Abstract:

The determination of evolution stage of debris flow gully is the first step of forecasting，evaluation and control of debris flow and the scope and frequency of the debris flow.Utilizing artificial threeply intelligenceBP neural network model,then selecting catchment area,main groove length,groove gradient ratio,average aspect,relative height difference,round ratio and relative cutting degree of the catchment evolution geomorphology as assessment index of the evolution stage of debris flow gully,the stage of debris flow gully was divided into four stages:young stage、developing stage、active period and decline stage.Authors pretreated the 80 debris flow data along Chengkun railway in Sichuan province (systematic classification、standardization) in order to avoid artificial error,secondly,network training the 80% of the data,and then built forecasting model,the simulation of the residual 20% of the data shows that the average relative error is 8.22% with satisfactory result.The evolution stage of the 6 debris flow gullies along Kundong railway was determined based on the model,the result showed that:the 6 debris flow gullies are in active period and the evolution score is between 3～3.5,so the monitor and forecast should be strengthened in these six debris flow gullies in order to avoid disasters.The intelligenceBP neural network model can be used as an advantageous method to determinate the evolution stage of debris flow gully.The result can provide theoretical basis and technical support for debris flow forecasting，evaluation and control.

Key words: evolution stage of debris flow/BP neural network/forecasting/Kundong railway