Numerical methods for load and response prediction for use in acoustic fatigue

Sammanfattning: Acoustic fatigue can occur in structural elements of an aircraft exposed to very high sound pressures. To deal with acoustic fatigue, mainly empirical methods have been applied and often late in the design phase. Current design guidelines have three main limitations. First, they do not say anything about the load intensities. The load levels can be determined either experimentally or numerically. Experimental testing tends to be expensive and time consuming. It is also desired to deal with acoustic fatigue early in the design phase. Therefore, it is desired to turn to numerical methods to determine the load levels. Second, the design guidelines assume that the spatial distribution of the load is uniform. In other words, the load is assumed to be perfectly in phase over the entire structural element. This assumption limits the accuracy of the response prediction and by extension the fatigue prediction. Third, the design guidelines are limited to a simple, single surface panel with linear response. In this thesis, both the load and response prediction are performed by numerical methods. The load is determined using Computational Fluid Dynamics (CFD). From the CFD simulations, both the load intensities and the spatial distributions are extracted. This solves the first and second mentioned limitations. The extracted load is used as force input to a Finite Element (FE) simulation of the exposed panel structure. Since complex structures and non-linearities can be handled using the FE-method, it avoids the third mentioned limitation. Two cases of separated flow are used as model problems for acoustic fatigue in this thesis. In both model problems, the simulations are compared to existing measurements. In Paper A, a ramped backward-facing step is used. The flow over the step induces a load on an aluminium sheet fitted downstream of the step. With the exception of the cut-off, or shedding mode, frequency being overpredicted, the spectral qualities of the load and the load intensities are well captured. The panel response prediction compares reasonably well with the existing measurements. In Paper B, a reduction in a range of low frequencies of the downstream load is observed when the ramped backward-facing step is lined with chevrons or serrations. The model problem used in Papers C-E is flow over an inclined fence at transonic Mach number and realistic Reynolds number for aircraft operation. A segment with cyclic boundary conditions of the flow setup is simulated in Paper C. This result in well predicted cross-spectra, but an energy concentration in the auto-spectra is not properly resolved. In Paper D, a full three-dimensional simulation of the entire setup is performed and it is concluded that the missing energy concentration in the auto-spectra is properly captured. In Paper E, the response of a realistic aircraft panel structure is simulated using FE random response analysis with the CFD-simulated load as input. The response is found to be sensitive to the cross-spectra of the input load. The strain predictions vary with strain gauge location. However, only one strain gauge is off by more than a factor of two, which appears to be the best one can hope for when using the design guidelines in favourable conditions and with a measured load. Therefore, the main conclusion of this thesis is that the method of using CFD to calculate the load which is to be used as input to an FE response simulation can produce useful results for acoustic fatigue.