Wow. Where to start?

I believe that customers in our very cost-driven industry (custom automated assembly systems) always look for the lowest possible price, disregarding factors such as cost of ownership. While there is nothing wrong with looking for a low price, I question the capabilities of many decision-makers to properly evaluate proposals and solutions and compare functionality, risk and maintenance.

Unfortunately, the predominant approach is that procurement departments don’t care about the solutions, as long as the requirements per spec are met on paper in the lowest quote. This leads in essence to the situations you describe, Jacques.

So, inexpensive fixtures do not necessarily cost more in the long run when done right. But, I do agree that poorly engineered fixtures will cost the client more than he cares to know—mostly inflicted by his own choice.

—Joerg Starkmann, sent via Linked-In, Automated Assembly Systems Group

Do you know the cost of accepting bad parts? Consult the ANSI NCSL Z540.3 2006 standard, which covers the technical requirements for the calibration of measuring and test equipment through the use of a system of functional components. Section 5.3, which covers the probability of false acceptance (PFA), dictates that the probability of accepting an out-of-tolerance part can’t exceed 2 percent.

Recognize that acceptance criteria should depend upon the consequence (liability, part cost, cost of subsequent operations, sorting, customer relationship damage, warranty absorption, etc.) of passing bad parts. Consider the following: If the part under test is at spec limit, there is 50 percent chance the part is out of spec. Shouldn’t the text fixture capture this possibility?

—Otto Geiseman, sent via Linked-In, Quality magazine group

It occurs to me that part of the issue with inadequate fixturing (whether cheaply or poorly designed) is the lack of appreciation for the role fixtures play in achieving the performance, quality, and cost targets for the product.

In more basic terms, I have found that some production fixtures are designed without consideration of correct or required assembly processes (e.g. positioning, pressure, timing, incoming material quality, etc). In addition, the capability of the fixturing to produce consistent results is either not measured or not appreciated. The results of bad or unacceptable fixturing might be noticed during in-line or end-of-line testing. Unfortunately, it may not get noticed until the customer sees the product.

I am inclined to think that such fixturing problems are, at least to some degree, a result of a design process that does not fully disclose the sensitivity of the product or assembly to assembly variations. In an ideal situation, engineering and production prototype testing ought to uncover such “potential to fail” process steps. I suspect, however, that “expediencies” in manufacturing are a result of insufficient “what if” analysis done during the design and prototype phases of a product evolution or that they are ignored in favor of a “good enough” policy in production or a lack of appreciation for the real cost of quality.

Obviously, the costs associated with fixing a problem at the end of a production process are way higher than the costs associated with identifying and mitigating such problems early on in the design evolution.

Of course, based upon your question, this is not so obvious to enough people. In fact, the countryside is strewn with products that failed simply because the assembly processes (including fixturing) did not reflect initial expectations.

So I agree that fixturing issues can lead to product failures. My experience tells me that it is up to the design and production and quality engineering staff to sort out what assembly processes are acceptable relative to stated targets.

—Ron Scicluna, director, PPI Consulting Group, sent via Linked-In

The cost of the fixture does not guarantee that it will or will not produce consistent results. The issue with fixturing is making sure it addresses all relevant material processing requirements. It is important to be able to distinguish between bad fixturing and bad materials going into the fixture.

As for test fixturing, inconsistencies here can obviously taint performance test results. The accuracy of a test fixture needs to be established at inception, as well as during its use. “Golden units” are especially helpful in this regard. Cutting corners here can cause major downstream issues, not to mention biasing acceptance data.

In addition, cost cutting when conducting a root-cause analysis can lead to misleading results, leading to ineffective remedies followed by more failure analysis. All this results in a product that costs more than expected. These costs are measured not only in unanticipated engineering or material costs, but also in costs associated with missing schedules, missing performance or reliability targets, and missing customer expectations.

Other cost-cutting actions with the potential of costing more in the long-run include:
* changing parts suppliers or assembly vendors.
* tweaking approved processes (e.g. injection molding settings) partway through production without sufficient validation.

—Posted by Ron Scicluna, director, PPI Consulting Group, sent via Linked-In

Otto’s remarks raise an interesting point. The closer a part dimension is to the spec limit, the more critical the true accuracy of the test fixture becomes. Not just repeatability or resolution. A bad gauge may repeat a bad measurement wonderfully over and over again. Conversely, the better the CPK is for a process, the less critical that minor test fixture error becomes, assuming that testing independent of the test fixture in question has properly established the true mean and CPK values for the process.

—Posted by Dan Smith, owner, FixLogix LLC, sent via Linked-In

I agree with your comments, Dan. The most important aspect of your response is the notion that independent testing is employed to establish the true mean and CPK of the process. Do you imagine this to be a “normal practice” in the product development cycle?

—Ron Scicluna

I have to ask what is the definition of a test fixture. In my world, a fixture is just that. It holds a part to be tested. If you are using something to test a part then it becomes a gauge.

If the fixture is to hold a part for testing, then there are multilple questions. How repeatably does the fixture hold the part? Can multiple operators repeatably place the part in the fixture? And, is the measurement process robust enough to measure the part and pass a gauge R&R?

In response to expensive gauges—and lets assume a custom gauge that fixtures the part—expensive doesn’t guarantee success. Poor design and application of measurement principles are probably the largest keys to gauge failures. On the opposite side of Jacque’s observation, I have seen cheap off-the-shelf items do a better job of holding a part than one designed by a brilliant engineer who loves bells and whistles.

Ron is absolutely right about independent validation. I specialize in writing CMM programs and believe me, I can write some wonderful programs that will repeat results until the cows come home. But, without proper validation of the measurements, it is easy to make an incorrect assumption about a datum or a feature that can produce scrap parts from good ones or good parts from scrap.

—Jeff Pfouts, CNC & CMM operator, Kovatch Castings Inc., sent via Linked-In

Getting back to Ron’s question. There are a number of companies that send initial parts out for independent verification at commercial labs. I’m not really sure if very many companies run parallel tests for process control. Hopefully, the independent calibration and inspection of machining centers and tooling (molds and dies) lends to that. The final arbitor is often a part that assembles properly and/or functions to design intent. As Otto and Jeff point out, there are many variables that contribute to inspection errors. The function of the part and tolerances should dictate how to weight those variables.

Going back to the original question about precision test fixtures. Custom test fixtures are most often used in high-production settings. If the process is in good control, the process can be sampled, perhaps with a CMM or even hand gauge checks. If the process is not in control, then a precision test fixture becomes more critical, as theoretically the parts should be checked 100 percent until the process is capable and producing only good parts. Typically, CMMs cannot keep up with production rates if 100 percent of the parts need to be checked, whereas a custom fixture can be designed to meet that volume of inspection.

—Posted by Dan Smith, owner, FixLogix LLC, sent via Linked-In

For every production engineer and manufacturing professional that has to live with the machine to run production, one universal law always holds true....Quality can be expensive but Cheap still costs more.

The initial question asks essentially about the advisability of employing precision test fixtures. It would seem that the required precision be established by the performance requirements and the acceptable variance thereof. This effectively establishes the acceptable yield. This drives material selection, secondary processes, and assembly processes. Much as we can overdesign a product, we can surely overdesign a test fixture. If the fixture is “precise enough,” then its cost should be acceptable. If the test and retest error of the fixture is way better than the acceptable standard deviation of the performance target, then it is too precise and perhaps too expensive as a result. Isn’t this the material point?

Consider a cantilever test. The test fixture required to expose a load-to-failure of ±0.5 kilogram might be way different and way more expensive than one allowing for ±2 kilogram. The required level of precision will be dictated by product acceptability limits.

As for the utility of CPK values mentioned in previous comments, if memory serves, these values don’t consider the standard deviation of the process. In this case, two processes can have identical CPK values, but produce very different results. Is my thinking flawed in this regard?

By the way, I consider a test fixture to be an assembly that reliably presents a DUT to an input stimulus or stimuli to facilitate consistently accurate performance measurements. The measuring device—guage—is an integral part of this fixture since the fixture must be specifically designed to interface with it.

—Ron Scicluna, sent via Linked-in

The cost of ownership model includes all costs allocated to a machine. Basically, this includes:
* The base investment and write-off
* Setup cost
* Energy cost( expected in write off period)
* Maintenance cost (best predictable in company with preventive maintenance).

In terms of maintenance cost, more expensive fixtures can be defended. The more expensive fixtures often have better warranty and documentation of expected lifespan. When planning maintenance, the interval can be lower. This can be due to longer lifespan or just a shorter range expected fail moment.

When you have high-quality demands, some statistics-based calculation can be used. The difference in standard deviation of the fixture performance is a good factor to use. Standard deviation, plus case-based factors and relations, gives an indication of reject rate differences.

If the cheaper fixture does not come with any information on its lifespan, it is simply not the same product and cannot be considered.

—Posted by Willem Pieter Horinga, sent via Linked-In