Are You Down With OEE?

By Robert Minardi

Monitoring individual machine performance is like counting calories. We all know we should do it, but few of us really enjoy it or put much thought into it.

Let me begin with an analogy. There are three main dietary macronutrients: protein, carbohydrates and fat. Each of these have their place in a balanced diet and each contain calories. I’ll assume we all know what a calorie is, or at least the concept. Now, imagine if I told you I was going on a diet, and the goal for my daily intake was 2000 calories.  Sounds reasonable, right? But what if I told you I was only going to count my protein intake towards my daily caloric goal? You should be making a strange face right now, because that doesn’t make sense. If I eat foods with either of the other two macronutrients, they’re going to contain calories that need to be accounted. (By the way, if anyone can find a legit diet where I don’t have to count the calories in Chubby Hubby ice cream, please, contact me as soon as possible.)

While this seems obvious, this strange accounting is essentially what we do when we base our equipment performance exclusively on a daily production number. It’s hard to look at a single number at the end of the day and determine why our machines performed the way they did. For example, what does running 130 jobs on machine X really mean? Is that the whole story?

Let me introduce you to a metric called OEE, which stands for Overall Equipment Effectiveness. It considers three factors: availability, performance and quality. It’s a little tricky and should be used judiciously, but it’s also a great tool for getting the most out of your equipment.

Let’s take a look at the formula and its components. For our example, we’ll use an automated freeform lens generator. I’ll keep the units for the calculations in hours or fractions of an hour, but you can use minutes if you prefer.

The formula for OEE is as follows:

OEE%=availablilty×performance×quality×100

In this equation, availability is defined as the amount of time your equipment actually runs, in relation to how long it’s supposed to run, as depicted here:

Table 1 below provides an example of how to document equipment “availability:”

Table 1

Variable

Amount

Reason

Total production time

8.5 hours

Because the boss says so

Scheduled downtime

1.0 hour

One 30-minute lunch and two 15-minute breaks

Unscheduled downtime

20 minutes

Generator had a diamond break (one third of an hour or 0.33 hours)

Unscheduled downtime

10 minutes

Accumulated downtime due to swarf needing to be cleaned out of collet (one sixth of an hour or about 0.17 hours)

Unscheduled downtime

15 minutes

Accumulated downtime due to coolant low pressure error (one quarter of an hour or 0.25 hours)

Unscheduled downtime

5 minutes

Conveyor belt jammed and had to be cleared (one twelfth of an hour or 0.08 hours)

The formula for the availability percentage for this data would look like this:

Availabilty%=((8.5-1.0)-(0.33+ 0.17+ 0.25+ 0.08 ))/7.5≈0.89×100≈89%

Notice the scheduled run time (denominator) is 7.5 and not 8.5. That’s bause breaks and lunches are scheduled. If you have a company meeting, or scheduled maintenance on a given day, you’d remove that from your “scheduled run time” as well because the equipment isn’t supposed to be running. All other downtime is considered unscheduled. The purpose of calculating your equipment’s “availability” is to discover why it isn’t running when it should be. A log should be kept with a short blurb explaining each instance of one minute or more of downtime.

Performance is the speed the equipment runs in relation to how fast it was designed to run. It answers the question: “When the equipment runs, is it running as fast as possible?”

Table 2 shows the performance of our freeform generator:

Table 2

Variable

Amount

Description

Total count

152 parts

All parts produced including defects

Scheduled run time

7.5 hours

From “availability” calculation

Ideal run rate

22 per hour

Ideal output in same units as “run time” (in this case hours)

In this scenario, performance would be calculated as follows:

Performance=(Actual Run Rate)/(Ideal Run Rate)

Performance%=[((Total Count)/( Scheduled Run Time))/(Ideal Run Rate)]×100

Performance%=((152/7.5))/22≈ 0.92×100≈92%

Now let’s take a step back here, because this is the tricky part. OEE is meant for equipment with constant cycle times—for example, a machine that stamps out widgets of equal specifications. Most machines in an optical lab don’t run like this. Freeform generators vary their speed based on things like material, cylinder power, prism, etc. A 0.25D SV in CR-39 will take less time than a polycarbonate digital progressive with a -5.00DC near zone. At least on the equipment I’ve worked with.

Also, we run work “first in, first out,” so we can’t batch up jobs with similar attributes and run them all at the same time. We do a continuous flow of unique jobs with unique cycle times, which we can’t easily account for. So, what do we do? We can fudge it a bit by using the fastest known rate for our ideal run rate and keep in mind that it’s just an estimate. Alternatively, I’d recommend you do a time study every few months across material types and lens designs to gauge machine performance. 

Ironically, performance in the form of machine speed is the variable many tend to focus on. We often ask ourselves, “How can we make our machine run faster?” The reality is, it’s the variable we can do the least about. The machine is optimized for performance when it comes from the manufacturer and there’s no re-tooling or adjustments we, as operators, can make to increase its speed. All we can do is make sure the machine is fed and produces quality work. Which leads me into the final OEE variable: Quality is simply the number of jobs that were good on their first pass, in relation to the total produced.

Table 3 records the quality performance of our generator:

Table 3

Variable

Amount

Description

Breakage

1

Lens too thin

Breakage

1

Incorrectly scanned; cut wrong power into lens

Rework

4

Not finished due to diamond break; had to rerun jobs

Good Count

146

Jobs that were not defective or in need of rework

Total Count

152

Total jobs produced

Calculation of quality in the above example would look like this:

Quality%=[(Good Count)/(Total Produced)]×100
Quality=146/152 ≈0.96×100≈ 96%

Putting all three of our variables together, we end up with the following:

OEE%=(0.89×0.92×0.96)×100≈79%

I know what you’re thinking: “What the heck? How did we end up with a 79% overall OEE if we’re in the 90% range for the individual categories?” That’s just the way the math works out. It sounds low, but to put it in perspective, 85% is a world-class number; with 60% to 65% being typical. From our example, we can see that “availability” is where we need some work. If we relied on our daily production number exclusively for that piece of equipment, all we’d know is that it didn’t produce as many jobs. Now we know exactly why.

On a side note, don’t fall into the trap of “robbing Peter to pay Paul.” What good is high availability if your quality is poor or vice versa? You don’t want to trade quality for production, as in the example below. In this case, we took steps to boost availability, so the equipment ran longer, but the quality score dropped. The net gain in OEE came at a price of more breakage!

Admittedly, this metric shouldn’t be used all the time. There’s a lot that goes into it. Until OEE is baked into the LMS, I recommend calculating OEE on new or underperforming equipment to get the most benefit. Also, it’s very much an individual metric. Meaning, you can’t compare a generator OEE to a polisher OEE. Nor can you compare your machine’s OEE score to another machine; even if it’s the same make and model. This metric is for comparing a machine’s overall effectiveness to itself and constantly working towards improving it.

Remember an OEE score is just another number. The power of this metric isn’t in the final score, but the factors that create it.

Robert Minardi, ABOC, has been in manufacturing for almost 25 years. He’s a certified Lean Six Sigma Black Belt with a background in quality control.


CURRENT ISSUE


Labtalk-November/December 2017