The correct classification of variables is related to the accuracy of statistical model parameter estimation. The well know existing deviation correction method is Misclassification Simulation Extrapolation (MC-SIMEX) procedure. Survival analysis is the main part of time to event analysis and including Cox Proportional-Hazards model (COX model) and Accelerated Failure Time model (AFT model) two main aspects. We found that the COX models are well explored. However, bias-correction in AFT models with MC-SIMEX mothed still need be investigated, since some of the results of bias-correction are not well accepted. In this study, we use non-parametric extrapolation function instead of linear extrapolation function within AFT model to help us correct bias in AFT model with misclassification. In addition, we evaluated the performance of a modified version of the MC-SIMEX in non-parametric extrapolation function within AFT model.