Structured thinking

A US machine vision technology developer has adapted a standard 2D barcoding system to detect early signs of fatigue in highly-stressed structural components. Charles Masi reports.

Engineers have learned much about how metal fatigues, and the limits of what can be asked of any structure, but it is virtually impossible to keep track of most working structures’ histories.

So, while we can predict how much a given structure can take, once it goes into service, we can’t keep track of the punishment it takes — or how rapidly it ages.

That means it is impossible to predict when it will fail with any certainty. In the absence of any way to predict when an item will fail, we are reduced to a high-stakes gamble in which the potential for catastrophic loss is balanced against the certain cost of early replacement of good components.

According to US company Direct Measurements, which has developed a fatigue indicator system using standard 2D barcoding techniques, if an inexpensive, easy-to-read indicator that performs over the life of structural components could be installed, the impact and benefits in structural-health monitoring could be enormous.

As structural components age, repeated stress cycles introduce permanent irreversible damage that accumulates. Whether it is a pipe in an oil rig or a strut supporting a passenger jet engine, fatigue damage in a structural component appears as a permanent strain, which is evident long before fatigue cracks begin to appear. That permanent strain, or ‘creep’, increases over time.

Materials science provides the tools needed to determine how much of the component’s service life has been used up, based on the accumulated strain. In addition, it is well known that the rate at which strain accumulates (given consistent stress conditions) increases as the part ages, becoming very rapid as it nears failure.

The hard part is to economically monitor the strain build-up. Strain gauges can measure the relatively small strains associated with early fatigue damage, but are not well suited for long-term, in-service operations. They are also too expensive and fragile to fix at every potential trouble spot and monitor carefully over time.

In the 1990s, researchers at DMI realised that machine vision technology was approaching the sophistication needed to measure distortions introduced by accumulated strain at the microstrain (part per million) level. Also, marking technology had reached the level where stable permanent marks were being applied to a wide range of structural components for tracking purposes.

The team came up with the idea of using a machine-vision-based reader to monitor creep in these components by measuring how accumulated strain was distorting these marks. The hardware consists of a basic machine-vision system — a high-resolution camera connected to a host computer through a frame-grabber card installed in the computer.

In parallel with the barcode-reading software needed for material tracking, a second software thread analyses the 2D barcode image, measuring the distortion caused by strain accumulating in the part. The system senses distortion by comparing the current image with historical images of the same part.

Images taken over a period of time show a steady elongation along the strain axis, compared to the crossstrain direction.

By analysing trends in this series, the software determines the absolute strain, and the strain accumulation rate changes with time. It also calculates a quantitative measure of fatigue.

Before this concept could be developed into a field-ready instrument the team had to overcome several obstacles. The first was the high cost of hardware. The optical requirements for an image sensor needed to measure strain are much higher than those needed to simply read a 2D barcode.

Time, however, solved this problem. Over the decade or so since the company conceived the idea, the cost of multimegapixel machine-vision cameras and high-quality optics has plummeted as the technology became a mainstay of inspection technology. So, as image-sensor production volumes went up, prices came down.

The fact that the technology needed to be portable presented another problem. Generally, the reader has to go to the structural component, rather than the component going to the reader. Not only did this present a packaging problem, but it meant that the image-analysis system could not be sure that the barcode would appear in a consistent location or orientation in the image.

The simple analysis algorithms available in the early 1990s were not smart enough to translate and rotate the image to a standard position or orientation fast or accurately enough. So the researchers had to write the software themselves. They also had to write software to compare current images with historic images, and calculate shifts due to accumulating strain. Not only did they need software to do the comparison, but they also needed an archiving system to hold the comparison images.

The handheld reader head contains an LED ring light to illuminate the mark, and a camera with fixed optics to image it. The operator places the reader over the mark, then depresses a key on the laptop host computer. Software on the laptop acquires the image and performs all of the operations needed to read the barcode and analyse its distortion. The results appear on the screen as well as being wirelessly transmitted to a central location over the operator’s corporate intranet. There, a decision can be made what to do with the component and sent back to the operator.

In a oil-drilling application, for example, drill pipes are re-used for several wells. As they age, however, they need to be moved from high-stress sections of the drill string to lower-stress (usually higher up the string) locations for the next well.

Measuring the fatigue in each pipe as it is returned to inventory makes it possible to deploy it for lighter duty next time. DMI and Lockheed Martin recently completed proof-of-concept testing using DMI’s fatigue gauge.

A test specimen with a 4.76mm diameter hole was used to simulate a stress concentration region near a rivet or fastener. A standard 2D barcode was applied around the hole to provide a strain measurement mark. To provide a control measurement, researchers mounted strain gauges on the specimen’s back side on either side of the hole. While the hole interferes with some of the information in standard 2D barcodes, it doesn’t affect the fatigue measurement.

DMI has now developed a special barcode that provides an information-free zone in the centre to accommodate drilled holes. Thousands of cycles were applied to the test specimen in a load frame at Lockheed’s structural test laboratory in Marietta, US. The DMI system measured much higher absolute levels than the strain gauges.

At first this might seem surprising, since they ostensibly measure the same strain in the same specimen. But the barcode hugs the hole contour much more closely than the physically larger strain gauges. This would seem to prove that the new system, therefore, more accurately samples the strain resulting from the stress concentrated around the hole’s edges.

The specimen was not tested to failure. In fact, microscopic examination after the test revealed no signs of cracking near the hole. This means that all of this data reveals stress accumulations prior to any visible damage. The company is confident that its instrument is now field ready.

In a sense, however, it is a solution looking for a problem. Ironically, DMI needs engineers facing structural-fatigue monitoring challenges in various industries to look at the technology as a possible solution.