Managing variation in chemical batch processes.

Two statistical quality control (SQC) techniques with an emphasis on variation reduction are developed for chemical batch processes. Conventional Shewhart SQC tools are inappropriate for chemical batch processes when dealing with (1) multiple variance components and (2) frequent manipulation of process variables. The first technique introduced in this research applies Equal Probability (EP) charts, a family of Shewhart-type mean and variability control charts with equal probability zones and simple run rules. The EP charts are designed to control a variety of chemical batch processes with different variance component structures. Average run length evaluation shows the superiority of EP charts to state-of-the-art Shewhart-type charts especially in detecting decreases in process variability. The second technique applies multiple linear regression (MLR) analysis, a well-developed methodology for building empirical linear relationships between control variables and quality indices, to identifying major variance contributors and facilitating variation reduction in batch dye-making operations. A variance component analysis framework is also developed to identify the limitations in and expand the applicability of conventional control charting approaches and process capability analyses. A generalized process capability index is defined for processes with two variance components. The usefulness of the proposed techniques is illustrated via computer simulations and real industrial examples from dyestuff manufacturing processes.PhDApplied SciencesChemical engineeringIndustrial engineeringUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/129960/2/9711939.pd