A fast reliability analysis for an Unmanned Aerial Vehicle performing a phased mission

It is becoming more common for Unmanned Aerial Vehicle (UAV) to perform phased mission where the phase s causes of failure may be different. The reliabilities of the phases are required throughout the mission in order to make future decisions for the mission. However, previous research of phased mission analysis has shown it to be very complex and take significantly long amounts of time. Also the analysis cannot be performed before the mission because information that is only available when the mission is active is required for the analysis. The aim of this research is develop new methods for a phased mission analysis which can obtain the phases reliabilities on a real structure UAV mission, where all the components are non-repairable, in the fastest time as possible. The present methods are explored and the outcome is that the methods based on Binary Decision Diagram (BDD) analysis are the most efficient. Therefore the BDD analysis is use as a starting point for the new method. The phase mission BDD based methods are improved by altering the procedure of the analysis. Also modules that can appear in many phases can be taken out to simplify the analysis. Search methods that lookup computations that have already been done before are investigated to determine how much impact it has on the speed of the analysis. A method that restructures the phase s mission fault trees to optimize the number of modules that can be taken out is developed. It is tested on a real UAV mission and it is shown to significantly simplify the analysis. This method is extended by situation where a mission is being reconfigured several times throughout a mission and the analysis also has to be done several times. Additional changes are made by using part of the analysis of the original mission for the new one to speed up the analysis. A method is developed which identifies parts of the analysis referred to as groups which can treated as a mini phase missions. Each group can be performed on separate processor in parallel that reduces the online analysis.