Hi,

Since I am quite new to this subject, I would ask you to look in to the following and give your opinion on my approach taken, if anybody has made comparable analysis, please provide me with any tips available!

I am analysing failures of our packaging machines. All failures (machine stops) are recorded in a database; amongst others failure location (sensor) and time are recorded. I have grouped the sensors per sub machine (packer, foil unit etcetera).
In order to make a comparison between different packaging machines and define stronger and weaker areas per machine, I want to calculate MTBF´s per machine, per sub area.
After selecting a period of time, my tool counts the number of failures per sub machine in this period.
Can I take the total duration of the selected period (in seconds or minutes) and divide it by the number of failures; would this result in a useful MTBF?
Can I then draw conclusions like: MTBF of packer area of machine A is higher than MTBF of packer area of machine B, machine B has a problem?

My concerns:
- Duration of machines stops are not taken into account in this approach. Should they? Or is better to use the above stated MTBF method for a quick and dirty analysis and then dig deeper into problem areas by analysing the duration etcetera?
- How to determine a suitable period for the analysis? How long should this period be to have results that are statistically relevant?

As said above, any answers or remarks are welcome!

Cheers,

Daan

Hi Daan,
Welcome to the Maintenance Forums.

You need to include the time your equipment is unavailable due to maintenance or in a standby status. Therefore, your Mean Time Between Failures (MTBF) should be:
Reliability = Total Hours - Downtime Hours - Standby Hours / Number of Failures

Probability a better marker in your case would be Production Efficiency, which is the ratio of actual output from a machine (which meets the required quality standards) to its rated output, during the time that it is operating:
Production Efficiency = Actual Production / Total Hours - Downtime Hours - Standby Hours / Rated Capacity (Units per Hour) X 100%

Regards,

Larry,

Thanks for your quick reply.

I see your point in calculating the Production Efficiency. My concern is that Production Efficiency alone does not give me enough information. It tells me that a machine is running good or bad, but does not give me an indication of the problems of an underperforming machine.
Therefore I thought of calculating the Mean Production Between Failures, so the average good products produced between failures. Together with the MTBF this would give me a good overview of the performance of our machines, and it enables me to compare the performance of the different machines.

The failures in our machines as I described them are operational failures (paper jams, glue problem etcetera) and are normally of short duration. When a bigger failure (breakdown) occurs, our time registration is switched from “production” to “maintenance”; I do not take this time into account in the MTBF calculation.

About the MTBF calculation: how do I handle the fact that I have multiple failure types per machine?
For example failures F1, F2 and F3 occur 8, 3 and 12 times respectively in a production period of 24 hours. Is the calculation than as easy as: MTBF F1 = 3 hours, MTBF F2 = 8 hours and MTBF F3 = 2 hour2?

Regards,

Daan

Daan,
First let me answer your previous question: -How to determine a suitable period for the analysis? How long should this period be to have results that are statistically relevant?”.
You should pick a period at least a few times the MTBF. That means if you have a MTBF of 8760 hours then your observation period would need to be, for example, 4 X MTBF = 35,040 operating hours. Obviously grouping like-equipment with the same duty cycle and operating environment will reduce the calendar time. 10 pieces of equipment = 3,504 hours, 100 pieces = 350 hours, etc.

For reliability discussions, you should look at the aggregate of failures. The failure rate of complex equipment will be the collective rate of each individual failure mechanism.
So your MTBF = total elapsed time / total number of failures.

R,

Larry has a good point on the suitable period for MTBF analysis. I notice that you have not taken duration of production stops or machine failures into account. I think we should because it will give a bigger view because MTBF=MTTR+MTTF where
MTTR= Mean Time To Repair to measure maintainability/servicability
MTTF= Mean Time To Fail to measure availability in eg running hours
MTBF= Mean Time Between Failure to measure reliability
Tq