Numerical Simulation And Ground Subsidence Prediction For Multilayer Goafs Underlaying Highway

With the development of the national economy, higher level of transportationsystem is needed. Road network is more and more intensive, and some of the highwayand railway will inevitably need to build across the mined-out area as a result.Deformation of rock mass overlying the goaf may induce surface subsidence,inclination, horizontal movement and deformation, which may cause failure ofroadbed and road surface collapse for highway across mined-out area. This kind ofgeological disaster caused great economic losses in our country.K70+500-K75+000section of Siping-Changchun highway, part ofBeijing-Haerbin highway is located in the northwest of Liufangzi bentonite and coalmining area. There are subsidence, cracks and roadbed subsidence in part of roadbecause of the goaf. The plan to extend the Siping-Changchun highway encountersdifficulty due to large deformation in Liufangzi segment. The maximum subsidencehas reached172.0cm (K73+288).Then whether subsidence will continue and itsinfluence on the highway become the key problem.Deformation of K73+100-K73+500section of Siping-Changchun highway istaken as the the object of study in this paper. On the basis of collecting and analyzinga lot of technical literature and the field data of subsidence, we analyzed geologicalenvironmental conditions, characteristics of surface subsidence, distribution laws andits formation mechanism for the research area.Then prediction system for multiplemining of random shaped face based on matlab is developed and applied to theresearched highway and its neighborhood. Deformation rules of surface and rock masson the goaf under the action of dynamic load are obtained by using the numericalsimulation software named FLAC. Mining subsidence forecast method based onimproved genetic algorithm and holt-winters model is proposed and applied in miningsubsidence forecast of Changchun-Siping highway. This research will providebeneficial references for similar engineering geological problems. The main resultsare as follows: (1) Comprehensive analysis for characteristics of surface subsidence, distributionlaws and its formation mechanism have been undertaken based on the data of fieldinvestigations. As coal drilling expanded in the direction of the highway，and evenexpanded over the highway, the subsidence basin gradually spread to the highway.It isknown that subsidence of a point on the surface reduced with its distance increasedfrom the center of the basin.There is a goaf with a width of190m under the sectionK73+177—K73+367. Above all, the subsidence of the southeast side is greater thanthe northwest side and there is a subsidence center at K73+288with subsidencereducing in both sides along the highway.(2) A prediction system for multiple mining of random shaped face based onmatlab is developed.When it is applied to the points along the researched highway andits neighborhood, there are certain differences between the predicted values and theobservation values. Considering that the highway is always under the influence ofdynamic load, the results got from the prediction system are regarded to beappropriate for the actual matters.Then the problem of prediction for multi-layer goafand arbitrary shape goaf is solved.(3) With the numerical simulation software named FLAC, deformation rules ofsurface and rock mass overlying the goaf under the action of dynamic load areobtained. Results show that the impact of dynamic load on surface and rockmass overthe goaf decreases with the increasing of buried depth, increases with an increase inmining width and decreases with increase of speed.(4) A method based on improved genetic algorithm and Holt-Winters model isproposed. With the aid of the improved genetic algorithm, parameters of Holt-Wintersmodel can be greatly optimized. Then the IGA-Holt-Winters model is applied inmining subsidence forecast of Changchun-Siping highway. The results show theconvergence of methods is faster and more accurate than that of the GM(1,1) modeland WLS-SVM model with relative mistake lower than2%. The model has betterprediction accuracy for long-term prediction of the mining subsidence.