Modelling and operation of intermediate storage in noncontinuous processes

Gyeongbeom Yi, Purdue University

Abstract

In this work, batch process production failures are tactically classified into three categories; periodic failure, long term failure and stochastic failure. Intermediate storage systems in production lines can be classified into three types; single input/single output (SISO) storage, serial train and multiple input/multiple output (MIMO) storage. Analytical models are used to investigate the forward and backward propagation of the effects of failures along the production lines. The modelling equations for intermediate storage are developed for the cases of no production failure and periodic failure. Using these results, the limiting size of intermediate storage can be calculated accurately for SISO storage but only the upper or lower bound are available for MIMO storage system, except for the case of symmetric structures. The open loop approach leads to several different operating algorithms according to the structure of the storage facilities and the failure properties. The operating algorithms are designed to satisfy physical constraints and to fully utilize storage capacity. They provide the shut-down initiation times and shut-down durations for the given failures. The effectiveness of these algorithms are demonstrated with simulation examples. The closed loop approach leads to two nonlinear model predictive feedback inventory control algorithms: a serial algorithm and a parallelized algorithm that can be applied to general storage structures. In both cases, the controller measures the hold-up and the flow profiles which are subject to process disturbances. The most up-to-date information is transferred to the processor which is calculating the current control input. Simulation examples demonstrate that the proposed parallelized controller can absorb process disturbance and convert all parameter variations into only cycle time variation, while fully utilizing available storage capacity. All operating or control algorithms developed in this study have a theoretical foundation in basic physical principles and mathematical analysis. Special emphasis is placed on insuring that the operating or control algorithms are explicit and compact so as to be useful for real-time application. This study thus contributes to contemporary manufacturing goals of minimizing storage size and operating cost as well as providing a framework for the automating inventory management. (Abstract shortened by UMI.)

Degree

Ph.D.

Advisors

Reklaitis, Purdue University.

Subject Area

Chemical engineering|Operations research

COinS