论文部分内容阅读
建立了血管支架变形影响因子与其变形结果之间的具有高度非线性识别能力的神经网络模型,通过引入学习因子η和动量因子ψ,采用附加动量项的权值修正方法,优化了网络训练算法,从而提高了网络训练速度和系统鲁棒性.结合实例对网络进行训练,并对预测误差进行了统计假设检验,检验结果表明血管支架变形神经网络智能预测结果与非线性有限元分析结果误差均值低于0.03%,训练后的网络能够较好地对血管支架变形进行预测.在此基础上,基于Pro/Toolkit工具,融合血管支架扩张变形神经网络智能预测模型,建立了血管支架力学性能快速评价工具,该系统实用性强、效率高,能大幅缩短血管支架产品开发周期。
A neural network model with highly nonlinear recognition ability between vascular stent deformation factor and its deformation result is established. By introducing learning factor η and momentum factor ψ, a weight correction method with additional momentum is adopted to optimize the network training algorithm. Which improves the training speed and system robustness of the network.The examples are used to train the network and make statistical hypothesis test on the prediction error.The test results show that the error between the intelligent prediction results of the vascular stent deformation neural network and the nonlinear finite element analysis results is low At 0.03%, the trained network can predict the deformation of vascular stent better.On this basis, based on the Pro / Toolkit tool, the fusion neural network intelligent prediction model of vascular stent dilatation and deformation, a rapid evaluation tool of vascular stent mechanical properties was established The system is practical and efficient, which can greatly shorten the product development cycle of vascular stent.