SMART DIAGNOSIS THE STRUCTURAL DAMAGES OF BUILDINGS: FUZZY-GENETIC APPROACH
Serhiy Shtovba, О. Д. Панкевич · 2004
Summary. A hybrid fuzzy logic- and genetic algorithms- based approach for smart diagnosis the structural damages of buildings is proposed. The approach is illustrated by a fuzzy expert system, finding the cause of stone construction of buildings. The proposed fuzzy-genetic approach seems to be prospective to creation of decision making support systems for detecting and diagnosis of damages in various mechanical and building constructions. PRINCEPLES OF FUZZY-GENETIC DIAGNOSIS THE STRUCTURAL DAMAGES OF BUILDING Diagnosis (or determination of cause) of structural damages of building is an important task in civil engineering. Instant and correct diagnosis of the damages makes further investigations, design, and reconstruction of buildings successful. Our approach to smart diagnosis of structural damages is hybrid. The one is based on such soft-computing techniques as: fuzzy logic (1) and genetic algorithms (2). This allows to combine the advantages of linguistic expert diagnostic knowledge with power of genetic algorithms in searching the optima. The similar approach is widely used in medicine - a lot of fuzzy-genetic diagnostic systems were created for various medical tasks (3). The proposed approach is constituted on the following principles: - description of the diagnostic model structure by hierarchical tree of fuzzy logical inference. Such hierarchical knowledge organisation allows to reduce the diagnostic model complexity: a large rule base with many inputs is changing into several small chained rule bases with fewer inputs; - presentation of parameters in linguistic variable form. According to this principle, linguistic terms assess the parameter values. For example, parameter crack may be represented as a linguistic variable with term-set {Vertical, Oblique, Horizontal}. - formalisation of linguistic terms by fuzzy sets. The formalisation is carried out via parametrical membership functions with bell-, triangular-, or trapezoidal- shapes. - formalisation of expert nature language judgements about relationship «state parameters - diagnosis» by fuzzy knowledge bases. Each expert judgement is represented as a fuzzy if-then rule in the following form: If antecedent proposition, then consequent proposition. The fuzzy proposition is statement likes x is oblique, where oblique is a fuzzy set. - tuning the parameters of fuzzy knowledge bases by genetic algorithms. The tuning is searching the weights of fuzzy if-then rules and parameters of the membership functions, that minimise the difference between actual and inferred decisions. It is supposed that a set of correct experimental data state parameters - cause of structural damage is available. We propose to employ a genetic algorithm, which finds next to global optimal solution quickly. Principal distinction of the genetic algorithm from classical optimisation methods is in the fact that it does not use the notion of a gradient while choosing search direction and it is based on crossover, mutation and selection operations. FUZZY DIAGNOSTICAL SYSTEM OF STONE BUILDING STRUCTURAL CRACKS