Grouting Stratum Classification with Support Vector Machines

A support vector machines (SVM) approach is proposed for grout stratum identification based on the grout instrument measurement datum in the grouting consolidation project. The important features of SVM model are obtained by analysis of grout fluid penetration mechanism integrating with grout physical model for pressure control aim. Then we introduce Lagrange multiplier and translate the SVM math model into an unconstrained objective function. The optimal hyper-planes is produced by quadratic programming technics and RBF kernel. In the real grouting project, we draw the osmosis-flow character curves in the different stratum by test , which are major grout parameters changing with stratum. And we obtain the primal datum of SVM classification. At last ,we separately choose 49 group datum for fracture vein rock and gritstone stratum. A part of them is used to training set of SVM ,the other is used to check up the classification effect. Simulations demonstrate identification error is less than 6.13%, so the method can applied to the automatic grout project.