Product Development

Hello,

I work for a company that manufacture hydraulic pumps and motors. I have only recently joined the business and we are due to launch a series of products.

I have been asked to think of a methodology we could use to determine the full range / capability of this new product family which would give us confidence that the product was also reliable.

There are 3 types of motor.

I have looked at HALT and HASS as well as MEOST and i'm not sure if they apply directly to what i want to achieve.

What i was thinking was to produce 10 of 1 type, test each of them 3 times and using the 30 sets of data per parameter, gain an insight into the product capability across a batch but also look at the variation within a motor.

I'm confused as i'm not so sure on which way i should go with this and im hoping you would all be able to provide some advice.

If i haven't explained this clearly then please let me know and i will try and provide as much information, as clearly as possible.

Cheers

Scott

**just realised this may be in the wrong place**

 

I recommend you start with a very clear, preferably quantitative definition of what you'd like to determine. "determine the full range / capability of this new product family which would give us confidence that the product was also reliable" is quite vague and too general.

 

Hi Scott,

I can see that you are talking about two Quality attributes of your product

1. Capability- The ability of the product to satisfy the specifications.
2. Reliability- The life profile of the product

Capability

Testing a batch of 30 motors to evaluate the capability sounds like a logical thing to do. However, if you have 3 types of motors which are different in design, the capability for each design is likely to be different. I would thus want to test 30 units of each type of motors to establish the capability of that type. Also Capability of each design of motor will need to be tested for multiple CTQs e.g. Torque, Power consumed at rated rpm, HP at the output.

Capability evaluation would entail the following steps:

1. Test 30 motors and collect raw data for each CTQ
2. Plot a histogram of the raw data
3. Superimpose specifications on the histogram.
4. Calculate the Capability indices.

The above Capability evaluation is more often used to evaluate Process Capability. It can be used to calculate the extent to which a product is likely to comply to the Specifications. The spread in the plotted histogram and the measures of dispersion in it will give you an indication of the variation within the type of motors that you are looking for.

Reliability

If you are performing Product Reliability Characterization, Reliability will be needed to be resolved by the multiple failure modes that the product failure is governed by. Your looking at HALT HASS and MEOST suggests that you are looking at Accelerated Life Testing. IMHO Accelerated Life Testing requires a detailed prior insight into aspects like

• The Physics of failure for each failure mode.
• The impact of acceleration factors / higher stresses.
• Whether the failure mode is valid across the domain of acceleration factors.

I believe that Accelerated Life Testing ought to be done only when you are tweaking the design and want to quickly evaluate the impact the change has induced in the life profile of the product. For exploratory analysis of Reliability testing at normal duty cycles works the best.

It is likely that the normal duty cycle testing is a long drawn one and you do not really have the lead time required for a life test.

In such a case you could look at the field failure history of similar designs. The failure of a motor is really attributed to the failure of components like the armature coil, bearings and seals. Every design usually has a carry forward element of similar designs working in the field. It is possible to capture field failure data for components with similar designs and determine the Life Characteristics of such components ( Weibull plotting is a popular method for this. I have also seen the use of Crow-AMSAA models towards this).

Once you have the Life data for individual components (in terms of say B10 life resolved failure mode wise), you could use statistical techniques or Monte Carlo Simulation to predict reliability of your product.

Hope this discussion helps towards Product Capability and Reliability evaluation that you are looking at. These are a few suggestions that came to my mind. There will be interesting technical aspects and views on this from the others. I look forward to comments on this topic from the other forum members too.

 

Very true, Bev!

Additional on testing for (controlled) parameter extremes, something important to keep in mind is noises of all sorts. In medical devices validation there's an expectation to account for noises in production and use. The very basic tactic is including samples from multiple (normally at least 3) production runs, although many manufacturers actually circumvent the underlying intention by running 3 "lots" in straight succession, merely stopping and starting the next day (or the next hour...). The more variability introduced, the better - raw material batches, day/night production, older/newer equipment, even seasonal changes if relevant, etc. etc.

Including noises from the use environment depends on whether a "nominal" performance or an actual one is being established. In medical devices use environment must be fully accounted for, however such approach will inevitably increase the resources required significantly.

Cheers,
Ronen.

 

I completely agree with you Bev that the samples selected need to be true representatives of the process (I would say product range in this case), and that the small sample may not adequately represent such a range. This variation is what Ronen is referring to as noise. Thanks to both.

Trying to understand the directed studies, you have explained that the input parameters need to be deliberately set to highest and the lowest (extremes). Would the input parameters that you talk about mean the contributing Product Characteristics?

For example the Quality characteristic of ‘motor vibrations’ would depend on input parameters like Runout of the shaft and grade of the bearings. The deliberate setting of the parameters would mean testing for vibrations with all grades of bearings and with inducing maximum possible Runout in the shaft.

Since the number of input parameters and the settings of each would sum up to a substantial number of combinations, you would need to configure a kind of DOE and test the product for all combinations. The study would then really be a robustness study that could indeed be done with small sample sizes and still capture the capability of the product function under all circumstances. I do see the point.

Thanks for your insights.

 

Is there a need for a DFMEA or a Risk Assessment? Sorry if I am qsking the wrong question.

 

OK Thanks.

 

you are 100% correct.

We have carried out a DFMEA and are now looking to prove out what we have done on prototype models. The plan is to maybe use Response Surface Methodology or similar (potentially some of your recommendations) to prove out the true spec/range of the product and then create a report to provide our customers and data to input into our product catalogue.

Following the testing we will look to amend the DFMEA if required.

 

Might anyone have sample templates for a robust project checklist for product design & release in a medical device manufacturing entity? Possibly something going thru the various stages thru final release?

My present project site has a variety of disjointed checklists scattered throughout the existing protocols. We are planning on the introduction of a typical Control Plan format after their Pfmea. A good, inclusive staged "checklist" would be another good mechanism to foster a more standardized product design approach and level of review.