The method of blasting is still considered an efficient and economic method of excavation due to making notable progress in rock excavation. Nevertheless, the challenges in blasting still persists. One of the major challenges include good material preparation. Rock fragmentation by blasting is considered a challenge given the heterogeneity of rockmass, stemming ejection, sub-optimal delay assignment and blast design parameters untailored to the rockmass. Fragmentation modelling was chosen as the study domain for this work. A thorough literature review on fragmentation modelling revealed some common Artificial Neural Network (ANN) architectures used by the authors. The gamut of crucial input parameters was narrowed down by the literature review viz. Spacing to burden ratio, bench slenderness ratio, burden to hole diameter ratio, stemming to burden ratio, powder factor, in-situ block size and modulus of elasticity. This combination covers rockmass parameters, explosive parameters and blast design parameters making the model very comprehensive. A regression model was developed as a benchmark case for ANN architectures. The fragmentation ratio, i.e. ratio between in-situ block size and mean fragment size was used to visualize the dependence of various input parameters using 3D surface analysis. Several ANN models of different architectures were made by changing the number of hidden layers and nodes. Using the pruning method and a novel ANN Performance Index (API), the optimal architecture was determined. It was found that ANN architectures outperform the linear regression model in terms of coefficient of determination and Root Mean Square Error (RMSE) values.