| 神经元网络及其参数训练用于钢丝绳无损检测 |
| |
| 引用本文： | 高红兵,杨克冲,杨叔子. 神经元网络及其参数训练用于钢丝绳无损检测[J]. 华中科技大学学报(自然科学版), 1992, 0(Z1) |
| |
| 作者姓名： | 高红兵 杨克冲 杨叔子 |
| |
| 作者单位： | 华中理工大学机械工程一系(高红兵,杨克冲),华中理工大学机械工程一系(杨叔子) |
| |
| 基金项目： | 国家自然科学基金资助项目 |
| |
| 摘 要： | 针对现阶段钢丝绳无损检测中存在的门限值与概率模型中特征参数的选择问题,提出运用神经元网络技术解决该问题的一个方案.并研究了神经元网络训练参数集的大小对门限值以及概率模型中特征参数准确性的影响,还给出了实验结果.
|
| 关 键 词： | 神经元网络 钢丝绳 断丝 |
On the Neural Network Method and the Size of the "Training" Parameter Set in Neural Networks for Steel Wire Rope Non-Destructive Inspection |
| |
| Gao Hongbing Yang Kechong Yang Shuzi. On the Neural Network Method and the Size of the "Training" Parameter Set in Neural Networks for Steel Wire Rope Non-Destructive Inspection[J]. JOURNAL OF HUAZHONG UNIVERSITY OF SCIENCE AND TECHNOLOGY.NATURE SCIENCE, 1992, 0(Z1) |
| |
| Authors: | Gao Hongbing Yang Kechong Yang Shuzi |
| |
| Affiliation: | Gao Hongbing Yang Kechong Yang Shuzi |
| |
| Abstract: | The basic concepts and specific features of neural networks are presented. The method of determining the threshold and characteristic parameters of the probability model in a wire rope non-destructive inspection by modified Adaline neural networks is suggested. The influence of the size of the "training" parameter set in neural networks on the accuracy of the threshold and characteristic parameters of the probability model is studied. Neural networks have been trained at lengths of 50,40,30,20,10,5,3 meters and 1 meter respectively on the tested wire rope. In the course of "training" , the threshold and characteristic parameters of the probability model are modified. An investigation on the numbers of misjudgment made show that the "training" of neural networks at a length of 1 meter is the best choice for practical applications. |
| |
| Keywords: | neural networks wire rope broken wire |
| 本文献已被 CNKI 等数据库收录！ |
