Abstract
Ring compression is one of the simplest experimental methods used to evaluate the frictional performance of lubricants and the flow stress of materials through calibration curves. In this paper, artificial neural network approach has been proposed for analysing the ring test. Multi-layered BP (Back-Propagation) network has been trained to predict the interfacial friction and the flow stress of the ring material. Inputs to the network are decrease in the inner diameter and reduction in the height of the ring, and outputs are the friction factor and the flow stress. To achieve better prediction, criterion for selection of the topological and learning parameters of the network has been discussed. Three types of prediction schemes have been presented including a mixed prediction using one network, a parallel prediction using two networks and a serial prediction using two networks. Simulations were conducted with respect to the “standard” ring geometry ratio 6:3:2 (OD:ID:To). Satisfactory results were obtained with prediction error of only 1.7% and 2%, at the maximum, for the friction and the flow stress, respectively.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call