A recent blog I wrote for a course at Pitt

This is the blog that I just wrote as a submission for homework of ENGR2402 course at the University of Pittsburgh, Introduction to Collaborative Scientific Programming. Also, available at http://collab.sam.pitt.edu/blog/31 . This gives you some information about what I am doing in my research at this moment.

Mixing granular flow with HPC 

Hi, my name is Tathagata Bhattacharya (I go by just Tatha). I am pursuing my PhD in the Granular Transport Group of Chemical and Petroleum Engg. under the supervision of Prof. J J McCarthy. Before joining the PhD program, I was working in a corporate R&D lab for couple of years following my masters. I love to do interdisciplinary research as my background involves a combination of mechanical engg (undergraduate), materials science (masters) and chemical engg (PhD). I am proficient in C and Matlab. I also learned C++ during my undergraduate studies but never used it seriously in any recent project. Therefore, I am also looking forward to brushing up my C++ skills through this course.

Let me give you some background of the work that our group is interested in. In our group, we seek to unravel the mysterious behavior of granular matter from a fundamental aspect, and to accomplish that goal, we employ modeling and simulation in conjunction with controlled experiments. Granular materials are ubiquitous in everyday life: they range from cosmetic powders to vitamin pills, from jelly beans to sugar, from beach sand to the rings of Saturn, and from breakfast cereals, grains to interstellar dust. Although very common, granular materials pose unique challenges and exhibit counter-intuitive behavior that makes research in this field exciting and stimulating. Granular materials possess the potential to exhibit new physics in certain situations if a controlled experiment is closely observed. Understanding the fundamental behavior of these materials can make or break a vast number of man-made and natural processes. In fact, this class of materials ranks second, behind water, on the scale of priorities of human activities and endeavors. Hence, even a fractional advancement in our understanding of the behavior of granular materials can have a profound impact on our economic and general well-being. It is estimated that more than 60% of the manufactured goods in the USA alone rely on particle technology (or the understanding of granular materials). So, that was the motivation behind our work. Let's dig into a little detail of what my work actually entails.

The main objective of my work is to gain an understanding of the factors affecting mixing, segregation and related phenomena of granular mixtures. Mixing of granular materials is essentially accompanied by segregation; however the fundamentals of the processes are not well understood. Small differences in either size or density lead to flow-induced segregation, a complex phenomenon without parallel in fluids. While a qualitative understanding of the mechanisms of segregation has existed for some time now, there are remarkably few models, which give quantitative predictions of the extent of mixing and segregation. Such information is particularly important in the analysis and design of industrial mixing operations in the chemical, mineral, metallurgical, pharmaceutical, food, ceramic, environmental and construction industries. I am using the Discrete Element Method (DEM) to investigate the mixing and segregation of granular material in a prototypical solids mixer -- a rotating drum. We also use other idealized equipments (like a chute or hopper), which have great engineering importance to test our hypothesis or models. DEM calculates the trajectories of individual particles based on Newton's laws of motion by employing suitable contact force models and a collision detection algorithm (like molecular dynamic simulations). In this approach, properties of particles such as size, shape, and density can be directly specified and we believe that DEM is a suitable tool for analyzing segregation. Details such as velocity and concentration profiles for every component in a mixture can be obtained. However, the number of particles and their shapes that can be simulated in DEM is limited by today's computer power, such that many engineering scale processes are beyond the reach of DEM. Therefore, this is a great opportunity where this particular course may come handy in addressing some of the limitations that DEM has. For example, we use a serial code (written in C, developed in-house, distributed under GPL) where it takes weeks (even months!) to obtain meaningful results. So far we have simulated a few thousand to a hundred thousand particles using that code. Profiling of the code revealed that the bottleneck seems to be the contact detection algorithm. Therefore, speed up is possible if we employ the tools of high performance computing (HPC) such as parallelization, GPU computing, etc, and use efficient search algorithms (computer scientists are very good at that!). So, my research work has all the elements of a typical physics problem where it can challenge the ability of present day computers and algorithms. A true multidisciplinary approach is necessary to address the challenges. Therefore, I am eagerly looking forward to learning the basics of these HPC tools from this course and if possible, collaborate with others with common interest. My expectations for this course are quite high where I expect that I should be able to obtain the working knowledge to modify my code and fit it with the HPC tools to speed it up. This will increase the usability and utility of the code to encompass many other engineering problems, which are beyond the reach of the code. Also, as the course name suggests, I am eager to know how open source projects are executed through a collaborative platform. I hope that we will get some glimpses of online collaborations through this course.