Systems can be modelled from many different perspectives. The focus of our group is on modelling of systems so that 1) the models can be reasoned with to deduce useful information about systems 2) the models can be analyzed to predict useful properties of systems. We discuss a couple of ongoing projects in this space.

A model driven approach to enterprise data integration
A large enterprise has hundreds of data sources that are produced in different contexts. These contexts have to be understood correctly for correct integration. Resolving the semantic variations of these contexts and reconciling them into a unified view is in fact the central problem of data integration. Conceptual models can play a key role in this process. With a suitably rich conceptual model, much of this process can be automated.  A conceptual model describes the concepts and relationships of a domain, and the rules and constraints that apply in a given context. The information captured in these models can be exploited for a number of purposes – answering queries, discovering model mappings, validating model mappings, matching and merging entities, and so on. We have developed a modelling framework and a core reasoning engine to process these models. We have also developed a few end-user tools on top of this framework: a query answering tool, an ETL generation tool, a data migration tool, and a tool to discover mappings between different models. A few more tools are in the pipeline.

Integrated Computational Materials Engineering

Integrated Computational Materials Engineering (ICME) is a new approach for materials development that envisages using physics based models, empirical models and human expertise in an integrated manner to significantly reduce the time and cost of development of new materials and their manufacturing processes. The properties of a material (such as strength, hardness, fatigue life) depend on its microstructure. The microstructure of a material in turn depends on its chemical composition and the manufacturing processes it is subjected to. These relationships are not well understood. While there are physics based models for predicting what microstructure comes out of some of the manufacturing processes, there are no such models for predicting what properties come out of a given microstructure. Often the only option left is to mine past experimental data and construct approximate models. These models can then be used for predicting what materials and processes might satisfy a given set of property requirements. One can then use process simulation to try out these short-listed materials and processes, and select the ones that best meet the requirements. All this will not be possible without a comprehensive IT platform. The platform should support among other things: building of a comprehensive knowledge repository on materials and processes, collection and integration of data from a variety of sources, data mining and model learning, knowledge services for material selection, process design and process simulation.