Gyroscopic Drift Forecasting Based on a New Adaptive Fuzzy C-Means Clustering Algorithm

In order to avoid the blindness and randomness in extracting fuzzy rules , an approach for constructing T-S fuzzy models was proposed on the basis of a new Adaptive Fuzzy C-means ( AFCM ) Clustering Algorithm. Firstly , subtractive clustering was utilized to determine the upper limit of clustering number and the initial clustering centers. Then an improved Fuzzy C-means clustering algorithm was adopted to optimize the clustering centers. Finally , the number of fuzzy rules and the clustering centers were confirmed adaptively through clustering validity method. On the other hand , the improved Fuzzy C-means clustering algorithm could eliminate the influence of noise on the clustering result. In the constructed model , the unknown parameters in consequent terms were identified by the weighted least square method. A gyroscope's drift data was used to demonstrate the detailed implementation of the proposed method , for which the number of fuzzy rules was confirmed adaptively and the accuracy of the prediction result was high comparatively. The results show the effectiveness and feasibility of the algorithm.