Abstract:
The aviation industry has expanded enormously during last few decades. The advent of new technologies has made the aircraft design process more complicated. There are certain modeling techniques that can help circumvent the existing design process and yet yield fair approximates. One such approach that helps us to arrive at the outcome with reasonably good approximation is known as surrogate modeling. Surrogate models help circumventing complicated analytical methods, time-consuming simulations and expensive experimental techniques typically used during the design process.
For the said purpose, a database of aircraft including military jets, commercial airliners and unmanned air vehicles was developed from commercially available data. Data collected was scrutinized into dependent and independent variables. Aircraft performance parameters needed to be estimated so they were dependent variables and aircraft geometric parameters being predictors were independent variables. Scalability trends using power laws were developed between dependent and independent variables. The scalability study provided the initial design bounds and developed the foundations for further developing of surrogate models. As aircraft design is a complex process and a single variable might not be sufficient approximations for all scenarios, therefore, surrogate models using multiple linear regression technique were developed. The developed models estimated the dependent variables with high confidence. Moreover these models were validated using a three step process which included verification of each model using quantitative criteria, comparing the models with analytical equations and checking the prediction accuracy of each model. It is thus validated that a multiple linear regression models are best fitted for modeling surrogate models for aircraft performance parameters. The approach proposed in this thesis holds a strong utilization potential among the aircraft design community and aero modelers during preliminary design phase.
