CN116090696A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117037432A (en) * | 2023-10-08 | 2023-11-10 | 四川省公路规划勘察设计研究院有限公司 | Risk evaluation geological disaster early warning method based on multi-method cooperation |
CN117037432B (en) * | 2023-10-08 | 2023-12-19 | 四川省公路规划勘察设计研究院有限公司 | Risk evaluation geological disaster early warning method based on multi-method cooperation |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wu et al. | Rock burst prediction probability model based on case analysis | |
Li et al. | Automated decision making in highway pavement preventive maintenance based on deep learning | |
Fijani et al. | Optimization of DRASTIC method by supervised committee machine artificial intelligence to assess groundwater vulnerability for Maragheh–Bonab plain aquifer, Iran | |
CN101354757B (en) | Method for predicting dynamic risk and vulnerability under fine dimension | |
CN108009674A (en) | Air PM2.5 concentration prediction methods based on CNN and LSTM fused neural networks | |
CN114971301B (en) | Ecological interference risk identification and evaluation method based on automatic parameter adjustment optimization model | |
CN111639823B (en) | Building cold and heat load prediction method constructed based on feature set | |
CN109886456B (en) | Debris flow prediction method based on PCA and mixed kernel function LSSVR | |
CN117236199B (en) | Method and system for improving water quality and guaranteeing water safety of river and lake in urban water network area | |
CN117009735A (en) | High-strength forest fire occurrence probability calculation method combining BiLSTM and nuclear density estimation | |
CN115730684A (en) | Air quality detection system based on LSTM-CNN model | |
CN117035465B (en) | Method and device for evaluating landslide susceptibility | |
Arabameri et al. | K-fold and state-of-the-art metaheuristic machine learning approaches for groundwater potential modelling | |
Li et al. | Spatiotemporal assessment of landslide susceptibility in Southern Sichuan, China using SA-DBN, PSO-DBN and SSA-DBN models compared with DBN model | |
CN116402951A (en) | Digital twinning-based side slope safety management method and system | |
CN116090696A (en) | Landslide geological disaster risk classification prediction method suitable for mountain railway line | |
CN115948964A (en) | Road flatness prediction method based on GA-BP neural network | |
CN109255485A (en) | Rainfall-triggered geologic hazard early-warning and predicting model and learning method based on RBFN machine learning | |
CN117238126A (en) | Traffic accident risk assessment method under continuous flow road scene | |
Moayedi et al. | A novel problem-solving method by multi-computational optimisation of artificial neural network for modelling and prediction of the flow erosion processes | |
Milan et al. | Using machine learning to determine acceptable levels of groundwater consumption in Iran | |
CN117037432B (en) | Risk evaluation geological disaster early warning method based on multi-method cooperation | |
Moradi et al. | Prediction of groundwater level fluctuation using methods based on machine learning and numerical model | |
CN116110210B (en) | Data-driven landslide hazard auxiliary decision-making method in complex environment | |
CN116933920A (en) | Prediction and early warning method and system for underground mine debris flow |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |