371613 Application of a Mixed-Integer Linear Program Based Planning and Scheduling Tool to Reduce Capital Investments in Batch Facility Debottlenecking

Thursday, November 20, 2014: 10:15 AM
406 - 407 (Hilton Atlanta)
John Robinson1, Kurtis McKenney2, Steve Harding3, Spencer Miller3 and Joseph Pekny3, (1)Amgen, Inc., Seattle, WA, (2)Amgen, Inc., West Greenwich, RI, (3)Advanced Process Combinatorics, Inc., West Lafayette, IN

Understanding capacity for new products in large scale production of therapeutic Monoclonal Antibodies is a challenging proposition.  Production generally proceeds through many sequential unit operations with parallel equipment for the slower cell culture unit operations and a single purification train.  Each unit operation may be supported by a variety of solutions which are prepared in a limited number of mixing vessels.  Storage and delivery of solutions to the applicable unit operation must respect both the connectivity limitations in the plant design as well as the temporal coordination with the rest of the production process.  Additionally, equipment cleaning using a limited set of Clean-In-Place (CIP) skids adds complexity to the planning and scheduling problem.  Often, there is not a single piece of equipment that defines the plant bottleneck, but rather a series of steps that may cross multiple pieces of equipment which limits performance.

Traditionally, understanding plant bottlenecks is a challenge even for products the plant has already run.  Debottlenecking a new product using back of the envelope calculations and broad assumptions is likely to result in inaccurate investment targets.  However, current tools for planning and scheduling support high resolution operational modeling of batch facilities and allow for accurate identification of plant bottlenecks.  Here, we describe how a sequential process of (1) identifying bottlenecks using a model and (2) testing alternative engineering solutions to bottlenecks significantly reduced debottlenecking costs by eliminating unnecessary capital project scope and simultaneously identified previously unidentified bottlenecks that would have been overlooked without the insight of a detailed model.


Extended Abstract: File Not Uploaded