Parametric modeling of a microaccelerometer: comparing I- and D-optimal design of experiments for finite-element analysis

Parametric modeling of a microaccelerometer is used to compare two optimization criteria in the design of experiments for finite-element analysis (FEA). I- and D-optimality criteria are used to specify a near-minimal set of 38 input points in multidimensional input space at which FEA is performed. The accelerometer is comprised of a rigid proof mass and four suspension beams. Its response is determined as a function of acceleration along the input and cross axes, temperature, and two structural variables, and the related performance compromises are identified. It is shown that both I- and D-optimality criteria provide good results, with displacement residuals spread over ranges of magnitude 1.6 and 2.3 /spl mu/m, respectively, in a large input domain over which the range of displacement has a magnitude exceeding 13 /spl mu/m. Additionally, the results suggest that for the particular device that was considered, the temperature coefficients of offset and sensitivity have smaller magnitudes when the suspension beams are attached near the center of each side of the proof mass.