Monday, May 2, 2011

Machines Monitoring

I have been working as an engineer in EMS (electronics manufacturing subcontracting) since 97 and always I had a special fascination for all kind of machines/robots used to DO our job - from Screen Printers, Chip Placers, Fine Pitch Placers, Reflow Ovens, Was Soldering, Conformal Coating, In Circuit Testers, Functional Testers, etc.

All of them are capable to follow a program/recipe and spit out log files containing lots of useful information.
Many times we rush through our duties and DO NOT pay enough attention to the data.
Many tried to prove that this behavior is wrong, but at the end of the day - quantity prevails quality big times and accountants rule the industry, and LLR countries are luring accountants.

I used to say - you as an engineer have to 'explain' any new idea in $ otherwise is absolutely dead - nobody will give a s**t on it and is absolutely dead. Yes, if you have a great revelation, engineering one - try put it plain and simple for money people: ROI (spend now, recover in x months...afterwards ++ profit), because nobody is excited by engineering facts - money talks.

Enough idle talk - I always liked to have a real time 'vision' of what machines talk and interpret their chat and take advantage of this knowledge. SMT, Wave Soldering, ICT are in place since 30 years ago, but despite huge transformation of the machines hardware, manufacturing is mainly still done using pen and paper - best case tons of EXCEL.
Excel you would say it's a good tool, but requires lots of people getting data from archaic paper, filling out forms, copy paste formulas, do graphs and having the report done......MUCH TOO LATE...and then if you need the data tomorrow start again and having it done....again TOO LATE.
Data is valuable if is FRESH and can take use of it....detect trends, identify root causes.

One of the simplest things that I would like to have and I have implemented, is to know the Overall Equipment Effectiveness - OEE - old concept - handy to know it.
It will give you more understanding of your inefficiencies and uncover ways to overcome them.

OEE=A * P * Q

A - Availability - represents time when machines are available percentage from total scheduled time

P - Performance or at the other end Speed Loss - time lost with short stoppages - due to improper manning of the line or materials unavailability

Q - Quality - it measures line yield - percent of good boards from total production

What can you DO with this OEE?

Take the right decisions - if Availability is always an issue, because machines are always breaking down you don't need to buy more machines to make up for lost capacity - you just need to put together a better maintenance plan and a better spare parts kits.

If you have problems with Performance, then machines are OK, but materials are not available JIT (Just In Time) or you don't have enough people to cope with all change overs and machines stops.

Quality - IT IS IMPORTANT - nobody will pay you for scrap products or let's say non working units and repair is hard work, require a lot of skills, is expensive and tedious. Put enough emphasis to minimize the faults - will reduce your costs.

What's a good OEE?

I used to say 90% availability, 90% Performance, 98% Quality will give you a nice and robust target OEE= 0.9*0.9*0.98=0.7938.
If you get it and are able to maintain it then you are OK.

All placement machines are capable to spit out log files with data about when a PCBA entered a machine, when it went out and lots of info about errors on each board, each slot or each feeder. These data are an easy grab and knowing, real time your output and calculating OEE will give you the power to change just pray for the best and  get same s**t every day. You don't even know if people are not slowing down the line on purpose - if the Supervisor doesn't want to tell you (of course he knows).

It is quite the same situation for TESTERS - In Circuit Testers in special - right after you've tested a board yo want to know your defects, top defects, machine utilization, number of retests, access a debug database for more advise.....just looking at the failure ticket given by the machine you will not get all these.

Is quite a pity that big companies like HP, Genrad, IFR made awesome test machines, but so bloody dum from data collection point of view. Log files generated are comprehensive, but all the work is left on Test ENG to figure out how to use them properly, wisely.

Just parsing all data from a HP3070 by example you could get - machine utilization, real yield and shift throughput, number of retest, top defects, SPC (Statistical Process Control) for any particular measurement you want.

Yes, I developed all these Machines Monitoring Solutions because I was sick and tired to be blind-folded and DO my job just based on subjective assumptions.   

or visit our site: - Products page has a comprehensive description of our HPDC&A solution developed for ICT machines real time monitoring.

All the best,
Nick - from Nick's Software