Improved multi-classes SVM and application in ring-key position selection in tunnels

The universal trapezoidal tapered ring has been widespread applied in the tunnel engineering. However, it is difficult to select rational ring-key position location. As it can be defined as a restrictive multi-classes problem, traditional method cannot be used directly. This paper proposed a novel method, CBPSO-RMCSVM, which combines chaos binary particle swarm optimization (CBPSO) with restrictive multi-classes support vector machine (RMCSVM), so as to satisfied actual engineering demand and improve the accuracy and efficiency. With the guidance of CBPSO-RMCSVM, a new decision support system of universal trapezoidal tapered ring-key position selection in shield tunneling was designed and developed. The system was applied to the Shanghai Yangtze River tunnel project successfully, and the engineering result showed the improved method had good performance.