Evolutionary Algorithms (EAs)

Abstract

Practical reliability optimization problems can be, in general, difficult to solve via standard mathematical programming methods, e.g. linear programming and integer programming because the objective and constraints are mostly non-linear, and the decision variables can mix continuous variables (e.g. test interval) and integer variables (e.g. redundancy). Evolutionary algorithms (EAs), inspired by natural principles of evolution, can perform population-based stochastic search to produce good solutions (not necessarily globally optimal) in polynomial time. Evolutionary search is the foundation of EAs. A genetic algorithm (GA) is perhaps the most popular and successful EA. It imitates the biological evolution and natural selection processes on a group of individuals (solutions), to eventually achieve quality solutions. The chapter briefly introduces two EAs: differential evolution (DE) and particle swarm optimization, which are as successful as GAs in solving various problems. DE was originally proposed as a population-based global optimization algorithm for real-valued problems.