MODEL-BASED JOINT INPUT-STATE ESTIMATION ON A THIN-WALLED CANTILEVER BEAM

Abstract eng:
In the field of structural dynamics, measuring external forces directly is often not feasible, whereas structural response data is far more accessible. Force identification is an inverse problem concerned with estimating forces using measured dynamic response. This paper presents a practical application of a joint input-state estimation algorithm. The algorithm gives minimum-variance unbiased estimates of states and input forces in a discrete-time linear system, where stochastic noise in the measurement data and model description are accommodated for. An experimental study of a thin-walled cantilever beam is presented to demonstrate joint input-state estimation. The finite element modelling is done in the software ABAQUS with shell elements. A modally reduced order model is created, containing 13 selected modes. We introduce the general approach of obtaining strain measurements in a modal format. A sensor network is designed from a discussion on optimal placement of accelerometers and strain gauges. It is also shown that the sensor network constitute a sufficient observational ensemble for estimation of states and forces. The studies demonstrate successful identification of forces generated by a simple mass-spring system. In addition, the strain is predicted at a location not co-located with the sensor network. The predicted strain agrees well with a reference strain measured at the same location. Lastly, we address some of the practical problems encountered and recommendations for further experiments.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket