人工神经网络快速预测蠕墨铸铁的性能

Fast prediction of vermicular graphite cast iron property based on Back Propagation neutral network

蠕墨铸铁具有优异的力学性能、铸造性能、抗热疲劳和耐磨性能,是大功率柴油发动机缸体的理想合金材料.蠕墨铸铁容易受到生产工艺因素的影响,蠕化率难以控制,必须依靠炉前快速分析技术加以严格控制.以蠕墨铸铁的化学成分和力学性能为研究对象,在大量实验数据的基础上,采用Matlab软件中的误差反向传播算法神经网络工具箱,通过二次开发建立了一个基于热分析的预测网络,实现蠕墨铸铁性能的炉前化学成分和力学性能的快速预测,并与理化实测数据进行分析对比.结果表明,人工神经网络能充分逼近复杂的非线性系统,准确快速地预测蠕墨铸铁的化学成分和力学性能,有利于蠕墨铸铁的蠕化率、化学成分和力学性能的炉前快速监控,确保蠕化处理效果的稳定性,提高产品质量,降低铸铁的生产成本.

Vermicular Graphite Cast Iron is the ideal alloy for high-power diesel engine block for its excellent mechanical property,casting performances,thermal fatigue resistance and great wearability.The chemical composition and mechanical properties of Vermicular Graphite Cast Iron were researched.The secondary development which is involved in Back Propagation neural network toolbox of Matlab software was used to set up a thermal analysis predicting network based on experimental data.The rapid prediction of chemical composition and mechanical properties of Vermicular Graphite Cast Iron was achieved through the secondary development.Compared with physical measurement data,the results show that chemical composition and mechanical properties of Vermicular Graphite Cast Iron is quickly and accurately predicted by Back Propagation neural network which is approximate to the complex non-linear system.So vermicular iron vermicularity,chemical composition and properties of Vermicular Graphite Cast Iron can be rapidly monitored to ensure the stability of the vermicularizing treatment,to improve the product quality and to reduce the cost of production.