Nowadays, may finite element packages have been developed and they are widely used for structural dynamic analysis. However, the results of analysis differ from those of the actual test. The discrepancy mainly comes from the uncertainty of finite element model such as joint properties, boundary conditions and nonlinearities. Generally, mechanical structures are composed of substructures connected by joints such as bolts and bearings. While the finite element representation of plain substructures is well developed and reliable, joints have a lot of uncertainties in being accurately modeled and affect the dynamic behavior of a total system. To identify joint structural parameters accurately, various techniques have been proposed in which experimental data are integrated with a corresponding finite element model. However, practical problems which are often encountered in conventional methods are that measurement noise is nuavoidable and may lead to faulty results, and that it is difficult to install sensors at joint positions. In order to overcome the difficulties, it is required to develop and efficient identification method for accurate joint structural parameters.
In this work, two different approaches are proposed for the identification of joint structrual parameters from partial response measurements and known excitation. One is the time domain method in which time domain signal is used to identify joint structural parameters, the other is the frequency domain method in which subset frequency response functions are used. In the time domain method, joint structural parameter identification problem is converted into a state estimation problem in nonlinear dynamic system. After regarding the joint structural parameters as independent states, unmeasured responses and joint structural parameters are simultaneously identified by use of recursive least square method. In the frequency domain method, unmeasured frequency response functions are estimated from the measured frequency response functions, and then joint structural parameters are identified from measured and estimated frequency response functions. In both methods, attention has been placed on problems associated with incomplete measured data and measurement noise.
The methods are tested and discussed numerically using example structures and the feasibility for practical applications has been demonstrated through an experiment. As the proposed methods make use of only a few measurements which are not necessarily related to the joints but related to easily accessible points, the measurement efforts can be reduced greatly in comparison with other conventional methods. As far as the results of this study are concerned, it is expected that the proposed methods will work as a successful means to identify joint structural parameters.
