Experimental Study and Neural Network Modeling of the Stability of Powder Injection Molding Feedstocks

Abstract

Due to their complex rheological behavior, feedstocks for powder injection molding (PIM) may exhibit non-homogeneous flow and separation. This can produce defects of green parts during mold filling, resulting in their cracking and warpage during debinding and sintering, and ultimately in poor physical and mechanical properties of the final part. An experimental rheological study has been performed to evaluate the influence of solids loading, shear rate, and powder particle size on feedstock stability. A micro-rheological explanation is given for the macroscopic effect of separation, and an instability index has been defined to describe quantitatively the threshold beyond which the variation of viscosity becomes unacceptable for PIM purposes. A neural network model has been developed for predicting the viscosity of feedstocks made from binary blends, when the powder characteristics, blend composition, and shear rate are known. The system enables determination of the process parameters for which powder-binder separation occurs in a given feedstock.