Displacement Forecast And Research Based On ARIMA Model And PSO-BP Neural Network

Increasing damage has been caused by landslide,and loss of life and property that is caused by landslide will decrease effectively if accuracy of landslide forecast has improved.So that,research of landslide prediction is obviously practical,and displacement forecast is the important part of forecast.This paper analyzes displacement data and deformation mechanism qualitatively.Conducting quantitative research of external factor causes landslide movement.Setting up forecast model based on ARIMA theory and PSO-BPNN algorithm.Applying separate and holistic method to forecast displacement of ZG93 and XD01.Comparing cumulative displacement data to pick up best forecast method.And using best method to forecast displacement data.This paper contains following research results:1.Concluding relation between landslide movement,reservoir water level and rainfallBaishuihe landslide has stepped movement characteristics: speed of displacement increases from April to September,and cumulative displacement data increases remarkably;but from October to March,landslide won't move noticeably,and cumulative displacement data is stable.From April to September,in general the movement increases first and decreases then.Rainfall and changes of reservoir level causes bad influence on landslide stability,but rise of reservoir level will increase stability.Influence on movement caused by reservoir water drop and rainfall has hysteresis.2.Analyzing qualitative influence on Baishuihe landslide movement made by internal and external impact factorBaishuihe landslide has height difference and enough space to move;structure of below strata alternate with hard and soft rock,and strength of rock is not high;joint surfaces offer opportunity for rainfall.Heavy rain will increase unit weight of soil,then decrease friction angle and cohesion;drop of reservoir level will increase porosity and permeability of soil,then lower rock compressive strength;uplift force,seepage pressure and pore water pressure which are caused by change of reservoir level and rainfall will effect stability of landslide directly.Under influence of external factor,slip soil strength will become residual sheer strength from recovered sheer strength,landslide will move fast;under effect of gravity,residual sheer strength becomes to recovered sheer strength,and movement will slow down.3.Researching quantitative correlation between out factor data and deformationAccording to Pearson correlation and canonical correlation analysis,research proves that rainfall in this month,rainfall of last month,rainfall of the month before last,reservoir water level of this month and reservoir level change of last month are major factor of landslide movement.4.Comparing landslide displacement prediction calculated by PSO-BPNN algorithm and ARIMA model to decide the best methodResults show that accuracy of separate method is worse than holistic method because trend displacement prediction data has large error.PSO-BPNN uses displacement impact factor as input data,and ARIMA model only uses displacement to make forecast,so PSOBPNN is better than ARIMA model.Comparing different predict method,holistic method and PSO-BPNN are best method.5.Forecasting landslide displacement through PSO-BPNN and holistic methodAccording to displacement prediction,movement barely happens from October 2017 to March 2018,and maximum movement will happen under situation of heavy rain and reservoir level fluctuation from April 2018 to September 2018.