【摘要】 针对激光超声表面波检测脊髓表面应力中，声弹性效应拟合方法预测脊髓表面应力准确性不高的问题，提出一种基于差异性优选特征并训练BP神经网络的脊髓表面应力的声表面波(SAWs)定量表征技术。通过运用有限元法模拟激光激发SAWs过程，提取脊髓表面应力变化引起的SAWs峰值、平均值、均方根、传播时间差等多个特征训练BP神经网络，建立起有效的神经网络预测系统，实现对脊髓表面应力0～18000Pa的定量表征。模拟结果表明，BP神经网络系统预测脊髓表面应力相对误差在5%以内，与声弹性效应拟合曲线预测结果相比，准确率提高60%以上，验证了BP神经网络定量表征脊髓表面应力的有效性与准确性。更多还原

【Abstract】 Aiming at the low accuracy of spinal cord surface stress prediction in the acoustic elastic effect fitting method on laser ultrasound, based on divergence feature optimization, as well as training BP neural network, a surface acoustic waves(SAWs) quantitative characterization technique of the spinal cord surface stress is proposed. By using the finite element method to simulate the process of laser stimulated saws, the BP neural network was trained by extracting the peak value, average value, root mean square, propagation time difference and other characteristics of saws caused by the change of the surface stress of the spinal cord, and an effective neural network prediction system was established to realize the quantitative characterization of the surface stress of the spinal cord from 0 to 18000 pa. The simulation results show that the relative error of BP neural network system in predicting spinal cord surface stress is within 5%,and the accuracy is increased by more than 60% compared with the prediction results of acoustic elasticity effect fitting method, which verifies the effectiveness and accuracy of BP neural network in quantitative characterization of spinal cord surface stress.更多还原