Thesis: David Michael Williams

Doctor of Philosophy, Keble College, University of Oxford, Trinity Term 2000

Numerical Modelling of Real-Time Sub Structure Testing

Summary

Current dynamic testing methods can prove unrealistic due to the scale at which test components are modelled, the rate at which they are loaded or the boundary conditions to which they are subjected. A new test method, termed "Real-Time Sub-Structure Testing" seeks to provide a more realistic testing environment for energy dissipative components. The method tests structural components at full or large scale and in real-time. The physical test interacts with a computer model of the structure surrounding the test component. In this way, the in-situ behaviour of the test component is evaluated in relation to the overall structural response.

The testing method requires fast and realistic modelling of the surrounding structure and a rapid interaction with the physical test specimen. For these reasons, a new non-linear finite element method has been proposed in order to model the surrounding structure behaviour efficiently. The method uses the Central Difference Method time stepping integration scheme together with a newly devised basis. The proposed basis consists of the structure’s elastic modes and additional Ritz vectors, which are calculated from the inelastic static displacement shapes of the structure. The displacement shapes correspond to the same static spatial distribution of loading as the intended dynamic excitation, and are intended to characterise the inelastic behaviour of the structure. The method has been validated against a Newmark event to event algorithm as well as Drain2DX. The non-linear dynamic response of a propped cantilever beam and portal frame structure was investigated. The response evaluated by the algorithm agrees very favourably with both validation analyses. The new algorithm was also shown to be faster than the Newmark procedure in simple benchmark tests.

In addition, a numerical model of the testing apparatus has been developed in order to simulate complete tests for the purposes of testing procedure development and validation. The model is developed using Matlab Simulink. Parameters for the model are deduced from published data, experimental component tests and open loop step response calibrations. The model behaviour was found to be very sensitive to the parameters used. However, after calibration against open loop tests the model reproduces the observed laboratory behaviour to a good degree of accuracy.

In an attempt to predict the behaviour of an actual test, the laboratory model has been coupled with the new structural solution algorithm to simulate a virtual test. The simulated results compare favourably with experimentally observed data demonstrating the usefulness of the overall simulation as a test modelling tool.

This thesis can also be downloaded from the ORA website