A framework for flow manufacturing scheduling systems.

For more than 20 years, the MRP framework has reigned supreme over the scheduling scene. It dominates plant installations, vendor software offerings, and the scheduling literature. It is the center piece of APICS certification exams. In spite of this paramount position, there is some evidence that MRP is not as widely used as one might think. A survey reported in Datamation found that only one U.S. manufacturer in seven used MRP II. This survey, however, begs the question: If American manufacturers are not using MRP--then just what are they using? The answer to this question is really difficult because there are no established alternative frameworks.

In recent years, there has been a wide spread interest in finite capacity scheduling systems. Unfortunately, there is no generally accepted framework for these finite capacity scheduling systems. This situation is made more difficult by the significant differences among finite scheduling packages. Some of these systems are designed for job shop environments and simulate the flow of orders through a job shop using routings and priority dispatch rules. In contrast, other finite capacity systems are designed for flow shop environments and simulate the impact of alternative processor schedules on inventories as materials flow through process stages.

Process flow scheduling (PFS) is a relatively new, general scheduling framework. PFS provides an alternative to the MRP framework for flow manufacturing. It's not a specific software package, nor is it a scheduling technique. PFS applies only in flow manufacturing environments--we are content to leave all the job shops for MRP or the job shop finite capacity simulations. Major users of process flow scheduling concepts are found in chemical, paper, plastics, rubber, food, textile, brewing, petroleum, and repetitive manufacturing industries.

PFS concepts are used in at least eight commercial scheduling packages, which are employed in over 600 plants. Moreover, many flow manufacturers have developed their own scheduling systems that use PFS concepts. These proprietary systems range from simple spreadsheet models to very sophisticated scheduling systems.

The origins of PFS are obscured. PFS concepts have been employed for many years in the process industries. In fact, PFS concepts have been used for more than two decades by process industry firms. By 1982, commercial software based on PFS concepts was available. In spite of the utilization of PFS concepts in proprietary and commercial systems, a framework unifying these concepts wasn't developed until the recent publication of a book on PFS.

Process description

For example, suppose we illustrate a process structure for the Rainbow Chemical Co., a totally hypothetical company, shown in Figure 1. The first process stage contains a separation process that continuously separates the feedstock F into three intermediate products I1, I2, and I3. The separation stage operates three shifts per day and seven days per week. The second stage is a reactor which first processes I1 into bulk product B1, then processes I2 into B2, and finally processes I3 into B3. This schedule is then repeated through the schedule horizon. When the reactor switches products, a transition material is produced that must be scrapped. This results in lost product and an increase in waste management costs. These costs must be added to the incremental labor and lost production time costs to obtain the product switching costs. The reactor, like the separator, operates three shifts a day and seven days a week.

[Figure 1 ILLUSTRATION OMITTED]

The third stage is a finishing line. The finishing line has a mixing operation followed by a packaging operation. The mixing operation blends yellow, red, or blue pigments to color the reactor products. The packaging operation packs the colored intermediate products into large and small containers and stores the resulting products in finished goods inventory at the plant. Every time the finishing line changes colors, the mixing and packaging equipment must be shut down and cleaned. The finishing equipment is relatively inexpensive and operates one shift a day, five days per week. The fourth stage is a distribution operation and consists of transportation equipment and inventories at field warehouses.

Each inventory in the system is required for different reasons. The feedstock inventory, F, is required to buffer continuous requirements of the separation unit from the periodic deliveries of feedstock. The intermediate inventories, I1, I2, and I3, are produced continuously by the separator, but are consumed one at a time by the reactor. Thus, inventory is required to decouple the two processes.

The reactor produces intermediates B1, B2, and B3 in a cyclic manner using a sequence and lot sizes based on reactor technology. Unlike the reactor, the finishing lines can easily switch between B1, B2, and B3. However, the finishing line has a natural production sequence based on color. As a result, inventories of B1, B2, and B3 are consumed as demanded by the finishing line's color-based scheduling sequence. These inventories permit decoupling the reactor sequence from the finishing sequence.

Finished goods inventories are maintained at the plant and in field warehouses. Finished goods inventories at the plant are required to buffer the finishing schedule from the warehouse replenishment schedule. Finished goods inventories at the warehouses are required to buffer distribution operations from customer demand.

The process stages are closely balanced, but since the reactor stage has the most expensive equipment, it is designed and built with the tightest capacity. Since demand is only 90 percent of reactor capacity, the entire process train is throttled back to 90 percent of the reactor capacity.

PFS terminology

PFS uses specific terminology to define the process structure for flow manufacturing systems. The example in Figure 1 will be used to illustrate the definitions for process trains, stages, clusters, and units. Rainbow Chemical has a single process train consisting of four full process stages and a partial stage designated as stage 0 for the feedstock. Each stage is separated from preceding and succeeding stages by inventories. These inventories allow the decoupling of the schedules for the different process stages.

The first and second process stages are relatively simple to describe. Although the first stage contains several types of separation equipment, all of this equipment shares a common schedule. Thus, process unit that forms a single-unit process cluster. The reactor stage also has several types of process equipment. Like the separation unit, however, all of the equipment in the reactor stage is grouped into a single process unit sharing a common process schedule. Since there is only one process unit in the reactor stage, it is also a single-unit process cluster. Note that each process unit must have a schedule and that each schedule is associated with a single process unit.

The third process stage is a little more complex. The finishing line has both a mixing and a packaging operation. There is no inventory decoupling these units. Thus, when the mixer is producing red, the downstream packaging unit must be packing red. Also note that a schedule for the mixer will only have to specify the three colors, while a schedule for the packaging unit must indicate both the color and the package size. As a result, the schedules for mixing and packaging are different, but tightly coupled. Because of this coupling of schedules, we group the mixing and packaging units together in a process cluster. However, since mixing and packaging have their own unique schedules, we call them process units. Accordingly, stage 3 is a finishing stage that has two process units grouped into a single process cluster.

The fourth stage is a distribution stage that consists of transportation equipment and warehouse inventories. Since transportation is provided by common carriers, the transportation operation is modeled as a single process unit with infinite capacity.

This example illustrates the use of PFS terminology to describe a process structure for a flow manufacturing operation. Equipment is grouped into process units that share a common schedule. Process units are grouped together to form clusters that have coupled schedules. Process clusters and their output inventories are grouped together to form process stages. Finally, a string of process stages may be grouped together to form a process train.

Inventories within a process train decouple the process clusters in successive stages so that the clusters may be scheduled somewhat independently. This facilitates linking continuous production to cyclic production, and scheduling successive process clusters in different sequences and lot sizes.

Scheduling strategies and procedures

The central concept underlying all PFS systems is the use of the process structure to guide scheduling calculations. There are many ways this concept can be implemented in different plants, and variations are even found within some plants. Let us examine one of the possibilities for the Rainbow Chemical Co. An overview of the scheduling procedure is given in Figure 2.

[Figure 2 ILLUSTRATION OMITTED]

The first step in developing a scheduling strategy is identification of the master cluster. The master cluster is the most important and/or the most difficult cluster to schedule in a process train. For Rainbow Chemical Co., scheduling the reactor in the least cost sequence and in economic run lengths has the greatest impact on plant profitability. Thus, the reactor is designated as the master cluster, and it is scheduled first. A forecast for B1, B2, and B3 is used to calculate economic run lengths. These run lengths and the target sequence are used to develop a trial reactor schedule.

Having developed a schedule for the reactor, a schedule is developed for the finishing cluster. The finishing cluster is scheduled next because it is the most difficult of the remaining clusters to schedule. The central scheduling problem for the finishing cluster is determining which process unit to schedule first. The unit scheduled first is called the key unit. If the mixer is chosen as the key unit within the finishing cluster, then color changeovers are minimized. On the other hand, if the packaging unit is chosen as the key unit, then package changeovers are minimized. Since color changes are more costly and difficult than package changes, the mixer is chosen as the key unit in our example. A trial schedule is now proposed for the mixer based on run length guidelines for each color. Because the mixer output directly feeds the packaging unit, the packaging unit schedule is the same as the mixer schedule except that it shows the package size.

Having developed trial schedules for the reactor and finishing clusters, we now check the inventories of the intermediate products B1, B2, and B3. The reactor pushes material into these inventories and the finishing cluster pulls material out. Material balance calculations can be performed for each period in the scheduling horizon and displayed as a plot of inventory versus time. If the inventory is above the planned maximum (the effective tank capacity) or below the planned minimum (safety stock), then an inventory violation has occurred, and the reactor or finishing schedules must be modified. If inventories are within the guidelines, then scheduling can proceed.

Next, a warehouse replenishment schedule can be developed. Since transportation equipment is readily available from common carriers, the replenishment schedule will be developed using distribution requirements planning (DRP.) This technique schedules replenishments just as an item's inventory reaches its safety stock level. The resulting requirements are offset for the transportation lead time and the resulting demand on the plant finished goods inventory calculated. At this point, plant inventory must be reconciled. The finishing cluster schedule pushed material into the plant finished goods inventory and the DRP schedule pulled material out. If the resulting inventories are within target minimums and maximums, then scheduling can proceed. If not, then the DRP, finishing, or reactor schedules must be modified.

The last steps involve scheduling the separator and checking its input and output inventories. These steps are unlikely to create problems because the separator continually processes a single feedstock. Thus the only real scheduling issue is the production rate for the separator. A trial rate can be chosen, and the inventories of I1, I2, and I3 checked. Finally the feedstock resupply schedule is developed and checked for feasibility.

This example reveals a key point about process flow scheduling: schedules were the authority to produce in all stages except the distribution stage, where DRP was used. Although PFS systems can use work orders, most PFS systems' process units are controlled by production schedules.

Let us now use this example to examine the three fundamental principles found in all PFS systems.

PFS principles

The first process flow scheduling principle is that scheduling calculations are guided by the process structure. While MRP calculations use the bills of material (product structure) to guide scheduling calculations, PFS uses the process structure. The scheduling procedures for Rainbow Chemical illustrate this point. The schedule was developed by progressively moving from stage to stage and iteratively building a feasible schedule.

The second process flow scheduling principle is that process clusters are scheduled using processor-dominated scheduling or material-dominated scheduling. This principle addresses the issue of what is scheduled first--processors or material. Processor-dominated scheduling was demonstrated by the reactor in the Rainbow Chemical example. The first scheduling step was to propose a trial Gantt chart for the reactor. After the reactor schedule was developed, the intermediate inventories of I1, I2, and I3 were checked. Thus, a processor-dominated scheduling approach was used, with the processor being scheduled before the materials. Processor-dominated scheduling facilitates scheduling process units in least-cost sequences and in economic run lengths.

In contrast, let us now examine the material-dominated approach used in scheduling the distribution stage. Here a material plan is first developed using distribution requirements planning. Following this, the availability of transportation equipment is checked. This material-dominated approach gives priority to maintaining minimum inventory levels over using equipment efficiently.

The ability of PFS to employ both material-dominated or processor-dominated scheduling makes it more flexible than MRP. The scheduling logic of MRP is limited to a global material-dominated scheduling approach. MRP first develops a plan for all materials and then checks the plan for feasibility by checking capacity. In contrast, PFS iteratively schedules both material and capacity as it works its way through the process structure.

The third process flow scheduling principle is that process trains are scheduled using reverse-flow scheduling, forward-flow scheduling, or mixed-flow scheduling. In theory, scheduling can begin at any stage in a process train. If, as in the Rainbow Chemical example, scheduling begins with an internal stage and proceeds to the outer stages, then an inside-out, mixed-flow scheduling strategy has been employed. On the other hand, if scheduling begins with the last stage and proceeds backwards through the process structure. then a reverse-flow scheduling strategy has been employed. Finally, if scheduling proceeds forward through the process train in the same direction as the material flow, then forward-flow scheduling has been employed.

Summary

Process flow scheduling is a new scheduling framework that unifies the scheduling practices of many flow manufacturing firms. This framework helps explain how some of the many firms that do not use MRP schedule their plants. PFS schedules are detailed, accurate simulations of the flow of material through high-volume manufacturing plants.

For further reading

Foley, M.J., "Post-MRP II: What Comes Next?" Datamation, Dec. 1, 1988.

Taylor, S. G., and S. F. Bolander, Process Flow Scheduling: A Scheduling Systems Framework for Flow Manufacturing, Church Falls, Virginia: American Production and Inventory Control Society, 1994.

Sam G. Taylor, Ph.D., is a professor in the college of business at the University of Wyoming, located in Laramie. Steven F. Bolander, Ph.D., is also a professor at the University of Wyoming.