This paper describes the basic ideas of sequential quadratic programming (SQP) algorithms for design optimization. There are two fundamental differences between the various algorithms: (i) the definition of the QP subproblem solved at each iteration, and (ii) the descent function used during step size determination. The performances of the algorithms can change dramatically depending on how the two steps are executed. Numerical implementation details of various computational steps are discussed.
Three programs based on SQP algorithms are used to solve 17 structural design problems having 7 to 96 design variables and 10 to 1051 performance constraints besides design variable bounds. Based on the performance of these programs, efficient procedures to execute various steps of the SQP methods are determined. It is concluded that the potential constraint strategy, where only a subset of the constraints is used to define the QP subproblem, is essential for large scale engineering design applications. With this strategy the SQP methods are quite robust and have great potential for routine application in engineering design.