Helicopter load signal and fatigue life estimation using low dimensional spaces

The accurate estimation of helicopter component loads is an important goal for ensuring safe operation as well as for life cycle management and life extension efforts. In this research, the use of computational intelligence, neural network, and machine learning techniques is explored to estimate helicopter component loads and their fatigue life, in particular the main rotor yoke load of the CH-146 Griffon helicopter. This paper describes efforts to reduce the number of dimensions of the input data using feature generation techniques in the load estimation methodology, beginning with intrinsic dimension analysis to determine the number of intrinsic dimensions in the data. The data set is then mapped using different implicit methods to a low-dimension representation of the original data, which is then used for load estimation and fatigue life analysis for comparison with the results of the original 26-dimension input data. The resulting load signal and fatigue life estimates from the low-dimension representations are in most cases equally if not more accurate than those for the original input data. These promising results show that the low-dimension representations retain the relevant data from the original input data set and perhaps discard spurious data resulting in more accurate estimates.