Abstract
The desirability function (DF) is widely used in multiobjective optimization problems especially when individual design objectives have conflicting responses. DF successfully transforms a complicated problem into a classical single parameter optimization problem by combining contradicting constraints. It can be applied to complex high-dimensional and multimodal problems where classical iterative methods are often insufficient to provide practical solutions. DF, however, requires prior knowledge about the trade-off parameter ranges and the preference parameters, namely, the target bound and shape. Thus, expert knowledge before optimization is often critical for convergence. In this work, multiobjective optimization problem is illustrated on thinned array antenna design. Single and two objective optimizations are compared with the all-active (no thinning) array. Adaptive desired function (ADF) is proposed, which requires no prior knowledge about the trade-off objectives. ADF is observed to provide faster convergence and better optimization results/behavior with its “fire, monitor, and update” approach when compared with the classical DF’s “fire, forget, and if it does not converge then restart” approach.