1 January 2006
IR is HOT For Monitoring Process Temperature
Automated infrared imaging technology gains inroads in niche temperature-sensitive segments
By Jim Strothman
While traditional machine vision systems using high-speed video cameras work well for many industrial applications, automated infrared (IR) imaging technology has the edge when it comes to major niche temperature-sensitive segments. Tops among those areas are monitoring critical vessel temperatures in process control applications, predictive maintenance of motors and other rotating equipment, and inspecting ceramic igniters used to fire natural gas-powered ovens and dryers.
"Companies with gasifiers—for cracking fuel under high temperature or high pressure, for example—can use IR cameras to make sure the process is safe," said Erik Goethert, machine vision veteran and custom measurement program manager for Boston Engineering Corp., a Waltham, Mass.-based engineering, design, and fabrication firm.
To prevent unscheduled stoppages of motors and other rotating equipment, heat-sensitive IR cameras focus on the machinery's bearing areas. When pre-set heat levels are exceeded by broken or worn-out bearings, alarms sound and/or reports are generated before a total failure occurs, Goethert said.
In another application, a manufacturer of ceramic igniters uses IR technology to locate cracks not visible to the naked eye. IR sensors find faults in the ceramic devices by heating the igniters only slightly using a low electrical current. "This makes the quality control process much less complicated," said Goethert, whose 15 years of machine vision experience includes working for the U.S. Environmental Protection Agency (EPA) and consulting for the Big 3 auto makers in Detroit on emission control- related custom measurements.
Integrates with off-the-shelf PCs
Imaging Organizations on the Web
Enterprise Content Management Association (AIIM)
Automatic Identification Manufacturers (AIM)
Automated Imaging Association (AIA)
ASPRS (Remote Sensing/Photo-gram)
DPIA (Digital Printing & Imaging)
IEEE (Electrical/electronic engineering)
IS&T (Digital photo/print science & technology)
Machine Vision Online
Robotics Industry Association (RIA)
Semiconductor Equipment and Materials International (SEMI)
SIGGRAPH/ACM (graphics, animation/visualization)
SME (Industrial engineering including MVA, machine vision)
SMPTE (Video engineering)
SPIE (Optical/electro-optical, imaging)
Surface Mount Technology Association (SMTA)
UK Industrial Vision Association (UKIVA)
Goethert detailed the IR applications at a recent InfraMation thermographers conference. "Automated IR imaging has the capability to improve many industrial tasks such as process monitoring and control, predictive maintenance, and quality control," he said.
IR cameras using analog, (IEEE 1394) Firewire, Ethernet, and/or other input/output (I/O) can be integrated with a basic commercial off-the-shelf personal computer to create an automated system, and more advanced systems can be programmed under a real-time operating system.
When using a vision system for closed-loop process control, real-time processing is typically needed for a deterministic system, he said. Determinism—the ability to process and execute within a specific time, such as every 200 milliseconds—is not guaranteed with Windows. "Windows may decide it's time to do a virus check, slowing your application down," Goethert said.
For deterministic applications, the Boston Engineering manager uses the National Instruments (NI) Compact Vision System, which uses a Firewire connection and can support up to three cameras. Any software developed for the PC system can be ported over and run using a real-time operating system. It also enables users to configure applications with NI's Vision Builder software for automated inspection (AI) or program it with NI's LabVIEW.
IR cameras have sensors that convert IR energy into a table or array of points, depending on the number of pixels on the sensor. Goethert uses cameras manufactured by FLIR Systems, which range from 160x120 to 640x480 pixels. The distance from the camera to the object, or field of view, determines the resolution or spot size of each pixel. This array of values is converted to an image with temperature information and displayed on a cathode ray tube (CRT) or liquid crystal display (LCD).
An 8-bit camera, for example, would show values ranging between 0-255, with the user pre-determining exactly what temperature the value represents. With more expensive cameras— such as 14-bit (214), which offers values ranging from 0-16,383—can relate directly to temperature. The image seen on the screen is the computer mapping these values to a color or gray-scale scheme. Extra radiometric information is also stored with each pixel for precise temperature measurement.
IR cameras can automate processes in at least two ways: 1) measuring the temperature of an object across its full surface or section of the surface, or 2) making traditional machine vision measurements of objects or processes that are best discerned in the infrared region of the electromagnetic spectrum.
Most IR cameras have the ability to return radiometric thermal imaging data for use in obtaining the temperature of the object—which is important if the actual temperature of the object at various points on its surface needs to be measured. The camera is being used as a non-contact temperature sensor, with the many temperature measurement points in the field of view. A camera with a very high thermal sensitivity that is actively cooled is suitable for these precise measurement applications, Goethert said.
Machine vision, on the other hand, is looking to identify features that become apparent with a temperature difference, such as the fill level of a bottle or can or the presence of a heater element in a seat or car window. To the machine vision system, the physical temperature is not as important as the presence or absence of a feature or pattern; the most important specifications are contrast—the difference between the temperatures or intensity of the feature and the rest of the object—and resolution (the smallest feature that can be viewed).
After an IR camera acquires data from the sensor, the image can be transferred to a PC through various methods: analog (RS-170, NTSC, PAL, SVideo), Ethernet, or Firewire (IEEE 1394). Analog acquisition requires a frame grabber board in a PC to convert the analog RS-170, NTSC, or S-Video signal into a digital form. Most suppliers provide boards that interface with many of FLIR's cameras. All data, including radiometric data, can be retrieved via Ethernet.
Firewire (IEEE 1394), a bus originally developed by Apple Computer, is good for handling video data because it can guarantee data delivery, Goethert said.
Once the image is in the PC's memory, it can be processed and analyzed to enable decision-making. Basic machine vision analysis uses the intensity levels along a line or an area in the image to determine the size of a feature, the shape, the number, or even whether a feature is present. Temperature measurement applications, on the other hand, depend upon the radiometric data the camera returns. Typically, the user draws a line or box in the image, and then looks up the specific temperature information. Users can then make decisions depending upon the specific temperatures.
While traditional machine vision applications for discrete manufacturing continue to grow steadily, "the infrared (IR) side is really opening up," said Kyle Voosen, NI's vision product manager.
ABOUT THE AUTHOR
Jim Strothman is the former editor of InTech and a freelance writer in Cary, N.C. Reach him at email@example.com.
Share the online version at www.isa.org/intech/060005
Machine Vision Basics
Whether by humans or machines, vision plays an indispensable role in manufacturing. Human and machine eyes (typically high-speed cameras) inspect parts and assure quality and consistency of manufactured products. Machine vision is needed when manufacturing process speeds surpass the limits of human eyesight (small sizes of parts, products, or fault areas).
In a typical machine vision application, a video camera positioned on the production line captures an image of the part or product to be inspected and sends image data to the machine vision computer. The computer uses sophisticated image analysis software to make measurements and determine quality and consistency. Results then go to other equipment on the factory floor, such as an industrial controller, a robotic arm, a deflector that removes the part from the line, or a positioning table that moves the part. This process is repeated for each part on the line—or continuously, for process material— as it moves into position in front of the camera.
Vision systems can perform inspections quickly enough to keep pace with machines that process thousands of items, or material feet, per minute, increasing quality control and productivity, according to machine vision sensor supplier Cognex.
In a typical machine vision application, a video camera captures an image of the item to be inspected, such as a bottle of cough syrup or "blister pack" of pills. The vision system can determine whether the correct date and lot codes have been printed on each bottle or package. If the system determines the product is faulty, a blower or mechanical arm pushes defective products from the line into a reject bin.
According to the Automated Imaging Association (AIA), in the automotive industry, machine vision is used for assembly, process verification, flaw detection, gauging, part location, and robot guidance. More specific applications include fuse box inspection, tire tread recognition, powertrain, and sheet metal inspection.
In the pharmaceutical industry, vision applications include inspecting filled and unfilled vials and ampules, packaging solid dosages, proofreading labels and inserts, and verifying date and lot code. General purpose machine vision systems are used in the medical device industry to inspect products such as orthopedic, prosthetic and surgical appliances, and ophthalmic devices including contact lenses and eyeglasses. –Jim Strothman