Research On Analysis And Prediction Model Of Rainstorm Debris Flow Based On Computational Intelligence And GIS

Debris flow is a common natural disaster in mountain area, and it has a strongdestructive, direct threat to people’s life and property security, serious impact on thesustainable development of economy. In this paper, based on GIS spatial analysistechnology and computing intelligence theory, models were constructed to analyzedebris flow disaster risk evaluation and provide technical support for debris flowforecasting.It has important theoretical significance and practical application value.Debris flow disaster system belongs to complex nonlinear system.And there existfuzziness and uncertainty in the system. According to the characteristics of differentstages in landslide inoculation and development process, multivariate factors wereintegrated to establish debris flow disasters risk assessing and predicting model, basedon computing intelligent theory and GIS technology. Observation data in the studyarea was simulated to verify the validity of the model. In this paper, the research workand main conclusions are as follows:(1) Based on event tree theory to build the development process of debris flowdisaster event tree model, using fuzzy language to express the probability of event treenode, and evaluate the fuzzy probability of occurrence of debris flow disasters, thedebris flow disaster risk probability is obtained by defuzzy method. Event tree modelembodies the mudslide subsquently phases of the process, and the calculatedprobability of debris flow disasters was in accordance with the actual.(2) the slope, relative elevation difference, the vegetation coverage, loosematerial reserves along the ditch,5d-accumulated rainfall, the maximum hours ofrainfall intensity and the day rain as the assessment indexes of debris flow disasterwarning model, by using the theory of extenics to establish debris flow warningmodel, advanced the theory of formal method for debris flow disaster evaluation.According to the characteristics of the factor set corresponding correlation functionsto calculate the correlation degree, the evaluation process more scientific andreasonable. (3) Using the simulated annealing genetic algorithm to improve GMDH（GroupMethod of Data Handling） network model, and use the improved GMDH model toforecast debris flow disasters. With KLDA discriminant analysis method to selectlager factor correlation and it as an input parameter, the biggest quantity debris flowrushing out once a time as the output parameter, using the improved GMDH networkmodel for debris flow disaster prediction, the model precision is higher than othermodels, such as BP and ANFIS).(4)A method is proposed based on local search strategy of hybrid ant colonyoptimization algorithm to optimize the bayesian network structure learning, and toimprove the bayesian network model, and the improved model was used for debrisflow disaster risk assessment, and the calculated risk of debris flow hazards was inaccordance with the actual situation. The method provided a new technology foranalyzing the uncertainty problem and the incomplete data problem.