灰铸铁石墨形态的神经网络识别研究

灰铸铁石墨形态的神经网络识别研究 Recognizing Graphite Morphology of Gray Cast Iron by ANN 江虹 JIANG Hong 1 first-author 曾立波 ZENG Li bo 1 张泽兰 ZHANG Ze lan 1 胡继明 HU Ji ming 1 武汉大学分析测试科学系,武汉,430072 武汉大学分析测试科学系,武汉,430072 Department of Analysis Measurement Science, Wuhan University Department of Analysis Measurement Science, Wuhan University 在分析分形维、纹理粗细度及二维回归系数共同组成的纹理特征基础上 ,本文运用人工神经网络方法实现了对灰铸铁石墨形态的识别 .相对于传统方法 ,该方法识别速度快、重现性强、数据更为客观准确 ,避免了由于主观判断所引起的人为误差 .在识别过程中 ,本文使用了一种确定隐层节点数的新方法 ,结果显示良好 ,对于人工神经网络中隐层节点数确定难的问题具有很好的实用参考价值 On the base of feature vector composed of fractal parameter, roughness parameter and regression, we complete the classification of the gray cast iron’s graphite morphology by ANN. compared with the traditional method, it works much quicker, and is with better repeatability, more objective and veracious results. Furthermore, we apply a new method to decide the hidden node’s number of ANN, which shows valuable to such a difficulty in ANN structure decision. Result shows available. 灰铸铁 石墨形态 分形维 粗细度 回归系数 BP网络 gray cast iron graphite morphology fractal roughness regression BP network 10.14188/j.1671-8836.2000.03.034 TG115.21 国家科技部科研基金资助项目!（JG-99-9） 2000-03-01 2021-04-01 3