AbstractThe use of neural networks in the modeling of the inflation stage of the blow molding process is discussed. The simulation is enacted in the reverse process direction, predicting initial membrane dimension requirements from final part thickness distributions. This situation has practical implications because tooling costs and machine down times can be minimized with the information obtained. The optimal network topology entails simultaneous pre‐ and postprocessing of the data. An adaptive window is employed for modeling of the effects of adjacent segments. The optimal window length is 16 for both the input and output layers in the network topology. Simulations are run for bottles blown using various constant and pulsed die gaps. © 1993 John Wiley&Sons, I