Fuzzy Rule Base Influence on Genetic-Fuzzy Reconstruction of CMM 3D Triggering Probe Error characteristics

One of the most important sources of coordinate measuring machine (CMM) errors is the probe used to collect coordinate points on measured objects. The error value depends on the probing direction; hence its spatial variation is a key part of the probe inaccuracy. This paper presents genetically-generated fuzzy knowledge bases (FKBs) to model the spatial error characteristics of a CMM touch trigger probe. The automatically generated FKBs are used for the reconstruction of the direction-dependent probe error w. The angles beta and gamma are used as input variables of the FKBs; they describe the spatial direction of probe triggering. The learning algorithm used to generate the FKBs is a real/binary like coded genetic algorithm developed by the authors. The influence of the number of fuzzy rules (FR) on the precision of the genetically-generated FKBs is investigated by varying the number of fuzzy sets (FS) on the premises and on the conclusion. The results of the learning are examined. Once the adequate number of fuzzy rules is found, an optimal learning is performed and a near-optimal FKB of probe error characteristics is proposed