Vibrational fatigue estimation using subspace-based system identification method in frequency domain

Nowadays, commercial vehicles such as buses, trucks, light trucks and even cars have been driven longer time periods than the past all around the world. Due to driving them longer time periods, failure of some components may occur while driving them. Especially, some of the failures may take place in mechanical parts when the vehicle is in motion. Due to component failures, vehicles are stopped for repair and maintenance for a while that causes loss of money and time. There are many reasons for the component failure in vehicles. However, one of the most important reasons is vibrational fatigue. The purpose of this study is to examine vibrational fatigue in frequency domain. In particular, subspace based identification method is used for identification of transfer functions in frequency domain. In this study, it is examined that how to get the transfer functions by estimating the structural transfer functions in order to predict stress values on some specific points of heavy duty trucks and how the components of the vehicle is affected by vibrational fatigue. Thanks to getting the transfer functions by using the estimation method, it is found that which part is affected by fatigue and the failure damage could be estimated easily. By using Matlab software, estimated transfer functions are found that have information on stress values (e.g., power density functions PSD) of the critical locations on a vehicle. The developed programs in Matlab environment is based on subspace based identification method, nonlinear least squares method and frequency domain identification method. The frequency domain techniques are selected over the domain time methods due to their less computational time in time domain. There are eight significant steps throughout the subspace based system identification method in order to obtain estimated transfer functions in the algorithm. The data are picked up from the vibration experiments to determine transfer function between selected locations of a vehicle. The data included acceleration/displacement measurements as the input. The outputs are the stress values at selected points in the vehicle. Then, frequency domain fatigue estimation methods are employed to predict the fatigue damage at the selected component. Thus, the parts of vehicles which are exposed to vibrational fatigue could be analyzed without any difficulty since the transfer functions were estimated in frequency domain and fatigue damage is calculated in frequency domain as well. Moreover, there is a need that the transfer functions should be adaptive, recursive least squares method was also studied. The transfer functions were made adaptive for different circumstances such as the vehicle acceleration, weight condition, road surface roughness, slope of the road, fill rate of fuel tank and they are also adaptive for different truck models. Finally, owing to transmitting the stress value information at selected components, fatigue analysis can be completed using them in frequency domain and predicts the fatigue life of mechanical parts in vehicles. The implementations are made on a heavy duty vehicle. It is shown that the developed algorithm can predict the stress variation in frequency domain accurately and the fatigue life of the selected components can be predicted in agreement with the experimental observations in tests.