No abstract

International audienceThis paper deals with the identification of a stochastic computational model using experimental eigenfrequencies and mode shapes. In presence of randomness, it is difficult to construct a one-to-one correspondence between the results provided by the stochastic computational model and the experimental data because of the random modes crossing and veering phenomena that may occurs from one realization to another one. In this paper, this correspondence is constructed by introducing an adapted transformation for the computed modal quantities. Then the transformed computed modal quantities can be compared with the experimental data in order to identify the parameters of the stochastic computational model. The methodology is applied to a booster pump of thermal units for which experimental modal data have been measured on several sites

-When studying mechanical systems, engineers usually consider that mathematical models and structural parameters are deterministic. However, experimental results show that these elements are uncertain in most cases, due to natural variability or lack of knowledge. Therefore, engineers are becoming more and more interested in uncertainty quantification. In order to improve the predictability and robustness of numerical models, a variety of methods and techniques have been developed. In this work we propose to review the main probabilistic approaches used to model and propagate uncertainties in structural mechanics. Then we present the Lack-Of-Knowledge theory that was recently developed to take into account all sources of uncertainties. Finally, a comparative analysis of different parametric probabilistic methods and the Lack-Of-Knowledge theory in terms of accuracy and computation time provides useful information on modeling and propagating uncertainties in structural dynamics.

No abstract

A benchmark is organised to quantify the variability relative to structure dynamics computations. The chosen demonstrator is a pump in service in thermal central units, which is an engineered system with not well-known parameters, considered in its work environment. The blind modal characterisation of the separate pump components shows a 5%-12% variability on eigenfrequency values and a less than 15% frequency error in comparison with experimental values. The numerical-experimental MAC numbers reach 0.7 at the maximum, even after updating. An example of modal results on the pump assembly fixed is presented, which shows a larger discrepancy with measurement values, essentially due to the modelling of the interfaces and boundary condition, and to the possible simplification of the main components F.E. models to reduce their size. Though a significant frequency error, the first overall modes are correctly identified. If this tendency can be confirmed from all the participants' results, the conclusion to be drawn is that, if the predictive capability of F.E. models to represent the dynamical behaviour of sub-structures is satisfactory, the one relative to structures that are built-up of several components does not allow their confident use. Additional information issued from measurements is needed to improve their accuracy.