Structural response predicton of a payload carried by a jet fighter

Military specifications are traditionally used to calculate the structural response of a payload carried externally by jet type aircraft both frequency and amplitude. The purpose of this thesis is investigation and prediction of structural response of equipment mounted in payload by using artificial neural networks (ANN). A representative payload which is similar to the Mark-83 warhead with guidance units in terms of mass and geometry is used to construct ANN models in real flight conditions. Seventy-five different flight conditions in terms of Mach number and flight altitude are used to train the ANN for low and high frequency components as military specification proposal. Also, 8 flight conditions apart from the training set of flight conditions are used to test the approach. Structural responses are collected with accelerometers mounted in payload by the data storage system. In signal processing, structural response is expressed as acceleration power spectral density (PSD) functions in the frequency domain. Procedures to predict the maximum response from measurements are determined with statistical limits in the literature. Besides the well known limits in the literature, third order polynomial normal and logarithmic transform are used to predict the structural response first time in this study and the performance of the different limits are compared. Structural response predictions with tolerance limit in frequency domain are compared with military specification and experimental results. A structural design case study is performed using ANN prediction, military specification and experimental data to demonstrate the loads magnitude difference. Loads for structural design are calculated with random loads using PSD data. Three sigma rule is considered between 20-300 Hz frequency bands. A support structure is optimized for three different loads by VRAND GENESIS tool. Supports designed based on experimental and predicted loads are observed to be lighter than the support designed according to the specification. As a result, it has been found that the developed approach in this thesis can be used in response prediction of a payload. The military specification response spectrum is conservative. Tolerance limit prediction performance is better than military specification. Distribution free and normal tolerance limits predicted low frequency acceleration spectral density magnitudes more accurately. Their prediction performances were better than those of the other tolerance limits and that of the military specification. Third order polynomial transform predictions are found to be reasonable with respect to normal prediction limit and envelope approach. The response prediction approach proposed in this thesis works well for payload carried externally by jet fighter.