CONSTRAINT-DIRECTED NON-SHARP SEPARATION SEQUENCE DESIGN

Knowledge-based tools are used to design separation sequences where non-key components are allowed to distribute. Specifically, frame-based, rule-based, and assumption-based truth maintenance systems are used. The design tasks include classification, opportunistic decisions, alternative generation and evaluation, and modification guided by constraints. By applying the appropriate information at each design stage, the design process and final design are easy to understand. The use of non-sharp separators, where non-key components can distribute, agrees with the intuition that separation and blending should be minimized. An apparent lower bound on separation purity is defined. The two lower bounds in a column and the relative volatilities are used in the Fenske equation to calculate the minimum number of stages and to model the non-key component distribution. The minimum number of stages is used opportunistically to generate a partial sequence ordering from which alternatives are produced automatically. Guided by the structure of these sequences and the modelled component distribution, separation purities are adjusted to meet the product specifications.