Process Industries: Better Living Through Machine Vision

The line between discrete manufacturing and process manufacturing seems stark, but in truth, it’s a lighter shade of grey. For instance, process manufacturing is often described as a generating a product that cannot be distilled back to its basic components. A soft drink is a common example, where you cannot take a Coke and return it to water, citric acid, and its other constituent components.

Most semiconductor manufacturers would say you can’t return a microchip to a silicon boule, either, and yet electronics are considered discrete products rather than the output of a process. For this article, we’ll take the definition of process manufacturing one step back from the finished product, and focus on the active elements of production and manufacturing rather than the finished product. Examples include monitoring water in a variety of applications, and the Yin to that Yang, ‘hot processes.’

What’s the Frequency, Kenneth?
Machine vision for monitoring processes can take many different forms. In many cases, these vision systems are more concerned with energy than products. For instance, the printing industry is specifically concerned with true colors (in addition to spatial registration, see “Printing Goes ‘Custom’ with the Help of Machine Vision” ). Cameras used in these applications often use specific sensors, electronics, and algorithms to focus on specific bands of the visible spectrum rather than the entire visible spectrum to make sure a printing press is running properly and provide closed-loop feedback to keep the printing “process” within customer-set tolerances. Like all machine vision systems, printing applications focus on reflected light, or electromagnetic energy, but unlike most machine vision systems, they’re very specific about the targeted wavelength band.

For many machine vision process applications, designers look outside the visible spectrum to analyze the electromagnetic energy that separates a successful from a failed process. Infrared is particularly useful because many materials and molecules have strong spectral ‘fingerprints’ in the IR portion of the spectrum versus the visible part of the spectrum. For instance, metal or plastic can be painted green and appear the same to a visible camera, but an IR camera will easily tell the difference between the two based on what portions of the IR spectrum are absorbed by the molecules in the material, and what are reflected back to the camera. (See “Physics Knowledge Helps Infrared Vision Be All It Can Be” ).

All Wet
Materials used in process industries are rarely homogenous. If they were, there would be little to ‘process’ since the basic material would be acceptable for the finished product. Water is a common feature to many process applications, from adhesives to food products, according to Douglas Malchow, Goodrich ISR Systems (Princeton, New Jersey).

Goodrich ISR’s products use short wave infrared (SWIR), rather than mid-wave (MWIR) or long-wave IR (MWIR). These systems offer several benefits, including the ability to see through common glasses -- the other bands require special Germanium, silicon, or quartz windows because they are transparent to IR radiation. They can be built into larger arrays for less cost than other IR-sensitive compound semiconductors, with better spatial resolution than common microbolometer-based cameras in the MWIR and LWIR spectral bands. (For more information, see “SWIR Imaging in Hot Process Monitoring” ).

“SWIR cameras are regularly used in applications that use water-based adhesives,” explains Goodrich ISR’s Malchow. “As water in an adhesive dries during curing, SWIR gives you better contrast of the adhesive than visible. By monitoring the water content during the curing process, vision systems using our cameras can tell the customer whether they need to extend the drying time, or can shorten it up to improve throughput.”

“In food processing, there’s a similar moisture feedback process,” continues Malchow. “Let’s say the raw materials have come from a wetter climate and that the factory is cooking the food. By monitoring the water content, the customer knows how to adjust the cooking time. At the same time, we can identify foreign contaminates such as stones and plastic based on the difference in IR absorption between the ‘wet’ organic materials, and the ‘dry’ contaminates.”

Market Intelligence News & Insights: 

How Will Apple’s $500 Billion Investment for US Manufacturing, Education, and AI Play Out?

On February 24^th, Apple announced a “commitment… to spend and invest more than $500 billion in the U.S. over the next four years”, with plans for new manufacturing plants, enhanced support for manufacturing partners via their Advanced Manufacturing Fund, and a myriad of other initiatives. As part of this plan, Apple projects 20,000 new hires, with roles focused on R&D, silicon engineering, software development, and AI/ML engineering.

Read More

When it comes to water, the infrared spectrum is particularly useful for automatically identifying its presence, quantity, and the physical state of a process. “H2O and OH molecules have very strong, fairly broad, absorption bands in the SWIR,” says Malchow. “Depending on the physical state of the molecule, the center wavelength for that absorption can vary, and you can use that variance to determine the state of the process. For instance, as a water molecule freezes, the molecules vibration frequency drops, which cases the IR absorption band to shift to longer wavelengths.”

Although water is common and has broad IR absorption spectra, analyzing processes and materials using spectral fingerprints is tricky stuff because the spectral absorption curves for many materials overlap. In these cases, customers may choose to add a MWIR camera to the SWIR camera to give the machine vision more data to make an accurate analysis of the process, and the physical state of the constituent materials.

“Although molecular vibrations are easiest to identify in the MWIR band, MWIR cameras must be deeply cooled and are not always compatible with factory environments,” concludes Malchow. “In addition, thermal emission from the transport conveyor and the material itself adds to the MWIR image’s background noise, so making spectral measurements in uncontrolled industrial environments is difficult in the MWIR.”

Hot Flashes
Temperature is a critical element to many processes from pharmaceuticals to steel production. While SWIR cameras are sensitive to reflected IR radiation as well as IR light emitted from ‘hot’ materials, most thermal applications use MWIR and LWIR to make absolute temperature measurements because emitted IR fuels these bands more than reflected light.

“For more standard, room temperature scenes, intensity in the SWIR can be caused as much by reflected light as emitted light,” explains Malchow. “That’s why we don’t often use our cameras to make thermal measurements unless the temperature is above 100C. In the case of monitoring molten iron pours, you can be sure that the high temperatures mean that you’re only seeing emitted IR radiation, and there’s a huge difference in the emissivity of the iron coming out of the pour hole, and the slag floating above it.”  End-point of the pour is detected when the intensity steps up as the brighter slag (at the same temperature) starts to emerge from the hole.

Raytek Corporation sells IR cameras that range from SWIR, through MWIR and out to LWIR at 18 microns. Temperature measurements have allowed Raytek to build many interesting machine vision process-monitoring applications, according to Jimmy Earle, Director of Global Marketing for Raytek/Ircon/Datapaq Automation Products.

“We have a handful of applications in the food and beverage industry that we can talk about that use IR cameras to find hot spots and direct water jets to cool those hot spots,” explains Earle. “Tobacco, for instance, has to be dried, and sometimes tobacco will get stuck in the dryer. When it breaks loose, its overheated to a glowing ember. We detect those hot spots and trigger water jets to put the ember out before it catches the bail on fire.”

“In a similar application, we use either our MP150 – multipoint line scanner – that operates in the 3 to 5 micron IR band, or microbolometer arrays in the 8 to 14 micron range to detect hot beads of glass from fiberglass manufacturing,” continues Earle. “Fiberglass is spun in small fibers from nozzles and sprayed on a supporting material. The fiberglass is then covered by paper on both sides and wound up tight for storing and shipping. If a hot bead collects on the nozzle and then breaks off, it can get rolled up in the fiberglass. During shipping, vibration can loosen the roll enough to allow oxygen to reach the hot bead, causing the roll to catch on fire. Semi trucks have caught on fire traveling down the freeway because of this problem. By catching the bead as it leaves the nozzle and spraying it before the paper is applied, we can avoid the danger and costs to the manufacturer.”

These are just a few examples of how machine vision is not only checking products, but also monitoring the process during manufacturing to limit risk and avoid costly waste. By combining spectroscopy and material science, machine vision will continue to gain importance to process industries.