engleski

Sensor placement optimization using convex L0 norm relaxations

Sensor placement optimization problem is a fundamental one in vibration measurement, structural health monitoring, and so on. Many different approaches have been proposed [1], from heuristic to approximations of the original problem [2]. We present an algorithm based on a series of the convex optimizations, where each iteration simultaneously minimizes a performance criteria and a sparsity-inducing term, as well as prioritizing grouping sensors in prescribed groups [3]. In the existing algorithms [4], [3], the non-convex sparsity inducing L_0 norm is replaced by a series of weighted L_1 relaxations [5], which doesn’t offer the ability of grouping the sensors. Grouping becomes important in multi-dimensional spaces to avoid eliminating a spatial coordinate from a potential sensor location. We approach grouping by using the L_2 norm alongside the relaxed L_0 norm.

sensor placement ; sensot selection ; mixed L1/L2 norm ; convex optimization ; L0 norm relaxations

nije evidentirano