work presents a novel approach for enterprise-wide planning
and scheduling problem in paper, pulp and printing industry.
This is essentially a large dimensional mixed integer nonlinear
optimization problem. In the current approach it has been systematically
decomposed into the interconnected smaller sub-problems by
using spatial and temporal considerations. Each of these sub-problems
can be solved relatively easily and with less computational
efforts using existing capabilities of the available solvers
and personal computers.
The proposed decomposition has a hierarchical structure,
which strongly relies on the algorithmic approach with
minimal use of heuristics and permits systematic
analysis of the problem at various levels. The rolling horizon concept is used
to deal with the dimensionality problem along with the changing future demands
and possible disruptions. The resulting set of large scale sub-problems are
solved using ILOG CPLEX 7.1 Solver. The efficacy of the
proposed solution approach is
evaluated by conducting studies on a large dimensional industrial problem (3200
orders, 5 paper machines and 3 month horizon, about 300,000 variables). The
proposed solution scheme provides satisfactory solutions
to the large scale industrial
problem for multiple scenarios in a reasonable time frame (about 2 to 3 minutes
on 2.4 GHz PIV PC). Thus, the proposed solution scheme can effectively deal
with large dimensional enterprise-wise order allocation,
inventory allocation, run
formation, trimming and pattern sequencing problems in a coordinated fashion.
This project was funded by Honeywell Technology Solution
Laboratory. (HTSL), Bangalore and IIT Bombay team working
on the project consisted of Prof Sachin
Patwardhan (PI), Prof K P Madhavan (Consultant), Mangesh Kapadi and S A Munawar
(project engineers). Also, HTSL, Bangalore and IIT Bombay are in the process
of jointly filing two US patents based on the technology developed.