(12c) Classification of Chemical Production Scheduling Problems and Approaches, and a General Solution Framework

Classification of Chemical
Production Scheduling

Problems and Approaches, and a General Solution Framework

Sara Zenner, Christos T. Maravelias

The
present talk consists of three parts:

(1)
A classification of chemical production scheduling problems

(2)
A critical review of the types of modeling approaches proposed in the
literature

(3)
A general representation and solution method for chemical production scheduling

Problem
Classification

Scheduling problems appear in a number of sectors
and applications, ranging from the oil industry to the pharmaceutical and
specialty chemical sectors, and from consumer goods
and the food industry to metal manufacturing. In addition, there is a large
number of different processing characteristics and restrictions (e.g., batch
vs. continuous processing, transportation vs. conversion processes, mixing
rules, storage constraints, etc) which lead to a wide range of scheduling
problems. In this paper we present a general classification of scheduling
problems based on a series of attributes, including: (i)
the industrial/market environment within which a company operates, (ii) the
interaction of the scheduling problem with the other planning functions of the
enterprise, (iii) the characteristics of the production environment, (iv) the
structure of the facility, (v) the production recipe, and (vi) the detailed
processing restrictions. We then discuss how three specific aspects
(bill-of-materials requirements, mixing/splitting rules, batch integrity
restrictions) can lead to problem classes that existing modeling approaches
cannot address and we critically review the shortcomings of existing
representations and modeling approaches.

Review Modeling
Approaches

Based on the insights provided by our problem
classification we review existing scheduling approaches in terms of a broad
spectrum of attributes. Most existing method classifications are based on a
subset of these attributes (e.g., facility structure and time representation).
In this talk we attempt to synthesize all previous approaches and propose a
roadmap that includes all known modeling approaches. Specifically, we propose a
multi-dimensional view where the
?universe? of scheduling approaches is treated as a multi-dimensional space
where each dimension corresponds to a key modeling feature. The primary
dimensions are:

a)     
Major entity(ies) modeled:
these typically include orders (or batches) and materials/resources.

b)     
Optimization decisions: the major decisions are batching (or lot-sizing),
assignment, and sequencing and timing; all approaches involve a subset of these
three.

c)     
Time representation: this includes the selection of the type of time grid (discrete vs.
continuous), as well as more subtle choices (e.g., unit-specific vs. common
time grid, local vs. global precedence variables, etc.). 

General
Modeling and Solution Framework

Based on our representation of modeling approaches,
we revisit the classes of problems that cannot be addressed using existing
methods. We explain why this is the case and highlight the ?holes? in terms of
problem representation. We then present a general framework, which is a
generalization of the approach of Sundaramoorthy and Maravelias (2011), to address some of these shortcomings.
Finally, we show how our approach can be used to represent problems that have
been addressed using different modeling approaches, and how other modeling
approaches can be ?reduced? to ours. We close with some computational
results. 

Sundaramoorthy, A.; Maravelias,
C. T. A General
Framework for Process Scheduling. AIChE J., 57(3), 695-710, 2011.
            (DOI: 10.1002/aic.12300).