Total Productive Maintenance

TPM is implemented as part of the IMPROVE phase in a DMAIC Six Sigma project. However, its purpose is to control the inputs (to allow stable output) in projects where the process is dependent on:

  • Performance (how well the machine runs when it is running)
  • Availability
  • Quality 
  • As you recall, these are the three factors that make up OEE, Overall Equipment Effectiveness. OEE is often used as a lagging indicator metric to gauge a TPM program. TPM is a critical principle for Lean manufacturing. If machine uptime (availability) is not predictable and product can not flow smoothly and reliably then there will be inventory and buffers must be kept to protect the customer. 

    Excess Inventory is waste, it ties up cash, takes up space, and may have shelf life. TPM has many other names such as Total Predictive Maintenance, Total Process Management, Total Preventive Maintenance, and others but they are all slightly unique and components of Total Productive Maintenance.

    Preventing downtime and errors is important and there are many tools such as NVH monitoring, infrared image surveying, ultrasonic tests, that can predict failures before they actually occur to keep machines "available" when they are needed.

    A robust preventive maintenance program is also key to a TPM program. Tracking and executing according the PM manuals are inputs to preventing unplanned downtime and quality defects. Similar to regular oil changes and tire rotations on a vehicle.

    It is beneficial that the users of the machines be involved in the TPM process. Maintenance departments should handle the major items but operators and regular users should have some say and responsibility to maintain a continuously improving OEE.

    The TPM status should be visual. Visual Management is another component in Lean Manufacturing. Computers, graphic charts, statistics are not necessary either. Although they have a time and place, visual management can be done with hand written charts, dry erase boards, magnets, and cards (such as Kanban cards).

    For example, hours on a machine can be hand written and the next due date, then it is easily visible the status of the PM for that machine.

    Facility Maintenance Metrics

    Mean Time Between Failures (MTBF)

    The Mean Time Between Failures (MTBF) is the average time between each failure.

    Some of the variables to iron out before applying is the definition for "uptime". A machine running at a fraction of its intended performance is likely not acceptable to be considered "uptime". Whatever decision is made, that it is applied consistently across all pieces of equipment.

    What exactly is a "failure"?

    A complete stoppage is one more obvious answer. Some may also consider a "failure" once the item or equipment experiences a slowdowns or reduced performance from an ideal level, but don't actually stop the machine. Again, whatever the definition is for failure, it should be uniformly applied to all pieces of equipment. 

    MTBF = (Total up time) / (number of failures)

    There are some items that are not repairable but they are replaced. Such examples are light bulbs, switches, torn belts. In such cases, the term Mean Time To Failure (MTTF) is used. 

    There is also the debate of planned downtime. Robust TPM programs have planned downtime for maintenance and predictive tools may create planned replacements or repairs in effort to reduce unplanned downtime and variability in uptime performance. 

    Ideally, the higher the MTBF the better. However, it is likely to plateau at a certain point due to planned downtime and intended maintenance. Then the challenge becomes how to reduce the planned outages and get better life out of the components or items involved so these planned intervals can be expanded. 

    Mean Time To Repair (MTTR)

    The Mean Time To Repair is the average time to repair something after a failure. As above, it is important to clarify what exactly constitutes a failure and downtime vs uptime. 

    "Uptime" at a significantly compromised rate of production due to poor maintenance is usually not acceptable. Allowing this to continue can show a better MTBF than the story in its entirety should show. 

    Mean Time To Repair = (Total down time) / (number of failures)

    The MTTR puts an emphasis on Predictive and Preventive Maintenance. Better preparation, spare parts programs, predictive analysis, are methods to reduce the MTTR. 

    Not all repairs are equal.

    What constitutes an acceptable repair? This should be defined in the definition of a failure as well. The machine should not only be "up", but it should be up to a certain level of sustained performance before the time can be counted as "uptime". 

    The GB/BB should help (allow a team member to be the author) develop a Standard Operating Procedure or a Work Instruction to clearly define the variable and metrics. As part of the CONTROL phase this is the type of deliverable that would be expected from the Six Sigma Project Manager.

    MTBF and MTTR Example


    Given that over a period of time the following information is available:

    Total Production Time (PT): 1,240 minutes 

    Total Downtime (DT): 1.5 hours (watch the unit of measures)

    Number of Failures (F): 25

    Determine the MTBF:

    The first step is to determine the Uptime (UT) which = PT - DT

    UT = 1,240 minutes - 90 minutes = 1,150 minutes

    MTBF = UT/F = 1,150 / 25 = 46 minutes

    There is another method to represent MBTF which equate to the same result. 

    MTBF = 1 / Failure Rate 


    Failure Rate = the # of failures divided by the total uptime = F/UT

    The Failure Rate = 25 / 1,150 minutes = 0.02174 Failures / Minute


    Using the same information from above, determine the MTTR:

    MTTR = Total Downtime / # of Failures = 90 / 25 =  3.6 minutes


    As a GB/BB, you should examine the data in its entirety. Perhaps the mean does not represent the measure of central tendency

    Examine, every interval between failure. Each amount of time between each repair is one data point. And analyze is amount of time it took for a repair. Each time to repair is one data point.

    • If the data is normal, the apply the mean.
    • If the data is not assumed normal, then the median or mode may be more appropriate.

    This analysis may also indicate outliers such as a period time that was unusually long or short between failures. Perhaps the team can brainstorm the reasons why using the 5-WHY tool?

    • Was the repair done differently?
    • Was the repair done be a different person or group of people?
    • Was a different part(s) used?

    This can shed light on best practices or components that should be used again for a closer Design of Experiments (DOE) to find the optimal combination or best procedure. 

    Maybe a little more money up front to use quality parts or a longer PM to address more inspections/repairs is worth their weight in gold in the long run. 

    Perhaps, a minor increase in the MTTR equates in a significant increase in MTBF. The team will have to determine if this is acceptable.

    Remember the goal of Six Sigma, is not just to shift the mean to a more favorable outcome, but to make the performance more reliable and others words with minimal variation!

    Free Download - MTBF and MTTR Calculator

    This will be downloaded in a .zip format. Click on the thumbnail picture on to the left.

    A extractor such as WinZip is required to unzip the package. Winzip can be downloaded for free here.

    How does this relate to OEE?

    Recall that OEE is made up of the product of:

    Performance * Availability * Quality

    Availability is the amount of time the machine is available to run as scheduled.

    Availability is the unit of time the machine is available to run divided by the total possible available time. This metric does not include any performance numbers relative to how the machine runs while it is running.

    AVAILABILITY = Operating Time / Planned Production Time

    A 30 minute scheduled interval to replace a belt is much better than a 40 minute unscheduled interval to replace a torn belt that could tear and rip apart an oil line or result in other unintended consequences.

    Assuming the belt replacement has been studied and the proper interval for useful life has been predicted (in other words, not over-changing and spending too much money and time or excess belt replacements), then a scheduled event is obviously more predictable and favorable then hoping and not knowing when the next failure will take place. 

    A scheduled event such as a PM, break, safety meeting, Gemba walk, is NOT in the denominator and does not penalize the metric of AVAILABILITY.

    Again, the team should also try to minimize these "planned" events to try and get the machine(s) more time to be utilized. But this affect Utilization which is different than the metric of AVAILABILITY (go to the OEE page to learn more).

    As related to the metrics above:

    AVAILABILITY = MTBF / (MTBF + MTTR) for Planned Production Time

    An unscheduled belt change would be in the figure of Planned Production Time; however, a scheduled period of downtime (again the schedule downtime should be minimal and strategically determined) would not be in this figure of Planned Production Time.

    Return to the IMPROVE phase

    Templates and Calculators

    Search Six Sigma related job postings

    Browse the store

    Return to the Six-Sigma-Material Home Page

    Recent Articles

    1. X-bar & R Chart

      Sep 12, 16 08:17 PM

      Using a an X-bar & R Chart to plot continuous data

      Read More

    2. Data Classification

      Jul 17, 16 12:12 AM

      Proper data classification is necessary to select correct statistical tools

      Read More

    3. 7-Wastes

      Jun 22, 16 07:13 PM

      Description of the 7-Wastes, also called Muda

      Read More

    Click for a Password

    Six Sigma
    Templates & Calculators
    $14.95 USD

    Six Sigma Modules

    The following presentations are available to download

    Click Here

    Green Belt Program 1,000+ Slides
    Basic Statistics
    Process Mapping
    Capability Studies
    Cause & Effect Matrix
    Multivariate Analysis
    Central Limit Theorem
    Confidence Intervals
    Hypothesis Testing
    T Tests
    1-Way Anova Test
    Chi-Square Test
    Correlation and Regression
    Control Plan

    Six Sigma & Lean  Courses

    Agile & Scrum Online Course