Incorporate Dependence Structures in the Runway Overrun Model using MATLAB Copula Models

Abstract

It is a well known fact that in aircraft industry, safety is very important. Nowadays there are several methods with which we can evaluate safety in an aircraft operation. Modern aircraft contain several sensors in order to measure flight data such as temperature, velocity, pressure, etc. during flight. An instrument called Quick Access Recorder records all the information provided by these sensors. With the help of this data, formulas and statistics, we can get an idea of how safe a flight can be. The main objective of this work is to evaluate the dependence structures between contributing factors that can cause an incident in a civil flight. This is done with a previously generated code available to calculate a simulation based on subset simulation for runway overruns. For this project, we have access to flight data from several different types of aircraft. This data is recorded over time and each set of data will have different time series so this has to be taken into account when obtaining the results. Data collected includes landing temperature, landing QNH, landing weight, etc. Our goal is to calculate the dependencies between the contributing factors and integrate them into the runway overrun incident model. The samples that are going to be used in this project, are generated by a statistical method called Copulas that will give us a variable dependency, instead of a constant one. The programs we will use are Matlab and MySQL. MySQL is needed for storing the aircraft data is stored in .sql files. Matlab are used to run the calculations, for this, an available software is used to calculate Copulas. In the next step the dependence structures are integrated in an available runway overrun incident model at the institute of Flight System Dynamics. Since the software to represent the dependence structure was developed in Linux, this is also the operating system chosen for this Bachelor's Thesis. Things that we will need to take in to account are the discrete and continuous factors, the fitting distributions and finally the contributing factors that are the ones that really have an effect on the incident probabilities. Subset simulation is performed, integrating the Copulas method and is to be compared to the method applied by the existing model.