Performance characteristics and reliability of detonating devices are strongly dependent on the morphology of active material components. The inner structure of a material is called its microstructure. It stores the genesis of the material and determines all the physical and chemical properties. However, the microstructure is highly complex and numerous image defects such as vague or missing boundaries formed during sample preparation make it difficult to extract the underlying structure precisely. Prior work has focussed on using popular image processing techniques like watershed segmentation, seed based region growing, graph-based methods like graph cuts and fast-fine cut to extract material properties (grain sizes, grain counts etc). Features extracted using these techniques rely heavily on the accuracy of the segmentation and they do not scale well with increasing data sizes. In this work, we show how techniques from Scalar Field Topology can be used to automate the analysis of large-scale high-resolution 3d X-ray computed tomography (CT) scans of high explosive images. After initial data preparation, we compute Morse-Smale Complex and apply persistence simplification to extract region specific statistics like grain count, grain sizes, boundary surface area etc and show how these material properties (as a function of persistence) can be used to train a robust classier/regressor to classify/predict performance of different materials. Our initial results show that Morse-Smale Complex and persistence simplification classify a wide variety of materials and it does not depend on the accuracy/quality of the actual segmentation.