模糊极小极大神经网络在火箭发动机故障分类中的应用

FUZZY MIN-MAX NEURAL NETWORKS AND ITS APPLICATION TO LIQUID ROCKET ENGINE FAILURE CLASSIFICATION

研究了一种用模糊集表示火箭发动机故障模式的神经网络分类器。模糊集是由模糊超立方体聚集形成的集合体，模糊超立方体是一个极小点和极大点构成的用隶属函数表示的ｎ维方盒。极小点和极大点的确定用包含有扩张与收缩阶段的模糊极小极大学习算法实现，这种算法能在一次循环学习中形成非线性模式边界，无需对已知故障模式重新训练就可融合新样本和精炼已存在的故障模式。模糊集用于故障模式分类自然地提供了故障更高水平分类的有用信息。液体火箭发动机故障分类的数值仿真解释了模糊极小极大神经网络的优越性能。

A neural network classifier that utilizes fuzzy sets as failure classes of a liquid propellant rocket engine is studied. Each fuzzy set is an aggregate of fuzzy set hyperboxes. A fuzzy set hyperbox is an n-dimensional box defined by a min point and max point with a corresponding membership function. The min-max points are determined using the fuzzy min-max learning algorithm, an expansion-contraction process that can learn nonlinear failure class boundaries in a single pass through the data and provides the ability to incorporate new and refine existing failure classes without retraining. The use of a fuzzy set approaching to failure classification inherently provides degree of membership information that is especially useful in higher level classification. The simulation of two examples on failure classification of a liquid propellant rocket engine demonstrates strong qualitics of the fuzzy min-max neural network.