Using FMEA to determine Risk in minimising Testing

T

topstriker

in the past we have allowed our electronics team have control over test. Although this may seem the simplest solution it has lead to a bank of test, ICT, PCB and post assembly that eat up NRE, operators and lead time.

I am attempting a new approach using FMEA. The electronics guys normally start with a test specification of all tests they want to carry out before developing the appropriate test plan. At this stage i plan to carry out and FMEA on the test spec. We will use several factors to calculate the RPN. We will consider if future tests will carry out the same fail. I think to determine the risk of fail i need to consider the number of opportunities for failure. I think maybe the best way is to look at the number of components measured through the test and score the test on the number of devices, type of devices and legs on the devices.

If anyone can suggest any other ideas worth considering how to grade the test then that would be great

Paul
 

Marc

Fully vaccinated are you?
Leader
It's the weekend and I know things are slow, but I figure I'll "Bump" this in case someone can help. My Thanks in advance!
 

Jim Wynne

Leader
Admin
in the past we have allowed our electronics team have control over test. Although this may seem the simplest solution it has lead to a bank of test, ICT, PCB and post assembly that eat up NRE, operators and lead time.

I am attempting a new approach using FMEA. The electronics guys normally start with a test specification of all tests they want to carry out before developing the appropriate test plan. At this stage i plan to carry out and FMEA on the test spec. We will use several factors to calculate the RPN. We will consider if future tests will carry out the same fail. I think to determine the risk of fail i need to consider the number of opportunities for failure. I think maybe the best way is to look at the number of components measured through the test and score the test on the number of devices, type of devices and legs on the devices.

If anyone can suggest any other ideas worth considering how to grade the test then that would be great

Paul

While I understand the desire for some sort of objective scale or values to apply to guide decision making, FMEA is always going to entail some unavoidable level of subjectivity. The ideal is to minimize subjectivity by use of data and experience, but subjectivity never really goes away, and Murphy is almost always lurking somewhere in the shadows, seeing things that you don't.

Rather than "grading" as you describe it, I think the best thing you can do is (given an understanding of process variation) determine the most likely failure modes and then the risk associated with each one, and plan your testing and prevention strategy accordingly.
 
T

topstriker

one thing to our advantage is a current product in production that has ran over 100K units with high yields. My Remit is to reduce test by 90%, a tall order and very aggressive. To achieve this i need to view how the customer uses the product and ensure our testing is around that interaction.

I need to determine the best way of scaling likelihood of failure. 1 option is to look at each test, determine how many components are tested to look at the opportunities for error. We can also put some weighting on the different types of components.

If we were making $500 iphone i would be a lot more conservative but instead we are making $40 items generally given away free by our customer
 
T

The Specialist

I agree with what Jim Wynne has written.

I also agree with the FMEA approach to failure management (through predictive measures).

What will assist you in this process is historical data from process and end-user reported failures.
You know your processes and the components of the product, and the most likely failures, through experience. This is what ultimately adds value to the FMEA process.

Ivolve plenty of (neccessary) people in your FMEA... Better than going it alone!
 
T

topstriker

thanks guys for your comments. reviewing test and customer returns, we see little evidence of production issues, like most products we see firmware hitting us and generally with issues not apparent during production. My remit has always been that we are testing to prove production quality not the design.

This FMEA will involve the electronics guys who also write the test spec and then get our firmware guys input. it looks like a good challenge to be taking on and maybe the best approach will be an aggressive outcome taken with conservative approach. I will work with the current product, set the unwanted tests to record only then see if the are captured at latter stages of production
 

Paul Simpson

Trusted Information Resource
Paul, Sorry for the delay from my part. I've made a couple of points against your original post. There's nothing in the replies I've seen that I disagree with but I start from a different place - :D. Anyway :2cents:
in the past we have allowed our electronics team have control over test. Although this may seem the simplest solution it has lead to a bank of test, ICT, PCB and post assembly that eat up NRE, operators and lead time.
There is a positive with the electronics guys 'owning' test - generally the buy in you get for quality is better. Any changes you make should be through working with them so you don't end up throwing the (engagement) baby out with the (cost of test) bathwater.

I am attempting a new approach using FMEA. The electronics guys normally start with a test specification of all tests they want to carry out before developing the appropriate test plan. At this stage i plan to carry out and FMEA on the test spec. We will use several factors to calculate the RPN. We will consider if future tests will carry out the same fail. I think to determine the risk of fail i need to consider the number of opportunities for failure. I think maybe the best way is to look at the number of components measured through the test and score the test on the number of devices, type of devices and legs on the devices.
Great. This is what FMEA is intended for. We are talking about a process FMEA, right? The more data you can gather the better.
  • You need to decide on severity. So if a faulty product gets out into the market what is the impact on the business. Any cost of quality / warranty information will help here. Severity rating will depend on the type of fault and should drive the risk appetite decisions (at the end).
  • Also start gathering data on likelihood of occurence of process failures. Hopefully you have plant data that says how many of each type of failure you get at which stage of the operation. You also should have data on how and where the fault is found in the process. This data gives you a good picture for likelihood of occurence and some information on likelihood of detection.
  • Then have a look at what is being found at the various stages of testing you are investigating. This gives you the rest of information on likelihood of detection and will tell you how effective these testing processes are. Bear in mind that if you do away with all testing then all the failures you currently pick up are likely to get out to the market.
  • Decide on your risk appetite. You made the point about $40 giveaway vs. $500 iPhone. This will help to decide whether you remove any testing.
Once you know the statistics of failures within the population you can model the impact of a reduction in sampling at the end of the line and predict how effective a reduced sampling plan will be in picking up faults - there are standard tables for this.

Good luck and let us know how you get on.

If anyone can suggest any other ideas worth considering how to grade the test then that would be great

Paul
 
Top Bottom