Quench collection and artificial neural network prediction of micro and ultra-fine aluminium agglomeration phenomena in combustion of solid rocket propellants: Experiments and modeling

Main objective of this study is to understand how micro/ultra-fine aluminium propellant formulations affect the condensed combustion products (CCPs) formation. The agglomeration of aluminum particles usually occurs on the burning surface of aluminized composite propellants. It leads to low propellant combustion efficiency and high two-phase flow losses. To reach a thorough understanding of aluminum agglomeration behaviors, agglomeration processes, and particles size distribution of Al/AP/HTPB and UFAL/AP/HTPB propellants were studied by using a quench collection experimental technique. The propellants were quenched at six quench distances (5, 23, 35, 47, 59,71mm) from the burning surface. Tests are performed in the 2-8MPa pressure range. Six propellant formulations are considered in this study in which five propellants are micro aluminized, out of which two are catalyzed and sixth propellant with ultra-fine aluminium. All formulations are based on a bimodal size distribution of AP particles. Propellant formulation variables like coarse-to-fine ratio and aluminium content were modified to assess these effects on aluminium agglomeration process at different pressures. The result shows that the particles are spherical and their sizes vary from 31µm to 115 µm for non catalyzed and 28 µm to 136 µm for catalyzed propellants. The synthesized ultra-fine aluminium powder is used with the intensity weighted harmonic mean size of 438nm which is produced by RF induction plasma method. Ultra-fine aluminum powder exhibits significant agglomeration with the size ranging from 11 – 21 μm. This will substantially benefit from exhaust signature point of view and also reduction in two-phase flow losses to thrust in contrast to micron sized-aluminum particles. To predict the agglomerate diameter of metallized/ultra-fine aluminium of composite propellants, a new artificial neural network (ANN) model was derived. Five Layered Feed Forward Back Propagation Neural Network was developed with sigmoid as a transfer function by varying feed variables in input layer such as Quench distance (QD) and pressure (P). The ANN design was trained victimization stopping criterion as one thousand iterations. The validated ANN models will be able to predict unexplored regimes wherein experimental data are not available. The resulting agglomerates sizes from ANN model, matches with the experimental results. The percentage error of testing data is less than 3.0% of the six propellants used in this work.