Predicting flow strength of austenitic steels with an IPANN model using different training strategies

Model construction and training strategies of an IPANN were developed to im prove the prediction accuracy of the hot strength of a series of austenitic steels with different carbon content deformed under a wide range of condit ions. The prediction accuracy is largely dependent on the training schemes and model structure because the flow strength varies with deformation condi tions and chemical compositions in a very complex way. The scheme for selec ting training data of every independent input was optimised, so that a gene ralised model could be achieved with less training data. With the strategie s introduced in this work, the effect of the carbon content and deformation was accurately presented in both the work hardening and dynamic recrystall isation regimes. (C) 2000 Civil-Comp Ltd. and Elsevier Science Ltd. All rig hts reserved.