MO-ISCSO: A Challenging Benchmark Test Suite for Large-Scale Multi-Objective Structural Optimization

Abstract

Current studies on the development of multi-objective algorithms for optimization of truss structures mainly depend on small-scale classic benchmark instances. This paper highlights the importance of establishing standard large-scale multi-objective structural optimization benchmarking suites for accurate validation of the proposed algorithms. A new benchmark test suite, called MO-ISCSO, is proposed for large-scale multi-objective structural optimization, based on the most recent optimization problems of the international student competition in structural optimization (ISCSO). Owing to the very small feasibility ratios of the MO-ISCSO instances, the effect of presence of feasible designs in the initial population of NSGA-II, GDE3, and AR-MOEA multi-objective optimization algorithms is investigated using the proposed test suite. The obtained numerical results indicate that seeding the initial population with feasible solutions helps the foregoing algorithms maintain a better balance between convergence and diversity. The statistical results form a baseline for future studies on developing efficient multi-objective structural optimization techniques.