基于集成神经网络的球墨铸铁珠光体含量涡流无损智能测定

为了实现球墨铸铁珠光体含量的智能无损测定，制备了具有代表性的球墨铸铁试样并按珠光体粗细程度将其分为三类。首先在分析了影响球墨铸铁电磁性能的主要因素的基础上，同时采用涡流无损检测法与金相法对球墨铸铁珠光体含量进行了测定，对检测数据进行回归分析表明二种方法的测量结果很接近；然后采用集成神经网络处理涡流检测数据并对珠光体含量进行了预测，预测结果表明基于集成神经网络数据处理的涡流检测是一种快速智能识别球墨铸铁中珠光体含量的有效方法。

Application of Integrated Neural Networks to Determine the Pearlite Content of Spheroidal Graphite Iron by Eddy Current Nondestructive testing

To determine the pearlite content of the spheroidal graphite iron automatically, The representative spheroidal graphite iron samples are made and heat treated in different way and then classified into three categories according to their thickness of pearlite. The factors that influence the electromagnetic property of spheroidal graphite iron are analysed and eddy current testing (ECT) is introduced. Experimental data are analyzed using regression analysis method and indicate that there are almost the same pearlite content measured by ECT and metallograph method separately. The integrated neural networks are used to process the data collected in ECT and to predict. The prediction results show that the integrated neural networks are effective means to predict the pearlite content of the spheroidal graphite iron in ECT.