Embedded modal analysis algorithms on the smart wireless sensor platform PEGASE

Vincent Le Cam (1, 2) , Michael Döhler (3) , Mathieu Le Pen (1, 2) , Laurent Mevel (2)

1 IFSTTAR/MACS - Département Mesure, Auscultation et Calcul Scientifique
2 I4S - Statistical Inference for Structural Health Monitoring
3 CIV - Department of Civil and Environmental Engineering [Boston]

Operational modal analysis and vibration based damage detection of engineering structures have become important issues for Structural Health Monitoring (SHM) and maintenance operations, e.g. on transport infrastructure. Methods from control engineering have been adopted and converted for the application on civil structures. Approaches like subspace-based system identification combine excellent theoretical properties under the unknown excitation properties of a structure with practical usefulness. In this paper, the implementation of covariance-driven stochastic subspace identification (SSI) on the smart wireless sensor platform PEGASE is described. Special care is taken about the fast implementation of this technique since the computations are embedded on the platform and perform in real-time. The most efficient and current version of subspace algorithms has been implemented. Efficiency and memory consumption are primary criteria in this implementation. First validated results will be given for each step of the algorithms: cross- correlation on natural inputs signal from sensors; Hankel matrix output; SSI implementation using the LAPACK library to get a SVD, pseudo-inverse, eigenvalues etc. Results validation has been correlated between PEGASE implementation and the previous processing in static situation: the same data was collected by wired sensors and data-loggers, then, later, processed on a PC using traditional Matlab software. In parallel, from an engineering point of view, a description of the PEGASE wireless platform will be given: generic usage, wide capacities, embedded Digital Signal Processing (DSP) processor and Library over a small embedded Linux Operating System, a very accurate synchronization principle based on a GPS/PPS principle, etc. Perspectives about a complete technical in-situ installation will also be given.

Domaines: Electronique; Instrumentations et Détecteurs [physics.ins-det]

Vincent Le Cam, Michael Döhler, Mathieu Le Pen, Laurent Mevel. Embedded modal analysis algorithms on the smart wireless sensor platform PEGASE. International Workshop on Structural Health Monitoring 2013, Stanford University / department of mechanic, Sep 2013, Stanford (CA), United States. ⟨hal-04481425⟩