Qualitative reasoning framework for process systems with spatial patterns

A novel methodology for the extraction of detailed information from numerical and experimental data on chemical engineering systems, and the creation of a descriptive language is proposed. The qualitative representation reduces large sets of quantitative data into an explicit symbolic description that supports both human and computer comprehension. This technique can be applied to any field exhibiting spatial variations and is exemplified using fluid flow as the field of interest. The representation provides enough information to describe the motion, while reducing the numerical complexity.; Important features and patterns in a flowing system may be hidden in the quantitative results. Symbolic representations of recirculation patterns in a flow are developed and the size and shape of the pattern is classified. This qualitative language utilizes the moments of an object and a shape matching procedure.; The qualitative representation can facilitate reasoning tasks in both experimental and computational domains. Bands of high particle concentration in a partially filled Couette are identified and detected. The amount of data required to describe the features and their motion is reduced by orders of magnitude. This is exemplified by calculating the lag between multiple bands and noting the comparison to a similar computation using Fourier methods.; Simulated results are also be analyzed with the symbolic representation, allowing quick and simple decisions on the viability of a specific design. This is shown using a CVD reactor example where the qualitative representation identified designs in which a uniform deposition would be produced. Multiple simulations can easily be queried in order to reduce the number of feasible design options to a minimum. This is exemplified using flow across a notch, where patterns that were present in simulations at different flow rates are recognized and the different regimes singled out.