Set Information Entropy Research Articles

Geographic division of exposure environment for concrete structure based on fuzzy clustering‐rough set information entropy

AbstractDeterioration of concrete structures is closely related to their service environment. Under different environmental conditions, the durability performance of structures will be different. To perform durability design based on local conditions and extend the service life of structures, it is necessary to divide their service environment based on the prevailing deterioration mechanism. In this study, considering Guangdong Province as an example, the environmental conditions affecting the durability of concrete structures are investigated and the geographical distribution of environmental conditions is analyzed, including ambient temperature, precipitation, corrosive gas, acid rain, and salt fog. The weights of the regional durability indexes are calculated by combining fuzzy clustering analysis with information entropy using rough set theory. Based on the assessment data of 1324 existing concrete structures, the durability geographic division map of Guangdong Province, including three regions (Chloride corrosion region, Acid rain‐carbonation coupling region, and Carbonation region), is constructed. This geographic division map is divided into a total of eight subregions by considering the subordinate factors in each region. The environmental characteristics of each region and its dominating deterioration mechanism for concrete structures are analyzed, and suggestions for durability design are provided.

Ant colony feature selection based on fuzzy rough set information entropy

Most heuristic feature selection algorithms converge easily to local-best, which cannot search the whole feature space effectively. In order to improve the parallel search ability to feature space, the information entropy theory of fuzzy rough set was introduced to ant colony model, and the ant search strategy, pheromone updating and state transition rules of the model have been modified to realize ant colony model based feature selection. UCI datasets experiments indicate that the proposed algorithm is effective to feature subset selection compared with three classical feature selection algorithms.