Line laser and CCD camera set up for triangulation.

Acquisition of 3-D information through line-stripe method.

In the last five years, technologies for vision systems have gained significant vigor, making machine vision a technology worth considering for locating and tracking seams for welding and for post-welding inspections. While machine vision is not an entirely new technology, these vision systems have not been not widely accepted for welding applications in the United States because they have been expensive and finicky.

What's new
One of the emerging technologies for U.S. manufacturers is weld seam inspection. This technology has been used in Europe for 20 years but did not catch on in the United States. That’s changing as manufacturers adapt their practices to adopt inspection systems that ensure quality based on the objectivity of a well-trained robot and a sophisticated computer program, and as U.S. welding shops adjust to having fewer employees on hand.

There are two broad categories of vision systems: 2-dimensional (x-axis and y-axis) and 3-dimensional (x-axis, y-axis, and z-axis), and 3-dimensional vision systems are replacing the less sophisticated 2-dimensional technology. The 3-dimensional systems have the ability to observe weld geometry and to recognize height and depth features in locating seams and in performing such tasks as inspection. Several manufacturers offer inspection systems for welding, including Vitronic Machine Vision Ltd. (www.vitronic.com).

That company’s Viro ^wsi system uses a measurement principle called 3-D stripe, a triangulation method that involves moving the part under a sensor or moving a sensor across a weld to scan the weld and acquire data. The system’s laser emitter creates the line stripes — sometimes called structured light — and has a matrix camera set at an angle. A matrix camera is a charged-couple device (CCD) that works as a video camera, not a still camera.

After the system is programmed, it can be used to check for many welding features, including the volume, thickness, length and height of a weld, and defects such as surface porosity, undercut, misalignment, excess convexity and incompletely filled grooves. The system also generates statistics on its inspections that serve as documentation for the inspection.

Inspection systems such as Vitronic’s work with arc, laser and electron beam welding, but they are not suited to every application. Each application that they are used on requires some engineering, says Robert S. King III, sales and marketing manager for Vitronic.

Automotive manufacturers are considered the leaders in the use of vision systems in the U.S. They have been using weld inspection systems mounted to robots for several years to inspect the numerous welds used on each vehicle they produce. Welded pipe producers also have used significant numbers of vision systems in their manufacturing process.

What’s been here awhile
Early users of machine vision systems had to understand how their vision systems worked so they could adjust the systems to the variables, such as plant lighting, that seemed to change constantly. That has changed. Machine vision systems that now are available are robust. They have been developed and are built with features that are designed to reduce variables to a minimum, such as automatic gain control for lighting that maintains a steady light on the object being viewed by the machine, and automatic calibration that helps the system to maintain reliability. These features have advanced vision systems, and have allowed them to be used for such applications as welding. Combined with motion and tactile sensors, vision systems are being used to provide a greater realm of perception to welding robots.

Seam location
Vision systems for robotic welding provide either guidance or inspection. In the area of guidance, one of the most common applications is seam location, says Pierre Huot, president of Meta Vision Inc. (www.meta-mvs.com).

This application requires a robot to move to a specific position over a seam. An analog or digital camera photographs the joint’s position prior to welding, and data from that pre-weld scan is sent to the controller for analysis. If the seam is not located at the position that was previously expected — the robot is “taught” where the seam location should be as a part of setting up the job — offset information is generated and is used to make corrections so that the programmed weld path matches the seam’s new location.

This type of system can be used when making short or long welds, especially welds that that do not change along their length, and welds on products that can be tightly fixtured so that the seams remain in good position as the welding system moves from part to part. The best examples of the products made with this type of system are the vehicle frames made in automotive manufacturing plants.

Other machine vision systems are used for applications in which weld joints are likely to change along their path or are likely to move during production.

Seam-or through-the-arc tracking
Seam tracking uses machine vision to follow the points at which welds are to be made and to place the welding gun or torch into the correct position.

Some seam tracking systems operate ahead of the torch, using a camera mounted on the robot arm. The machine vision system uses the camera to scan the joint, and provides real-time feedback so the robot can make corrections to its path before welding begins. Other seam tracking systems, especially those used with GTAW systems, operate through the arc.

Through-the-arc-tracking (TAST) systems include weld current sensors and software designed to interpret current feedback. That feedback is analyzed and used to modify the torch’s vertical position so that a constant stick-out is maintained. For these systems, the robot arm is programmed to weave across a seam that is to be welded. As it weaves, the welding current oscillates, and the combination of the current oscillations and the position of the robot are used to determine the position of the seam. By comparison, arc voltage control systems monitor welding voltage, and the software uses the proportional relationship between welding voltage and arc length to maintain the correct torch position over the seam.

Huot says seam-tracking systems are used to weld pressure vessels and long joints, such as those used in the large tanks needed by nuclear power generating plants, fuel and water tanks and for tube and pipe.

According to Mike Sharpe, manager of materials joining at Fanuc Robotics Inc., through-the-arc systems are appropriate for any welding job that has a large gap condition or variability in the joint. Newer, laser-based tracking systems have been developed to work with a robot, software and welding machine, and to adapt the welding process dynamically as the joint geometry changes. These systems make positional corrections and refinements to the weld schedule to ensure that the joint is properly filled.

Analog cameras? Pierre Huot, president of Meta Vision Inc., says analog cameras are not expected to go away soon. While the use of digital cameras is increasing, analog cameras have a long and strong legacy, and are likely to be used into the future for such applications as machine vision systems and for security cameras. Huot says he expects analog cameras will be phased out eventually, but, for the present, the investments that many users of vision systems put into analog cameras and the costly software developed to support them will keep them around.

Repairing turbine blades for airplanes Liburdi Engineering Ltd. (www.Liburdi.com) and its sister companies manufacture welding machines and provide engineering services to companies that use them to repair turbine equipment, such as turbine blades from jet aircraft engines. The company was forced into developing vision systems for its specialized equipment, says Robert Pistor, principle engineer. “It was necessary. It’s the only way to automatically weld turbo machinery components. There is so much variation in geometry that you cannot find an optimum set of parameters to weld that part without adapting to the features that change. So we introduced our first vision system in 1988 and have been continually adding features and keeping up with the technology — cameras, computers and recognition algorithms,” says Pistor. Repairing a jet engine blade is much different than manufacturing a new one, he says. When a blade comes out of an engine, its shape differs from the original geometry because of wear and tear. Blades may be bent or eroded. Repairing them requires laser welding and a vision system that can adapt the weld schedule to the shape based upon how much material remains on the blade, how thick the welded joints will be, and how thick the part features are. One of Liburdi’s systems is the laser seam tracker, or LST. It shines a laser line, or structured light, onto the blade. The camera looks at the laser line and makes a vision scan. Once the scan is completed, the weld computer automatically updates the weld schedule and overrides the current. If the system used ambient light instead of a laser, someone would have to polish the blade so that the edge would be reflective. A camera would then measure the airfoil shape, and create a weld path and weld schedule appropriate for welding the particular shape of that particular blade. Pistor sees the company’s technology moving to 3-dimension on the fly — using a laser-based scanning system that will generate a tool-path as the part is being welded. He says the current state of vision technology involves a pre-scan of the part and a computer doing calculations before the welding is performed. As he sees it, “the future is simultaneous scanning and welding.”