Planning for One Piece Flow Cells

Bas Timmermans from the Netherlands asked a question about OEE and one piece flow cells:


How would you best plan such a cell? Would you make an hour-by-hour chart based on the cycle time and plan only the hours needed, leaving some contingency time at the end of a shift for catching up the losses during that shift if any (representing OEE)? Or would you calculate the OEE into the takt time? From what I’ve read about Toyota I get the idea that they use overtime to compensate losses. Is that correct?

Bas has many if not all of the bits of the answer. Let’s try to put them together. In some cases we’ve written about some of these topics in the past, highlighted as links below. First, OEE is not typically measured at the cell level but by the critical machine or machines in the cell. This would be determined by which machines have limited capacity, poor quality yields or have the potential to limit the output of the cell due to any of the six big loss factors. For review, OEE is overall equipment effectiveness. It is calculated as:

OEE = availability (A) x performance (P) x quality (Q)

OEE takes into account the six big losses, which are 1) Unplanned breakdowns, 2) machine set up and changeovers, 3) idling and minor stops, 4) reduced speed, 5) defects and scrap, and 6) start up and yield losses. These correspond to the A, P and Q categories respectively. For more information please see this article. Unless there is a real need to measure OEE as such at the cell, I would recommending one or more of the six big losses specifically instead.
When planning for one piece flow cells these losses need to be taken into account just as when planning any system. Takt time is the average rate of customer demand and this number can be used to calculate everything from crew size to standard work in process quantity to lead-time through the cell. The theoretical takt time is calculated purely based on the net available time and customer demand for a particular period:

takt time = net available time per period / customer demand per period

However in practice there is something called actual takt time which is used to operate the production line or one piece flow cell. The calculation for actual takt time takes into account various factors including losses. There are other approaches to planning capacity for one piece flow cells when there are known losses, such as this example of taking changeover time into account for takt time calculation.
Once the takt time has been set and the line is running, hour by hour charts or in the case of high-speed, high volume automation real-time electronic performance displays are used to track the planned versus actual production output and reasons for any losses. Making the losses visible is and lining up organizational resources to rapidly address these problems is a key lean management behavior. Even Toyota does not aim for 100% but rather expects to run somewhere between 95% and 99%, depending on the line and the product. A 100% rate indicates that there is too much inventory, time or other slack in the system, hiding waste.
Many lean companies realize that running a factory 24 hours 7 days per week is an apparent efficiency hiding other system-wide wastes. This focus on high asset utilization can result in a loss of flexibility, remove preventive maintenance time and weaken the mindset to “finish today’s work today.” When there is barely enough time between shifts to communicate and hand over the work people begin to lose their sense of responsibility for fulfilling customer orders within their shift. The ability to catch up and meet production requirements during each day is an essential planning discipline and the 2-hour gap between shifts is one example of how on-time delivery in enabled when running a lean operation.
Hopefully that answers your question Bas. If not please post a follow up. As always, others are welcome to chip in.

5 Comments

  1. Rob

    January 7, 2009 - 4:46 am

    Incorporating OEE into your efforts to increase manufacturing
    excellence has several additional benefits. OEE’s
    inherent causality will ensure that improving your OEE
    will directly impact your success. It also seamlessly supports
    many of the modern manufacturing improvement
    techniques (including Six Sigma, lean manufacturing,
    continuous improvement, etc). Because it’s easily calculated,
    OEE can be quickly understood and implemented in
    your planning structure. It’s not subjective or discriminatory,
    making it easily used for comparison across industries
    or shared with customers as necessary.

  2. Bas Timmermans

    January 7, 2009 - 12:36 pm

    Thanks Jon for your fast respons. I’d like to check my understanding of your reply to my question with an example. Say I have a customer takt of 5 minutes (demand per shift is 100 products, shift is available to run for 500 minutes). Suppose the pacemaker in the line has a cycle time of 4 minutes and an average OEE of 80%. I would set the line speed at 4 minutes and base my hour-by-hour chart on that. This means that I plan every shift to be ready 100 minutes before the shift ends. In reality mostly random (unpredictable) losses will occur on most days, using some or all of the 100 minutes ‘spare’-time, or even more into overtime. On a relatively good day, runtime can end before end of shift and remaining time can be used for improvement work or to let people go home early to compensate for overtime.
    Would you agree with this example?
    Off course the hour-by-hour chart will also be used to track the reasons for missing the hourly targets so improvement initiatives can be started to eliminate the root causes and shortening the planned run time further and improving the OEE.

  3. Jon Miller

    January 11, 2009 - 10:06 pm

    Hi Bas
    As a theoretical example it’s basically correct.
    However if the 80% OEE was truly due to “mostly random (unpredictable) losses” this implies breakdowns, minor stops and scrap. These are very bad things and I would say these need to be corrected right away, even consider taking the machine off-line and buffering so that the line can run at the normal 5 minute takt time.
    This is why OEE can be dangerous if looked at only as an average number. The 85% OEE mark is often said to be world class, but really only if it is a product of AxPxQ = 92%x92%x100%. Losing 20% due to planned changeovers and planned downtime is ok. This can be planned for. But if you have truly unpredictable breakdowns and quality problems adding up to 20% that is far from world class.
    It is best to measure Availability, Performance and Quality separately and consider how to schedule capacity on the line according to the type and severity of losses leading to 80%.

  4. ankit

    February 1, 2009 - 10:27 pm

    i want to calculate the OEE by using the cycle time so how could i do that. Please guide me.

  5. Jon Miller

    February 2, 2009 - 3:26 pm

    Hi Ankit,
    You will need more than cycle time to calculate OEE. You will need available time, downtime, good parts produced, bad parts produced and time per piece (cycle time).
    OEE = A x P x Q, where
    A = (total available time – downtime)/total available time
    P = total pieces produced/((net run time)/(cycle time per piece))
    Q = (total pieces produced – total bad pieces produced)/total pieces produced
    Cycle time only comes into play in the calculation of P or Performance. The P value is a measure of the actual speed as compared to the possible speed or nameplate speed. A machine capable of producing 80 per hour but capable of producing 100 per hour would have a Performance of 80%. The 20% losses derive from various minor stops, slow downs, or willfully reducing performance for whatever reason.
    Cycle time or the time per piece can be used to determine the possible speed if the nameplate value is not available. For the example above the cycle time would be 36 seconds per piece
    3,600 seconds per hour / 100 pieces per hour = 36 seconds per piece
    If you know that the machine was running for 7 hours and produced 600 pieces you can calculate Performance value as
    P = Actual output / possible output
    P = 600 pieces / ((7 hr x 3,600 sec/hr)/(36 sec/piece))
    P = 600 / 700
    P = 85.7%
    If you are running a continuous process where the nameplate speed is known, there is no need to use cycle time in the equation above. If you are doing multiple products with different cycle times then it is better to capture the possible speeds of each product to get a more accurate picture. The best way to use cycle time for OEE is not for calculation but or observation and improvement.