This thesis is concerned with the optimum design of sensor arrays (i.e., electronic noses or tongues) using simulation experiments. The proposed design method adopts a set of new criteria, which more adequately represent the desirable properties of a sensor array in practice. A number of best non-dominated array designs are selected through a multiple objective (criteria) Tabu search algorithm. The evaluation of a candidate sensor array is based on its multivariate calibration model, which is efficiently estimated from well-designed simulation experiments. The method can be used to optimize the design of both linear and nonlinear sensor arrays. |