In manufacturing plants like the one I have been consulting for recently, that has limited human resources, having a well-planned and synchronized production schedule is critical for the smooth flow of parts through the processes within the manufacturing plant. But as important as the production schedule is, there are other key factors that must be considered before a viable production schedule can to be developed. Perhaps the most significant factor that must be considered is the degree of operational stability of the equipment being used to produce the products. That is, if the equipment is unreliable and unpredictable with excessive amounts of downtime, then it is virtually impossible to develop a practical and dependable schedule for producing parts. Without stability, the level of predictability will be very low.
Another important factor that enters into the production flow equation is the degree of flexibility of the work force, especially when a limited number of human resources are employed. That is, if the work force in place is limited on the number of different machines and parts they are able to run, then scheduling becomes much more difficult. One could even say that in circumstances like this, the human resource level might be considered the system constraint.
Finally, two other factors that enter into the creation of the scheduling process are the quality level of parts (i.e. yield) and the demand requirements placed on each machine/operator combination. If the yield losses are excessive, then these must be added into the scheduling assumptions and planned accordingly. For example, if the scrap rate is 5 % on average, then 5 % more parts must be produced to meet demand requirements.
While all of these factors are important to the development of a viable production schedule, there must also be an active improvement plan in place to eliminate the barriers to scheduling. An analysis and review of existing machine downtime and quality information must be at the forefront of this improvement plan. A system for capturing this key data must exist, but unless it is used to its fullest extent to determine the focal points for improvement, it is just a database.
So the question becomes one of what should the data analysis look like or what information is important? Clearly the reasons for equipment downtime should be available for review, but in what form should it be displayed? And what about the quality losses….shouldn’t that be available for review as well? Should the review be only daily information or should trend analysis also be available? Obviously daily results are important so that a real time investigation of the issues can be completed. Memories fade, so as soon as the data is available, it must be investigated. How an operation is performing, as a function of time, is equally important because single event problems are not nearly as damaging as chronic ones. Because of this, historical plots (e.g. run charts) are very important as well.
My recommendations for how the data should be analyzed are very straightforward and simple to understand. Daily results for both production and quality should be discussed at daily production meetings and then once per week a simple Pareto Analysis of both downtime and Quality can be very helpful. In addition, time based production run charts should be developed to monitor production results as well as control charts on critical quality measurements. Those responsible for each machine group should be required to develop and discuss action plans on how they plan to improve both production and quality.
Once we have created a more stable production environment, then we can focus on developing a discrete scheduling system. Drum Buffer Rope (DBR) is a methodology that can provide relief in a resource starved environment. There are several key principles behind Drum-Buffer-Rope scheduling that must be understood including:
1. In any set of resources, some will be more heavily loaded than others which are referred to as Capacity Constraints, or CCR (capacity constrained resource). In some plants there is really only one CCR while in others, there might be several. The easiest way to locate the constraint is to walk the process and finding the largest pile of WIP.
2. The most capacity constrained resource will always dictate the rate of work flow from raw materials through to finished goods. This is an important concept because knowing where it is will permit us to focus on individual processes and identify each of their specific constraints.
3. There is no value in having any resource that feeds the constraint produce at a faster rate than the constraint. Common measurements like manpower efficiency or equipment utilization might encourage this, but the only outcomes we observe from using these metrics are increased WIP, extended lead times, more expediting, and deteriorating on-time delivery performance. Plus, the excessive WIP ties up floor space and cash as well as increasing the chances of part's damage and undetected quality issues. Both of these problems can be hidden by the seemingly endless waves of WIP.
4. The production rate of resources that are fed by the constraint is dictated by the output rate of the constraint. In other words, the output of the entire process is dictated by the constraint. Because of this, it is absolutely necessary to keep the constraints producing at all times.
These four points lay the foundation for a scheduling system named Drum-Buffer-Rope, so let’s now look at each individual component of DBR.
Once the constraint has been identified, a finite schedule based upon the capacity of the constraint (i.e. the Drum) must be developed for the work that has to pass through it. The schedule can be something as simple as deciding the number of parts to produce and must also include the timing of when each of the parts are due to be completed and shipped to the customer. In addition, the schedule must also consider normal yield losses. Essentially the constraint (the drum) sets the production pace for the entire process.
DBR’s buffers exist in two distinctly different dimensions, time and stock. And because the constraint controls the throughput of the entire process, every minute wasted is a waste of the whole plant’s production capability. So in order to get the most out of the plant, employees must get the most out of the constraint(s). For this reason, we can never let the constraint be starved of parts. One logical answer to this is to make sure that there’s always a buffer of work in front of the constraint so that it is NEVER starved of work. And although it is a stock buffer, it is created as a result of managing time and not stock. In other words, parts must be scheduled to reach the constraint prior to it running out of parts. Let’s look at a simple example.
Suppose that when raw materials are released into the gating operations of the process, it takes an average of 27 hours to flow through the different process steps until they reach the constraint. But because of normal disruptions and statistical fluctuations, many times semi-finished products would not get there in time. Things like unplanned downtime, quality issues or even operator absences happen routinely. In order to counter these delays, we might want to release the parts and materials hours ahead of when they due at the constraint to protect the constraint from starvation. In other words, we want to establish a buffer in front of the constraint. The question becomes, how long in advance should they be released? One way to calculate this buffer is to determine the “normal” average length of time from release of materials until they reach the constraint and calculate one third of this time. In our example, the size of our buffer would be 27 hours divided by 3, or 9 hours. This means that if everything flows smoothly, then the work will arrive at the constraint 9 hours earlier than needed thereby creating a 9-hour stock buffer. The advantage of using this buffer is what happens when there are disruptions and statistical fluctuation in the processes in front of the constraint. As long as the delays don’t exceed 9 hours, then the work will still arrive early or on time at the constraint.
One of the most effective ways to monitor the flow of parts so that they will reach the constraint on time is to create a visual buffer management system. In our example we said that it takes 27 hours on average from release of raw materials into the process until the semi-finished parts arrive at the constraints. We calculated a buffer size of 9 hours to assure that the parts would arrive on time at the constraint. The visual buffer management system divides these 9 hours into 3 segments of 3 hours each and color codes them as green (1-3 hours), yellow (4-6 hours) and red (7-9 hours). If the parts are in the green zone (i.e. the first 3 hours), then they will most likely arrive at the constraint on time. If the parts are late arriving at the constraint and fall into the yellow zone (4-6 hour point), then plans should be put in place to expedite the parts in the event that they exceed the 6 hour point. If they pass into the red zone (7-9 hours), then the plans to expedite should be implemented. This system works quite well as long as the part’s status is closely monitored.
The concept of the rope is probably the simplest DBR concept to understand in that the only parts and materials that should be released into the gating process should be those needed to support the Drum schedule. And while this may seem obvious, in reality for many people it isn’t. In many manufacturing plants work is released into the gating operations simply to keep operators busy so that workers and machines aren’t idle. The performance metrics manpower efficiency or equipment utilization are many times behind this behavior. Many times batch sizes are inflated to avoid lengthy changeovers so materials are “pushed” into the process.
So to summarize:
1. Identify the constraints in the system (this corresponds to the first of Goldratt’s 5 Focusing Steps, “Identify the constraint.”)
2. Examine the orders within the system, consider the finite capacity of the constraint, and schedule the work in detail through the constraint. That is, schedule backwards from the customer due date through the entire process. This corresponds to the second of the 5 Theory of Constraints Focusing Steps, “Decide how to exploit the constraint.”
3. Calculate a time buffer and add it to the average set-up and run times of the processes supplying the constraint. This is also part of TOC’s Step 2 (Decide how to exploit the constraint).
4. Using the rope, only release materials and parts into the gating processes of the plant based upon the needs of the Drum (constraint) and the timing of the Buffer.
5. Monitor the buffer penetration by dividing the total buffer into 3 equal quantities and visually record each 1/3 as green (1/3 of total buffer time), yellow (2/3 of total buffer time) and red (Total buffer time). Green indicates no action is required, yellow requires the development of a plan to expedite the parts, and red requires execution of the expedition plan.
There are many variations and refinements on this basic technique, but as described above, it’s one of the keys to dramatically improving the flow of parts through processes as well as improving on-time delivery performance while at the same time shrinking lead times and work in progress inventories.