Packaging Line Scheduling Optimization

Oct. 5, 2015
Scheduling software brings artificial intelligence to address packaging processing complexity at Pfizer

Packaging pharmaceutical products is a multi-faceted endeavor. The packaging process comprises several operations including primary packaging filling and processing, which includes labeling and often buffer or desiccant insertion; and secondary packaging that includes box forming, bottle or blister insertion, dosing and safety sheet handling, cartoning, palleting, etc. Much of the packaging process is accomplished by packaging hardware — systems operated by dedicated staff — usually in a clean-room or controlled environment. Packaging plants are often dedicated to the task and field multiple packaging lines.

Pfizer’s operational staff is continually searching for ways to enhance the company’s pharmaceutical manufacturing processes and operations. As part of this ongoing effort, the company determined its scheduling process for packaging of pharmaceutical goods warranted particular investigation. Pfizer’s scientific inquiry into this area of potential operational improvement began with research to determine what alternate planning and scheduling solutions were available.

Pfizer’s investigations led to an intelligent scheduling technology available from Stottler Henke. To better vet the software platform, Pfizer conducted empirical trials to determine first, if the technology could be adapted to Pfizer’s manufacturing environment, and second, confirm whether or not an artificial intelligence (AI) supported scheduling engine could produce results better than the company’s current scheduling process. Using real-world Pfizer manufacturing data, direct head-to-head comparisons with their existing solutions were made. According to Pfizer, Stottler Henke’s Aurora solution produced superior results to the tools the company currently employs including improved graphical reporting capabilities — something Pfizer’s schedulers didn’t have access to before.

Most under-appreciate the difficulty of production scheduling: It is one thing to be able to create a schedule that does not violate any constraints (so is therefore valid), but it is quite another achievement to generate a near-optimal schedule. Unfortunately, the consequences of using a less-than-optimal schedule can be quite significant and can quickly result in millions of dollars of otherwise avoidable costs and other expensive outcomes.

Pharmaceutical manufacturing processes, including packaging, offer an ideal environment to highlight the potential risks associated with inefficient scheduling: At first glance this activity appears relatively simple, but its hidden complexities often open pitfalls the unwary trip into on the path to productivity. Such time and resource “excursions” occur because pharmaceutical packaging processes belong to a class of problems understood collectively as resource-constrained scheduling.

Mathematically, the resource-constrained scheduling problem is NP-hard (nondeterministic polynomial-time hard). This means that there is no realistic way to guarantee that the result provided is optimal. Therefore the scheduling approach and algorithms a manufacturer chooses significantly affect how close one can get to the optimal schedule, and thus its duration. Unfortunately, the algorithms used by most commercial scheduling software have proven inefficient for scheduling resource-constrained problems. As a result and due to the inherent complexity of resource-constrained scheduling, the durations of resource-constrained endeavors can be 10 - 20 percent longer or more, and thus more than needed. Moreover, as the schedule is executed, there may be deviations between the schedule and the actual process execution. Operations may take longer than assumed, equipment may fail, and various other unforeseen problems may arise. The scheduler must update the production schedule to reflect changes in resource availability and notify the staff.

Although there is a variety of quality, customized scheduling systems available, off-the-shelf systems rarely fulfill the scheduling needs of any single domain. This is, in large part, because domain knowledge is crucial to the efficient and effective solution of scheduling problems in general. Industries and domains that realize advantages afforded by intelligent scheduling systems have in the past been either those that can afford a full custom solution or those that fall within the off-the-shelf domain coverage.

To make scheduling software accessible to a broader audience, new scheduling systems need to be created quickly and easily. This requires the insight to understand what is alike in different scheduling processes while still accounting for the many differences. Stottler Henke solves scheduling problems the way human experts do, only by orders of magnitude faster. Essentially, the software provides a framework that takes advantage of the large degree of commonality among the scheduling processes required by different domains, while still successfully expressing their significant differences, i.e., with parts of the scheduling process broken out into discrete components that can easily be replaced and interchanged for new domains.

Aurora Pro-Plan, Pfizer’s specific solution, quickly finds a good schedule by leveraging both domain-independent and domain-dependent heuristics. The base scheduling framework is a foundation on which a number of different systems can be constructed. Thanks to the software’s modularity and therefore inherent flexibility, it’s been successfully applied to myriad industrial domains, including improving pharmaceutical packaging processes. The technology has proven a good fit for Pharma because it is designed to address resource-constrained scheduling problems endemic to its packaging process.
The software can be applied to global production scheduling and optimizing production, not only at the individual plant level, but also across networks of plants. The platform Pfizer uses provides many benefits, including minimizing changeover time and inventory, optimizing production, and maximizing overall equipment effectiveness.

A solid scheduling basis also allows the software to more easily handle complex situations, such as new orders being inserted in real-time. Where traditional scheduling systems use simple algorithms and criteria when selecting the next activity to schedule and when assigning resources and times to each activity, Stottler Henke’s platform can optimize over many metrics, such as carrying costs, while minimizing the production runs by ordering production to minimize those changeover times that are dependent on the tasks selected to be performed before and after the changeover. The software is designed to interface with other enterprise applications, supporting inventory carrying costs and demand changes; and supports equipment downtime and line allocation changes.

When optimizing the schedule for a packaging plant, there are multiple lines that may consist of homogeneous or heterogeneous packaging machines. Optimization needs a main goal to optimize; this is either throughput or minimizing changeover time. On initial review, it may appear that these goals are equivalent, but due to differences in machine types and machine availability, throughput optimization may differ from changeover optimization. For example, when maximizing throughput, it may be better to have slightly higher total changeover so that the fastest machine can be used more.

A multitude of constraints must be considered when performing the optimization process. To maximize throughput of the plant, the throughput of the machines needs to be known, and then the products/SKUs that are being packaged need to be provided. One critical piece of information when optimizing a schedule is the changeover matrix that provides the times when transitioning from one SKU to another.

What needs to be packaged is another important input to the optimization process. This estimate of what the plant should provide can be based upon historical or forecast information or on some combination of the two. In addition, there may be minimum and maximum frequencies with which any SKU must be produced. Another important factor is often the carrying cost, so that products with higher carrying cost may be biased to be produced as late as possible.

Even though the improved scheduling provided by the software provides Pfizer with a significant and sustainable competitive advantage, Pfizer decided to share this technology with the rest of the pharmaceutical industry because the company understands that it is in the best interest of industry and consumers to produce, manufacture and package medicines in the most efficient way possible. To this end, Pfizer has gone to the effort of working with others, including Stottler Henke, to establish the Intellicentic consortium, consisting of various technologies Pfizer has found to be successful in improving its manufacturing and packaging processes.

This success is being rolled out to all of Pfizer’s packaging facilities around the world. Due to the success of Aurora-ProPlan, further plans are being made to leverage intelligent scheduling throughout the entire pharma manufacturing process.

About the Author

Rob Richards | Ph.D.