Generating Intelligence from Data Brings Machine Vision to Process Industries
| By: Winn Hardin, Contributing Editor
To a machine vision system designer, every industry is a process industry because every manufacturing or automated measurement system generates data that, when properly analyzed, can be used to make the manufacturing process better.
But the subset of manufacturing typically called "process industries" doesn't agree. It’s okay. They were here first.
Generally speaking, process industries are involved in recipes and raw materials, not discrete products using bills of materials and counted in finished units. Examples include chemical processing, food and beverage, biotechnology, and metallurgy, to name a few. These industries are not typically considered major consumers of machine vision technology because, in chemical processing, for instance, constituent fluids often are rushing through pipes, swirling around mixers, and being heated, pressurized, and otherwise modified at molecular levels that resist standard imaging techniques. But that’s where the imagination comes into play.
For example, imaging systems may not be able to see inside a closed furnace to make sure the heat is evenly distributed and the materials inside correctly modified. But it can look at the material undergoing the heating process and extract intelligence about the effectiveness of the process. The Fourier transform allows machine vision systems to measure frequency versus spatial data, and there are many more machine vision tools that can help the process manufacturer succeed in today’s market.
Process or Not? That Is the Question
“No matter whether it’s polymers or wheels, a machine vision system is about making adjustments to the process based on the available information,” says Nicholas Tebeau, Manager of Vision Solutions at LEONI (Lake Orion, Michigan). “The ideal being a true closed loop where, for example, I measure the temperature of an object and if it’s cold, the system turns up the oven. It’s not always so straightforward in process manufacturing industries, but you’re still doing something with information to improve the overall process.”
While furnaces and mixers are closed, and cameras are of little use for extracting meaningful information from the maelstrom inside, as long as there is some visual factor associated with a processed material, then machine vision can benefit a traditional manufacturing process.
For example, LEONI recently was asked to determine whether a polymer-coated surface had the right friction coefficient. Taking a picture of the polymer surface wasn’t a viable solution, but by running an image through a Fourier transform and looking at frequency patterns, Tebeau was able to verify that the individual polymers on a metal surface were sufficiently ground to minimize friction.
In another polymer-based application, a customer needed to know that a Teflon powder was cooked to the correct temperature to deliver the requisite material properties for a downstream application. The color of the Teflon powder was directly related to the temperature and duration of the heating process. By measuring the color, LEONI was able to verify the Teflon manufacturing process even though it had no direct way to image the Teflon during heat treatment.
Calibrating the Process
Each year, U.S. agricultural equipment manufacturers buy several specialized line-scan cameras from Imaging Solutions Group Inc. (Fairport, New York). “John Deere manufactures the equipment farmers pull behind a tractor to put seeds in the ground,” explains ISG Co-founder Kerry Van Iseghem. “The seeds can’t be too close, no matter if the tractor is running full-out mid-row or slowing down to make a turn. To verify that the equipment is seeding the ground at the right interval, John Deere uses ISG’s line-scan cameras to count seeds and spacing as test units are put through their paces. We keep telling them they should install the camera on every tractor and they won’t have to worry about spacing ever again, but we haven’t won that battle.”
In another application, a customer wanted to verify steel I-beam extrusions as they emerged from the metal forging equipment. Often, thermal cameras would be used for measuring very hot objects such as glowing hot steel I-beams. However, this customer opted to use a 3 megapixel (MP) CMOS area-scan camera so that the vision system could check the presence of the beam as well as measure its dimensions. “CMOS was cost-effective, didn’t bloom, and didn’t have saturation like you would with a CCD camera,” Van Iseghem says.
Merging Machine Vision with Spectroscopy
For many processes, there is no substitute for collecting images in the infrared area of the spectrum. While IR cameras will yield some of the necessary information, for many process applications, end users need to look at many parts of the visible and IR spectrums.
EVK DI Kerschhaggl GmbH (Raaba, Austria) offers its HELIOS camera with near-infrared (NIR) spectroscopy and industrial image-processing capabilities. HELIOS displays hyperspectral imaging data as chemical colors that consequently can be processed by industrial image-processing methods.
In recent years, the HELIOS has helped McDonald’s limit the number of burned French fries by helping potato processors identify fries with high sugar content at their ends, which leads to burning in the fryer. The application won an International FoodTech Award in 2012. The same technology has been used in plastic sorting and recycling, as well as the mining industry for identifying mineral-rich rocks.
“Spectroscopy is a proven technology in lab environments,” says Sieglinde Kepplinger, Product Marketing Manager at EVK DI Kerschhaggl. “Our HSI line-scan camera systems enable real-time applications in the processing industries by helping them to ‘picture the invisible.’ Our cameras’ internal classification and sort functions make it easy to replace color cameras. And the easy-to-use software allows customers to develop their own applications, even with limited knowledge of hyperspectral imaging technology.”
While process industry traditionalists may not agree that every application is a process application, by looking beyond the immediate, the visible, and the obvious, the machine vision designer can help bring large-area quality inspection to traditional process industries and further improve their productivity and profitability.