Paper Machine Scheduling Yields Gains at James River Naheola Mill

Built in the 1950s, the Naheola mill, located in the heart of the Southeastern pine forest belt, is James River's largest manufacturing site. Product valued at more than $2 billion/year is shipped from this facility by rail and truck. The mill's product is manufactured on five paper machines and two paperboard machines. The towel and tissue are converted at this fully integrated facility into Northern bathroom tissue, Brawny towels, McDonald's napkins, and other familiar-brand consumer and institutional paper products. The paperboard is converted by James River's Dixie plants into cups and plates, and appears as packaging for products sold throughout the world.

In recent years, as computer hardware and software grew increasingly affordable and capable, James River management became interested in new techniques for the development of manufacturing schedules. James River's customers were becoming more demanding, and an expanding product line made production scheduling more complex. With competitive pressures and an interminable industry recession squeezing profit margins, management needed to provide added customer service while minimizing schedule-related costs-all with minimal capital outlays. A new look at the scheduling process was required.

"HIGH-TECH" SCHEDULING. The use of scheduling decision-support software had been considered for several years at the Naheola mill. The headquarters logistics group, which had beneficially used computer-based systems for centralized long-range planning, was also bullish on the potential of such systems for detail-level mill scheduling. However, mill personnel established tough requirements for any scheduling program to be operated at Naheola. It would have to:

• Demonstrate significant benefits over existing scheduling methods.
  
• Be sufficiently user-friendly to be used routinely by mill production schedulers.
  
• Be thoroughly automated to require as little demand on the users as possible.
  
• Earn the support of mill operating personnel by addressing their concerns.
  
• Be robust and reliable, thus system "crashes" and bizarre results would have to be infrequent.
  

Rather than forcing business processes to fit within the constraints of a few standard, "one size fits all" scheduling modules, the Chesapeake tools allow for the development of customized scheduling models. Such models can more accurately reflect the often necessarily unique and complex business systems involved in manufacturing a suite of high value-added products. This allows more realistic abstractions of the "real world" for more optimally trading off the often-conflicting demands and cost factors involved in scheduling.

Naheola's MIMI-based scheduling systems despite their underlying sophistication, are operated with relative ease by mill logistics personnel through a pointer, menu, and prompt list-driven graphical user interface. The user interfaces were designed in cooperation with the schedulers to make both the schedule development process and model maintenance follow a logical flow and be intuitive and self-explanatory.

To preserve model integrity, ensure accurate results, and make the system easier to use, extensive error-and-reasonableness checking are conducted on all input data, whether manually entered or brought in from other systems. Intelligible messages point to the likely source of data or solution problems so that the user will seldom need to refer to the manual.

"PARALLEL" EVALUATIONS. Initially, Chesapeake applications consultants assisted James River personnel in setting up the MIMI package and prototype scheduling models. Mill schedulers and information technology personnel also attended training programs at a Chesapeake office and received one-on-one training at the mill. Subsequently, Chesapeake advisors were consulted by phone and at user group meetings, but the system development effort was accomplished almost entirely by Naheola-based James River personnel.

The first system implemented was designed to schedule the paper machines. These machines can run numerous grades, and many grades can run on multiple machines. The "customers" for the paper machines' parent rolls are primarily the mill's own converting operations, although some product is shipped to other locations. The scheduling problem involves capacity balancing, minimization of total grade change time, avoiding runouts of parent rolls supplied to converting, and trim optimization.

Once an operable prototype model was ready, a period of parallel scheduling began. Schedules were developed simultaneously, using the existing and MIMI-based methodologies for the five tissue and towel machines, for a number of scheduling cases. For both modes, scheduling-related costs were identified and quantified, the principal costs being:

• Grade/color change waste
  
• Parent roll run-out costs (due to disruptions in converting operations)
  
• Trim waste
  
• Slitter setup time
  
• Deviation of scheduled production from ordered amount.
  

After the system was made production-ready and put to use, a general improvement was noted by mill operations personnel. For example, trims seemed tighter, product changes were easier, grade runs were longer, parent roll inventory levels decreased, and the ability to meet required delivery times improved.

SCHEDULING FOR BOARD MACHINES. Based on the results for the paper machines, mill management approved development of a system to also schedule the paperboard machines. MIMI's optimization techniques and expert system and database capabilities, implemented on a fast workstation, gave mill management confidence that substantial savings could be realized over the existing paperboard scheduling system. Furthermore, once all the pulp consumers (i.e., the board and paper machines) were being scheduled, the concept of coordinating pulp production with pulp consumption via an integrated package had great appeal.

Using the MIMI package, Naheola logistics and information technology personnel developed a paperboard scheduling system that they integrated with a state-of-the-art order entry and tracking system. Once again, parallel comparisons with the existing scheduling method showed quantifiable, direct reductions in schedule-related manufacturing costs of about $3.6 million annually.

These cost reductions, which went straight to the company's bottom line, stemmed from reduced trim waste and reduced deviation of scheduled production from ordered amounts. Further benefits, deemed significant but not formally quantified, were derived from reduced carrier (truck or rail) wait time and reduced product dimensional variation and work load due to a reduction in the number of slitter changes.

More recently, parallel comparisons for several scheduling cases were made between the Naheola system adapted for James River's second largest mill (Camas, near Portland, Ore.) and the existing scheduling system employed at that mill. These comparisons yielded potential cost reductions of about $1.4 million/year in reduced trim waste by adopting the MIMI-based system developed for Naheola. Other James River mills have also expressed interest in adapting the Naheola scheduling systems for their own use.

SCHEDULING SYSTEM CONFIGURATION. The scheduling systems at the Naheola mill employ most of the modules for production scheduling offered in the software package. These modules provide operations research-based algorithms (allowing, for example, linear programming of waste minimization, sequencing optimization, capacity balancing, and run-out solutions), as well as tools for expert system logic, procedural software development, database management, and graphical-user interface development.

The MIMI software at Naheola has run for more than three and one-half years under the UNIX operating system (MIMI also runs under Microsoft Windows NT) on a Hewlett Packard Model 730 Series 9000 workstation (140 million integer instructions/sec, 40 million floating point operations/sec). Of note, the hardware and system software have run without failure the entire time. Users access the system through the mill's twisted-pair Ethernet network from their IBM-compatible PCs, which run X-terminal emulation (Reflection X) software.

Both the mill's paper and board machine scheduling processes are now well-automated, producing schedules that are best for the mill. Clerical chores relating to scheduling have been reduced, and mill schedulers are able to focus more on refining the business rules, constraints, parameters, and cost estimates that are used in determining schedules.

Examples of business rules would be: "make the easiest grade changes possible as long as they avoid parent roll runouts in converting," or "if possible, avoid slitting rolls that are too narrow on the ends of reels so that there's not a danger that they will tip over and hurt someone." Constraints involve machine capabilities, minimum run lengths, minimum and maximum inventory levels, etc. Parameters include values such as machine rates, minimum and maximum trim widths, etc. The scheduling systems typically run dozens or hundreds of scheduling scenarios, calculating the total costs of each one and choosing the one with the lowest cost.

FUTURE PLANS. Now that both the board and paper machine scheduling systems are online, development is well under way to integrate the scheduling of the "back end" of the mill (the pulp mill and utilities) with their customers-the paper and paperboard machines. As a first step, this will provide the pulp mill and utilities with forecasted pulp demand, which will be available over the mill's proprietary millwide process database system. It will show plots and tables of projected pulp demand by type of pulp in one-hour increments up to six days ahead. Among the resulting benefits, the mill expects that the pulp mill will be able to operate at a more constant rate (reducing process upsets and product variation) and that the use of expensive dried pulp as make-up, which compensates for pulp mill shortfalls, will also be reduced.

GLENN C. PICKARD is manufacturing systems consultant, James River Corp., Pennington, Ala., and ROBERT L. YEAGER is president, Integrated Marketing International, Stamford, Conn.