Experimental identification of structural uncertainty – An assessment of conventional and non-conventional stochastic identification techniques

• Structural uncertainty identification via different stochastic methods is assessed.
• Two series of experiments are employed under similar operating conditions.
• Non-parametric, conventional parametric, and new CCP methods are employed.
• The methods cannot fully capture uncertainty based on a single experiment.
• Proper identification methods based on multiple experiments need to be developed.

An experimental assessment of several stochastic identification methods with respect to their ability in capturing the structural uncertainty is presented via their application to a lightweight aluminum plate structure. The uncertainty analysis is based on both non-parametric (Welch based) and parametric (VARX based) conventional stochastic identification methods and also on a CCP-VARX based method which operates on multiple data records. The methods effectiveness is assessed via multiple series of experiments under similar operating conditions, focusing on both in-series and inter-series uncertainty. The results of the study indicate: (a) small, yet statistically significant, in-series uncertainty and larger inter-series uncertainty; (b) inability of both the conventional VARX and CCP-VARX parametric methods to fully describe multiple experiment uncertainty through a single identified model.

Figure optionsDownload as PowerPoint slide