WIND LOAD IDENTIFICATION ON A TOWER STRUCTURE

Force identification, as a specific type of inverse problem, i nvolves reconstructing the force distribution on a structure from measured data. Having an accurate knowledge of the exact loading on a structure or mechanical system is important in any situation where the loading forms a crucial component in the design of the structure, i.e. when the unknown load can be regarded as critical and dictates the design, but measuring these loads is not always possible resulting in the necessity to obtain them by means of such reconstruction. Since the wind load may constitute a critical loading for high towers or mass structures and since, furthermore, it is in general dif ficult to measure a time-varying force on a structure, especially if the force acts as a continuous load ing, the identification of time-varying wind loads on a structure presents itself as an interesting and ch allenging force identification application. Obtaining these forces requires an indirect computation using measured response data in conjunction with a mathematical model of the structure. In its broadest sense force reconstruction can be viewed as being comprised of two main steps, the first of which involves the creation of a coefficient matrix wh ich relates the measured data to the unknown forces. The content and structure of this matrix depends on the specifics of the formulation used and are dictated by, for instance, the domain in which the solution is obtained (frequency/time), the available information used to establish the input-output l ink (impulse response functions, the output from a strategically chosen forward problem, etc.) as well as other less fundamental issues pertaining to the flexibility of the formulation in regard to prescribed initial conditions, the input quantities used (displacements/accelerations), and so forth. The second step, referred to as regularization, comprises t he manipulation of this, mostly illconditioned, coefficient matrix in an attempt to make it bett er suited for meaningful inversion. These manipulations are a function of the specific type of regulari zation used which, in turn, is a problemspecific issue. In the context of a wind load identification application the i nverse problem will be stated in the time domain and the coefficient matrix can be created either by usi ng a state-space model of the structure, with impulse response functions serving to establish the input-output link, or by using an external FE