本文提出一种粗糙集理论和动态前馈神经网络相结合的神经网络构造方法。充分发挥了粗糙集理论和神经网络的优势，弥补了各自的缺点。并应用于实际工业过程，在乙烯装置裂解炉燃料气热值控制中取得了良好的应用效果。-This paper presents a rough set theory and dynamic feedforward neural networks combined neural network constructed. Give full play to the rough set theory and neural networks the advantage to make up for their shortcomings. And applied to practical industrial processes, in the ethylene plant cracking furnace fuel gas calorific value of control to achieve a good application effect.