Rapid prediction method for nonlinear expansion process of medical vascular stent

Abstract

A neural network model with high nonlinear recognition capability was constructed to describe the relationship between the deformation impact factors and the deformation results of vascular stent. Then, using the weighted correction method with the attached momentum term, the network training algorithm was optimized by introducing learning factor η and momentum factor ψ, so the speed of the network training and the system robustness were enhanced. The network was trained by some practical cases, and the statistical hypothesis validation was made for the predictive errors. It was shown that the average difference between the intelligent predictive result of vascular stent deformation neural network and the nonlinear finite element analysis result was less than 0.03%, and the trained network could perfectly predict the vascular stent deformation. Further more, the rapid evaluation tool for the vascular stent mechanics performance was established using the Pro/Toolkit and the intelligent neural network predictive model of vascular stent expansion. The proposed tool system with strong practicality and high efficiency can significantly shorten the product development cycle of vascular stent. © 2009 Science in China Press and Springer-Verlag GmbH.