SBIR Phase II: Uncertainty Analysis of Manufacturing Process Models

This Small Business Innovation Research (SBIR) Phase II project proposes to create a robust software system for performing uncertainty analysis of process simulations for manufacturing. For simulations that are large or that contain many parameters, even the best Monte Carlo, or importance-based sampling methods for uncertainty analysis can be prohibitively expensive. Consequently, systematic uncertainty analyses are rarely implemented for complex systems. This proposal presents a plan to produce a commercially viable package of a new method for quantifying simulation uncertainty, based on polynomial chaos expansions. The method can determine the probability density functions of black-box model responses and can identify quantitatively which of the parameters contribute most to uncertainties in responses for multivariate inputs and outputs. The unique sampling approach enabled by the use of polynomial chaos expansions allows more accurate resolution of probability distribution functions at a very small fraction of the cost to achieve similar results with more traditional uncertainty-analysis methods. While illustrative examples from the chemical manufacturing industries will be used to demonstrate the software functionality, the methodology has broad application to such fields as circuit design, risk management, allocation of experimental resources, chemical plant design and operation of production systems. Due to the ability to handle arbitrary or black-box simulations, the methods can be applied as easily to economic market analysis, or global climate modeling, as to chemical process design.