Scott J. Mason, Eray Cakici, Jun Jia, Abraham J. Lachowsky, Letitia M. Pohl, C. Richard Cassady, Pingjian Yu
University of Arkansas
The objective of this project is to model and analyze E. & J. Gallo Winery’s Modesto, California cellar tank piping network to determine the routing path for wine through the cellar tank piping network that minimizes wine damage and optimizes cellar tank and piping network resources. Also, this project develops a heuristic-based solution approach capable of routing wine flows in seconds.
The process of converting grapes on a vine to bottled wine typically takes E. & J. Gallo Winery (“Gallo”) four to five weeks, depending on volume and/or product mix requirements. Demand is planned according to a 13 week blend schedule-knowing the “big picture” enables planners to decompose demand into smaller, more manageable time buckets. Planners establish what needs to be done each week in order to meet the overall blend schedule. Planners issue work orders to the cellar with specific due dates-work orders with due dates in the current week become the various “operations” that must be performed in the current week. These operations require an extremely inter-dependent set of decisions that are dynamically made by multiple parties. The key component linking all of these decisions is tank assignment.
At any point in time, some of the tanks in the Gallo Tank Farm are storing varying quantities of wine and blends at varying stages of their production process, while another group of tanks may be available for use by the current week’s operations to be performed. Researchers at UA will develop a methodology to help Gallo assign work orders to tanks that comprehends tank attributes (size, location, valve type, processing capabilities/ preference, etc.), work order attributes (operation code, quantity/size, due date, subsequent process steps/operations, etc.), and current tank farm status (contents of each tank, tank reservations, tank availability (now or at some time in the future), etc.). Our research goal is to provide a recommended assignment of work orders to tanks for all phases of the wine making process.
The project formulated and developed heuristic solution procedures for routing flows through an extensive network of tanks and pipes. The key challenge associated with solving our optimization models to optimality is the size of the piping network. One method of dealing with this challenge would be to break the Modesto campus down into “sub-wineries” based on some type of departmental structure. Models and algorithms can be found in the final report.