基于最小二乘拟合的模糊隶属函数构建方法

Construction of fuzzy membership functions based on least squares fitting

针对当前模糊隶属函数构造方法中存在的问题,提出一种构造模糊隶属函数方法.采用最小二乘法拟合离散数据来获得隶属函数.为减小拟合误差,采用了3项措施以达到预期目标.所构建的隶属函数,对任意输入物理量可直接得到其对应模糊语言变量的隶属度,从而有效避免专家指定隶属度的主观臆断性及不一致性.该方法简单、求解精度高,具有广泛适用性和较强的应用价值.仿真结果证实了该方法的有效性.

This paper proposes a approach for constructing fuzzy membership functions considering some problems existing in some current methods. The approach employs least squares to fit discrete data, and their membership functions are obtained. To decrease fitting errors, three measures are adopted. For any input datum, its corresponding membership degree can be obtained directly by the constructed membership function. Subjectivity and unconsistency generated by experts can be avoided. This approach has simplicity, high accuracy and wide applicability. The simulation results show the effectiveness of this approach.