The Birth of In-Process Testing

An automaker’s search for something better than hot test redefined quality on the assembly line

Published: Tuesday, October 4, 2016 - 12:01

There was a time when manufacturers thought that “hot test”—a test at the end of the assembly line of a fully functional engine—was the only way to ensure that each unit had been assembled to perform as expected.

A lot has changed during the past 20 years. Manufacturers, from automotive to medical devices and even printer cartridges, today understand that just about anything can be tested during the process of assembly, with the goal to catch defects at the earliest point on the line. But it was with the automotive sector that in-process testing began during the early 1990s.

At that time, Sciemetric was building a reputation in the marketplace for test and measurement equipment and was already known to many of the big automakers. One of those big names was looking for an alternative to the traditional engine hot test. Hot testing was expensive, took up lots of floor space in the plant, was bad for the environment due to emissions, was a very subjective and often ineffective test, and added no value to the product.

We had already developed a signature analysis solution for other auto manufacturers, using sensors to capture the waveform of a process—in this case, to verify valve seat and valve guide assembly operations. Signature analysis proved ideal to monitor force vs. distance when parts were pushed into a cylinder head casting, to ensure the right fit without damage.

If it can work for this...

The question came up: Could process signature analysis, which wasn’t even yet a recognized domain, be applied in a systematic fashion up and down the whole assembly line?

We offered our help to figure out the answer with the automaker that wanted something better than the hot test. Our team participated in the “simultaneous engineering” process, to think through every step of the process required to assemble a functional engine.

We looked at all key assembly operations on the engine assembly line to understand which processes could be measured more precisely to detect errors, identify bad parts, or expose bad assembly operations immediately rather than waiting to detect them downstream.

In each instance where some chance of error persisted, we imagined how we could measure and qualify the process with signatures or waveforms to verify perfection and spot problems quickly.

Catching the errors that can’t be avoided

At the end of the exercise, we boiled it all down to about 12 in-process testing or monitoring stations along the line, and added sensors and data acquisition at each station to measure key values like force, angle, distance, pressure, and torque.

It wasn’t perfect. Of those first 12 stations, only nine or 10 made the final cut because they offered a good return on investment and a worthwhile improvement in quality and first-time yield at an appropriate cost point. But that was enough to dramatically improve first-time yield within a few months of the launch of a new line. Defects were caught immediately at or near the point of assembly, and therefore the root cause could be diagnosed and eliminated systematically. Not only did waveform analysis improve quality by catching more errors, it also allowed a new line to be calibrated much faster than before.

The results from those nine or 10 stations were so impressive the automaker decided to integrate in-process testing technology into every plant it had around the world as new lines were built. This multibillion-dollar investment took a decade to implement. But the return on investment was fantastic, and the cost was recouped almost immediately through the cost savings of dropping the hot test. Equally important, the quality and productivity of these assembly lines were superb—best in class for the industry.

Little errors can add up to big problems fast

It’s interesting to consider how dramatic the impact of little errors can be on first-time yield when compounded down the length of an assembly line as complex as that for an internal combustion engine, which can have hundreds of assembly stations.

Let’s say there are 100 stations in series. Say each station produces 99.0-percent first-time yield. This may sound good, but by the time you reach the end of the line, the rolling throughput yield is just under 37 percent—which means the scrap or rework rate is more than 63 percent! That’s obviously disastrous. This effect is illustrated in figure 1.

Figure 1: Assembly line rolling throughput yield. Click here for larger image.

The moral of this story? The more stations on a line, the greater the need for a quality assurance process that ensures each station is operating at 100 percent.

When walking the plant floor grew too tiresome

The next big step came when I just got tired of walking the floors of these huge plants every time I wanted to see the data from a station.

Through the 1990s, those testing stations were islands. There was no data management, not even reliable networking. Ethernet hadn’t even been adopted yet as an international standard, and IBM token-ring was the prevailing standard for many plants (IEEE 802.5).

We figured there had to be a better way.

Our first attempt at bringing all that data together for centralized management and analysis was purpose built for the automotive industry; it was the Modular Engine Test Engineering and Repair System (METERS). This was the first generation of traceability, to be able to share data between the plant floor and the front office, and dig into it for the insight to drive even greater process improvement.

Achieving the ‘perfect’ assembly line

During the early 2000s, the next generation of this data management software suite was renamed QualityWorX, to reflect the fact that this kind of data analysis and management could be applied to any discrete manufacturing process, not just automotive.

Little by little, we built more and more data gateways into the system to take in-process testing to the next level so everything that’s pertinent can be measured and stored for greater insight and correlation. One of the latest steps has been to add machine-vision image storage.

Industry 4.0

In recent years, “Industry 4.0” has become the hot new buzzword, but anyone working with in-process testing and waveform signature analysis has lived this concept for decades.

We are quickly moving to the point where we can collect, correlate, and analyze enough data to make any assembly line perfect. And it all began 20-some years ago in automotive—the industry known for pioneering the moving assembly line in the first place with Ford and the Model T in 1913.

About The Author

Nathan Sheaff

Nathan Sheaff is founder and CEO of Sciemetric Instruments. He developed Sciemetric's innovative Process Signature Verification technology used to detect and analyze defects on manufacturing lines across many industry sectors. Under Sheaff's leadership, the company has grown into a multinational enterprise, with subsidiaries in the United States, Europe, and Asia, and an impressive list of Fortune 500 customers that includes Ford, Hewlett-Packard, Jaguar Land Rover, Caterpillar, Honda, John Deere, and Medtronic.