BOOTSTRAP METHOD FOR ESTIMATION OF SPECTRAL BANDWIDTH WITH LIMITED OBSERVATIONS

Characterizing the dynamic response to external forces is an important component of both the design and analysis of safe and serviceable structures, requiring accurate estimates of the natural frequency and critical damping ratio. The simplest methods for determining these parameters are based on the response power spectrum of the structure to an external loading, such as wind. The location and width of the power spectrum’s peak are related to the natural frequency and critical damping ratio of the structure, respectively. Unfortunately, the amount of continuously stationary data required to create a response spectrum with reasonable statistical accuracy is nearly impossible to gather in the field. This paper explores using the bootstrap statistical processing technique to improve estimates of the damping ratio for structures with limited data records. The following will briefly describe the standard method for calculating the critical damping ratio of a structure, as well as listing the sources of error in those calculations. The basic principle of bootstrap resampling is explained, and various methods for applying it to this problem are suggested. Simulations highlight the improved bandwidth estimates via bootstrapped spectra for limited data