Q: Our machine can run mixed output, so several different products, each shift. Till now we just recorded the time the machine was running and the output we had. Our software now requires that we register activities and quantities to account for every minute in a shift. But we are not tracking the number of minutes it ran each product.
Arno Koch • NOT your software requires this! It is required to get an accurate OEE! And thus that is how it has to be done in the software…
Running Mixed Output
If you run mixed output, so different products each shift, you might face up with products having different maximum speeds within that shift. So to know how much product you could have made (the target for a production run), you need to know how many minutes the machine ran at what maximum speed. By calculating the target output for a run, and comparing it to the actual output for that run, you are calculating a performance rate for this particular run. If you do this for every run in a shift, the performance rate is a weighted average of multiple runs (no matter the different maximum speeds per product!).
If you would just add all output and all production time, you lose the weightfactor! It LOOKS as if you get a correct OEE, but it isn’t. Many of the home build systems I saw had this algorithm error.
Example
Let’s take an example: a mouldingpress. It is being operated with two different dies: one die for a large part, making one part each stroke and one die for a small part making 4 similar parts each stroke.
The press can make two strokes a minute. So the standard for the large part is 2 parts per minute and for the small part it is 8 parts per minute (2 strokes x 4 parts) Just for the example we assume changeover is timeless and everything else is perfect.
Let me show you two different days;
First in the way as you describe:
‘… we just recorded the time it was running and the output we had.’

Day 1 (480 min) 
Day 2 (480 min) 
Time Running 
480 min 
480 min 
Output 
2340 
1260 
Obviously day 1 was your best day.
Now, can you tell me what your OEE was on day 1 and on day 2?
Lets see how you can solve this (and how good software should do…)

Day 1 
Day 2 
Minutes Part Large 
120 min 
360 min 
Max. Output Part Large 
(120×2=) 
(360×2=) 
Actual Output Part Large 
180 parts 
540 parts 
Minutes Part Small 
360 min 
120 min 
Max. Output part Small 
(360×8=) 
(120×8=) 
Actual Output Part Small 
2160 parts 
720 parts 
What was your best day?

Day 1 
Day 2 
Availability 
100% 
100% 
Performance 
75% 
75% 
Quality 
100% 
100% 
OEE 
75% 
75% 
The point is: To calculate the performance rate, it is necessary to know what the EXPECTED output was versus the ACTUAL output. In a mixed production this means you’ll have to know how long you ran what product since that is what determines the expected output.
Let’s make it a little more difficult:
Parallel Output
Some machines, like presses, can produce different products in one stroke.
Imagine a press that can have two different dies in one stroke (so you could have a changeover of half a press…)
So in one shift the press can first produce 1 A and 4 B parts, after that 2 C and 3 D parts
and next run 2 C and 8 E parts each stroke. Dazzling???
Your software should be able to handle this situation seamlessly due to the mechanism as described above!
Q: We have a similar machining environment and what we do is the following:
If we have 4 variants ad running at ideal cycle times x1x4, planned for quantities a1a4, and in reality what got produced was b1b4, then performance is calculated in terms of total time as
(b1x1 + b2x2 ….) / (a1x1 + a2x2 …)
Thus it kind of gives a weighted indication of what the performance is, without skewing it like an average.
Awaiting your opinion on this, and whether there is a more accurate method of doing it.
Arno Koch • Indeed, the way to calculate a correct performance rate in a mixed output situation is to calculate a weighted average expressed in time.