Accurate durability prediction is an important requirement in today's automobile industry. To achieve the same, it is imperative to have a good estimation of time histories of strains, accelerations etc. at various locations on the vehicle structure. This is usually difficult to obtain as a typical data acquisition exercise takes lots of time, cost and effort. This paper aims to address this problem by predicting the strain time histories accurately at various locations on the vehicle chassis from a few channels of measured data using Artificial Neural Networks (ANN). The predicted strain histories were found to be quite accurate as the error in fatigue lives between the measured and the thus predicted time histories at various strain locations were found to be less than 15%. This approach was found to be very useful in collecting huge amounts of customer usage data with minimum instrumentation and small sized data loggers. This has given a big fillip to customer usage data collection in the automotive industry, where the size of the loggers has been a constraint in the collection of such data (especially in the case of motorcycles). Further the predicted time histories were used for component level simulation, servo hydraulic vehicle level simulation, diagnosing problems with respect to failures of components in the field, arriving at a correlation between road and the test rig etc.