Infrared imaging helps automate part inspection and tracks manufacturing performance in real time.
Edited by Jessica Shapiro
Manufacturing and process engineers boost efficiency and cut costs by turning to machine-vision systems like automated infrared (IR) imaging. The technique is being used in a host of industrial production applications, including process monitoring and control, quality assurance, asset management, and machine-condition monitoring.
In IR imaging, an infrared camera converts heat radiating from an object or scene into a visual image depicting temperature variations. In many cases, temperature readout is a direct indication of process progress or final part quality. Operators use this data to adjust processes and reject bad parts. Calibrated temperature data from IR cameras can isolate or mark bad parts for rejection.
IR cameras can detect heat radiating from below the surface of an intricate assembly. This lets the system or operator determine if internal parts are operating correctly. Automakers have exploited this attribute for checking the operation of seat heaters. Contact temperature sensors slow down production and are inaccurate if not placed precisely. In contrast, IR cameras give visual images and accurate noncontact temperatures of heater elements inside seat assemblies.
The permanently mounted IR camera swings into measurement position when a car on the assembly line reaches the inspection point. A nearby monitor displays a thermographic image of the seat-heater element and its temperature. Other automotive heating elements checked by IR thermography include defroster elements in rear windows and side-view mirrors. Beyond automotive applications, such systems are used to inspect component placement on PC boards and joining lines on plastic welds.
Like modern visible-light cameras, thermographic cameras and associated software can recognize the size, shape, and relative location of target objects using pattern matching. The software compares current visual images and temperature readings to stored images that include temperature readings representing good parts. Excessive variation between the two readings triggers an alert.
Onboard firmware gives the camera Supervisory Control and Data Acquisition (SCADA) capabilities for autonomous process monitoring and control. Users can configure the camera with built-in logic and send digital outputs to alarm and control devices.
Get with the program
A standard Web browser lets users receive real-time digital video output and remotely configure and control multiple independent target spots and alarms. Software running on a PC expands these capabilities for more complex needs.
For example, FLIR’s IR Monitor software monitors and reports over a local-area or wide-area network (LAN or WAN) so that any authorized user on the network can instantly see thermal images and temperature data. The software also analyzes temperature trends and compares them to stored, user-defined data. It also sends alarm messages and images via e-mail to remote locations using simple mail-transfer or file-transfer protocols (SMTP and FTP).
Users can design monitoring and control systems using various developers’ toolkits. National Instruments’ (Austin, Tex.) LabView developer’s toolkit can turn an IR camera into a machine-vision tool with minimal investment in application development. System kits (SDKs) based on ActiveX, Visual Basic, C++, and other codes can shorten the time it takes to create applications for IR cameras.
Cameras that have a Gigabit Ethernet interface and are GigE Vision capable are compatible with several machine- vision software packages. These include VBAI from National Instruments; MIL from Matrox, Dorval, Que.; Common Vision Blox from Stemmer Imaging, Puchheim, Germany; and Halcon 9.0 from MVTec, Cambridge, Mass. A&B Software, New London, Conn., has a GigE Vision SDK that lets users and integrators do their own programming.
Many of these third-party software packages come with the most commonly used machine-vision functions preprogrammed. One example is pattern-recognition software that aids inspection for soldering and welding metal parts, as well as for laser welding plastic parts. The software can learn a weld path, then verify it is correct on production parts. Alternatively, program developers can save a “perfect” part’s image and have the software compare it to a new part.
IR cameras with logic and memory can analyze images and perform alarm and control functions, but data communications are SCADA’s backbone. Many cameras have data I/O ports along with wireless and fiber-optic connections. However, Ethernet has become the de facto communication standard for such systems because it’s reasonably fast, flexible, and inexpensive to implement.
Many IR cameras have a 100BaseT (or higher) Ethernet interface and may provide FireWire (IEEE 1394) connectivity as well. Some units accept Power-Over-Ethernet (POE), so they don’t need local power supplies.
Ethernet lets multiple cameras synchronize operations using the Simple Network Time Protocol (SNTP). In this scheme, each camera has a unique Internet Protocol (IP) address that lets the network detect it and accept the data the camera generates.
Bringing quality back in line
If X-rays showed a significant problem in parts from a particular mold, quality-control personnel relayed this information to the production area so workers there could adjust the mold’s temperature. The extra steps and delayed feedback often resulted in high scrap rates.
To streamline the process and cut down on scrap, the manufacturer installed an IR-camera system in the mold area. Data from the IR system let the mold operator check mold temperature distribution in real time and correct the problem before more castings were affected.
Some camera manufacturers have recently added EtherNet/IP for command, control, and data collection. EtherNET/IP is compatible with the Common Industrial Protocol (CIP) developed by the Open DeviceNet Vendor Association (ODVA). Using CIP commands over Ethernet gives system developers an industry-standard network communications protocol that simplifies the integration of IR cameras with PLCs. The resulting machine-vision systems can perform a wide variety of monitoring and control duties.
Vision systems need the speed and data capacity Ethernet provides because still images from IR cameras are actually frames of video output. Newer IR-camera electronics process video signals at up to 1,000 frames/sec. The 1-kHz video frame rates capture fast-moving parts on production lines by snapping new images every millisecond. In contrast, more common 60-Hz frame rates capture new frames about every 17 msec.
The camera analog-to-digital (A/D) converters combine short integration times with 14 to 16-bit resolution, critical features for characterizing targets in motion or if temperatures are changing rapidly.
Blow-molding machines are one application that calls for relatively high-speed image capture. One IR-camera setup monitors preheating of polyethylene-terephthalate (PET) bottle preforms. The blow molder produces up to several thousand bottles an hour, so robotic mechanisms and bottle parts move quickly. An IR camera with a 60-Hz frame rate takes images and temperatures at several locations on each preform as it exits the preheat station. This ensures preforms are at the right temperature before they enter the blow molding station.