Stationary Random Vibration Load Identification, Computer Simulation And Experimental Study

Loading identification has long been a difficult problem for multi-input multi-output stationary random vibration problems contrary to the fact that system identification is well developed and advanced. In this paper, the Inverse Pseudo Excitation Method (IPEM) is utilized for reconstructing the excitations. Numerical simulations are used to select the optimal combination of reference points, which tested to be very useful in the setting up and processing of experiments so as to obtain reasonable prediction of input loading from selected structural responses, e.g. a few structural acceleration responses. The selected combinations of reference points are then measured on a uniform cantilever beam and a four-story structure made of organic glass subjected to random excitations as well as frequency response functions and coherence coefficients are employed as criteria to adjust the setting status of drive rods, accelerometers etc.Excellent experiment results are achieved from above mentioned tests and generalized epsilon decomposition is tentatively proposed to solve the problem of matrix singularity. It is shown that IPEM combined with computer simulation is a good approach to identify the power spectral density matrix of the applied forces while and before actual experiments proceed.