Gross Error Discrimination of Dynamic Measurement Data Based on Fuzzy Clustering

摘要

Describing gross error with Fuzzy Mathematics' membership degree, this paper proposed a fuzzy description method of dynamic measurement error data; combining with clustering analysis, the gross error-distinguishing model of fuzzy clustering is built. The partition of error characteristic class is completed on the base of analyzing the fuzzy relation, the transformation of fuzzy relation and the fuzzy equivalence relation. The gross error-classify method based on fuzzy quantity C-means clustering proposed consist of the Membership function and the fuzzy relation between fuzzy clusters, the fuzzy clustering sets and the quantity evaluation function, the fuzzy variance of fuzzy class and the fuzzy variance between classes, the object function, and the Convergence conditions of algorithm. The logic structure of the algorithm is clear and the algorithm needs less operation time. The algorithm has certain self- adjust function since the data characteristic analysis is introduced into the algorithm. In the dynamic circular division measurement system, the use of the gross error-classify method enables the dynamic measurement accuracy to improve by 80%, and to reach 7 arc-seconds.