Performance-driven measurement system design for structural identification

Much progress has been achieved in the field of structural identification due to a better understanding of uncertainties, improvement in sensor technologies and cost reductions. However, data interpretation remains a bottleneck. Too often, too much data is acquired, thus hindering interpretation. In this paper, a methodology is described that explicitly indicates when instrumentation can decreases the ability to interpret data. The approach includes uncertainties along with dependencies that may affect model predictions. Two performance indices are used to optimize measurement system designs: monitoring costs and expected identification performance. A case-study shows that the approach is able to justify a reduction in monitoring costs of 50% compared with an initial measurement configuration.