Reliability Testing Protocol on a new Peristaltic Pump

I am planning to conduct reliability testing on a new peristaltic pump.

can somebody help me with a template of reliability testing protocol.

Sample size needed for testing in regular environment

sample size needed for accelearated testing ...

Re: Reliability Testing Protocol

Did you mean "... in a regulated environment"?

Medical Device regulations may have specific requirements. I am not from that industry, so I cannot speak to that. However, I am familiar with reliability. The following is the process for determining the sample size.

1. Determine the type of reliability testing you want to perform (i.e., demonstration, estimation, accelerated).
2. Specify the desired reliability level at a specified time.
3. Specify the desired confidence.
4. Specify the assumed distribution (e.g., Weibull) and estimate the critical parameter (e.g., Weibull shape).
5. Calculate the required sample size.

I was thinking about executing both demonstration and acceleration tests....

For demonstration I was thinking about 85% reliability and 95% confidence...it comes to around 19 pumps... But I may have hard time selling this testing with 19....I am thinking about running these pumps simulating atleast 5 years of usage ... And determine the reliability based on mttf or mtbf....I am guessing pump life cycle data would follow weibul... If it does where I can find shape and scale factors for this pump.

Sample size for accelerated testing ...may be 5 pumps ...after temperature, vibration and humidity exposures ... I am thinking about running these pumps simulating 2 years of usage ...if i find any failures.. it would be hard to know if this failure may occur at regular conditions....

Should we also define failures ....

Acceptance criteria .. Etc ...with reliability testing I may see some un expected failures .... So I have to b careful...

Cognizant said:

I was thinking about executing both demonstration and acceleration tests....

Click to expand...

You can also do an accelerated demonstration test by applying an Acceleration Factor to the demonstration time.

Cognizant said:

For demonstration I was thinking about 85% reliability and 95% confidence...

Click to expand...

These are reasonable values provided they are typical and acceptable in your industry.

Cognizant said:

it comes to around 19 pumps... But I may have hard time selling this testing with 19....I am thinking about running these pumps simulating atleast 5 years of usage ...

Click to expand...

You can reduce the quantity by extending the test time. Also by reducing the confidence level.

Cognizant said:

And determine the reliability based on mttf or mtbf....I am guessing pump life cycle data would follow weibul... If it does where I can find shape and scale factors for this pump.

Click to expand...

MTBF and MTTF are based on the exponential distribution (Weibull with shape factor = 1). I don't recommend it for two reasons: 1) It is usually misunderstood. By the time you reach the MTTF/MTBF 2/3 of the product has already failed. A reliability statement or an L5/L10 life is better; 2) It assumes that all failures are random and are due to overstress conditions. What do your failures indicate? Are they due to latent manufacturing defects, overstress or wear out?

Cognizant said:

Sample size for accelerated testing ...may be 5 pumps ...after temperature, vibration and humidity exposures ... I am thinking about running these pumps simulating 2 years of usage ...if i find any failures.. it would be hard to know if this failure may occur at regular conditions....

Should we also define failures ....

Acceptance criteria .. Etc ...with reliability testing I may see some un expected failures .... So I have to b careful...

Click to expand...

You should definitely define a failure. Is it total loss of primary function? A degradation in primary function? Loss of secondary function? Noise?

A computer example: Your laptop gives you the blue screen of death 10 times a day? However, you can reboot and the computer works. Is that a failure? What would you customer say? What if instead of this mode, you lost wireless communications ten times per day and had to reset the wireless. What if the cooling fan emitted an audible hum? Or the screen intermittently dimmed then brightened?

One does need to consider the risks involved in failure. Is this a medical pump implanted inside a person's body? If it fails does the patient immediately die? Your proposed 85%/95% would be fine if it is a pump external to the patient or is simply a non-medical pump.

As alluded to above the exponential distbribution can be a dangerous assumption to make, without verifying that observed failures are indeed exponentially distributed. Again, in a low risk application, the exponential assumption may be fine for initial testing, but you would need to observe the full lifetime in real world use to see if there is a burn-out phenomenon.

Also, what is the proposed preventive maintenance regimen for this pump? That can affect burn-out (end of life) failures.

Thanks all for your responses..

Pump is not implantable .., it supplies saline to the device during treatment ..,

In order to test this for life I may have to run long time ...

Also for ALT ... Using Arrhenius equation.., How can I find the activation energy for this peristaltic pump.,,

Cognizant said:

In order to test this for life I may have to run long time ...

Click to expand...

Yes, that is definitely a challenge in reliability analysis.

What folks end up doing is either

analyzing the life of individual components within the mechanism, and then model what that gives for the system life

or

just run several items on the test bench in parallel and assume exponential as good enough, and give failure rate as total failures seen (not every item has to run to failure) divided by the total unit runtime. So if I run 5 units for a year (with replacement of failures) and have 3 failures, I assume this is the equivalent of running one item for five years and getting 3 failures.

Thanks Miner/Steve for your responses ......

Usually exponential is for electrical components (resistors, capacitors, motors etc) .....since this pump is mechanical pump with roller blades and is driven by the motor ... which distribution is more appropriate .....Weibull or exponential ?.......

also in additional to the demonstration testing with no stresses .. should I also initiate accelerating testing on another sample ...I amy have hardtime finding samples for both tests ...

is it worth combing both tests ..... do accelerated demonstration test ....

Mechanical failures tend to be wear out mechanisms with potential early failures from latent manufacturing defects. The Weibull distribution is often a good fit. Slow wear out will have a shape parameter > 1 but < 4. Rapid wear out will be > 4.

Btw it is no longer true that electronics always follow an exponential distribution. It is becoming more common to see electronic wear out failure modes.

I am planning on running accelerated life cycle test with three different variables, temp, RH, speed of the pump. Is there any equation for this to calculate acceleration factor ...

Also I m planning to keep the pumps in chambers (accelerated temp 50c, 85%RH) for x hrs then take the pumps out of chamber and run the pumps for x hrs at room temp at max speed (observed in the field).... Is this the right approach or should I run the pumps while hey are in chamber ? I