Abstract
A combined multi-point approximate function, which integrates the formulas of original and modified multi-point approximation, is proposed in the paper. It is then applied to solve truss structural optimization problems with continuously cross-sectional size variables and discrete topology variables. With the two level approximation concepts and genetic algorithm (GA), the primal truss optimization problem is at first transformed into a series of approximately mixed-variable problems using the proposed approximate formula; after that the topology variables and cross-sectional areas of the truss are determined by GA and dual method respectively through a layered strategy. The final optimization solution can be reached after iterations. The numerical examples show the truss optimization with topology variables can be obtained after less structural analysis. Which presents that the proposed combined multipoint approximation has higher quality that the former modified multi-point approximation.