论文部分内容阅读
针对时差定位法受很多因素影响的弊端,将神经网络技术应用到声发射源定位中。提取最能揭示声发射源的特征参数和运用主元分析技术来降低输入样本的数量;采用增加隐含层神经元个数探讨它们的误差变化来确定隐含层;运用附加动量法和优化选取初始阈值等措施进行网络设计。将设计好的网络运用到实例中,通过与实际缺陷位置的比较,结果表明,选择合理的网络结构和输入参数可准确定出结构损伤位置,且精度有较大的提高,计算更简单有效。
Aiming at the shortcomings of the time difference positioning method affected by many factors, the neural network technology is applied to the location of acoustic emission sources. Extract the most revealing characteristic parameters of acoustic emission sources and use principal component analysis to reduce the number of input samples; determine the hidden layer by increasing the number of hidden layer neurons to explore their error changes; use additional momentum method and optimize selection Initial threshold and other measures for network design. The designed network is applied to the example. Compared with the actual defect location, the results show that the location of structural damage can be determined accurately by selecting reasonable network structure and input parameters, and the accuracy is greatly improved and the calculation is simple and effective.