Premium
Genetic algorithm‐based ground motion selection method matching target distribution of generalized conditional intensity measures
Author(s) -
Ji Kun,
Wen Ruizhi,
Zong Chengcai,
Ren Yefei
Publication year - 2021
Publication title -
earthquake engineering and structural dynamics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.218
H-Index - 127
eISSN - 1096-9845
pISSN - 0098-8847
DOI - 10.1002/eqe.3408
Subject(s) - algorithm , crossover , selection (genetic algorithm) , tournament selection , genetic algorithm , matching (statistics) , standard deviation , conditional probability distribution , mathematics , fitness proportionate selection , set (abstract data type) , intensity (physics) , pattern recognition (psychology) , computer science , amplitude , fitness function , artificial intelligence , statistics , mathematical optimization , physics , quantum mechanics , programming language
This study developed an approach for selecting sets of ground motion recordings that match a target conditional multivariate distribution of ground motion intensity measures (IMs). This was achieved by applying a genetic algorithm (GA) that treats IMs of interest of each recording as a “chromosome” and the set of the desired number of recordings as a single “individual.” The fitness function was constructed by measuring the mismatch between the target and the individual's means and variances for all IMs. Then, through Roulette wheel natural parent selection, one‐point chromosome crossover, and individual mutation, new generations of ground motion sets were produced and the process was continued until the optimum combination of recordings was obtained. Example application illustrated that the proposed GA method could efficiently search and find a desired number of recordings to represent the target conditional IMs’ distribution, including the mean and variance. The IMs considered included response spectrum (range: 0.05‐10.0 s), amplitude/intensity‐based IMs, cumulative‐based IMs, and duration. Comparison with existing GCIM selection method indicated that the standard deviation of the recordings selected using the proposed GA method was closer to the target and more stable among replications. The results demonstrated that the proposed GA method represents a promising approach for searching pairs of recordings that could simultaneously match the target conditional distribution of various IMs.