Automated Weld Integrity Analysis Using 3d Point Data

This paper describes a method for non-destructive evaluation of the quality of welds from 3D point data. The method uses a stereo camera system to capture high-resolution 3D images of deposited welds, which are then processed in order to extract key parameters of the welds. These parameters (the weld angle and the radius of the weld at the weld toe) can in turn be used to estimate the stress concentration factor of the weld and thus to infer its quality. The method is intended for quality control applications in manufacturing environments and aims to supplement, and even eliminate, the manual inspections which are currently the predominant inspection method. Experimental results for Tfillet welds are reported. INTRODUCTION It is a common myth that manual welding in factories has been entirely displaced by robotic welding. While it is true that for large production volumes a robotic solution is commonly used, this technological investment cannot always be justified for small production volumes. Deploying a welding robot can easily cost as much as $1 million or even more in some cases. Therefore, even large companies continue to use human welders. Even if cost were not a factor, there are still some structures that are simply too big to be welded by a robot. These structures have parts that are either too long or have welds in positions that can be difficult for a robot to reach. Some structures also contain many sub-components, which have to be manually positioned in place before they are welded. Thus, in many factories manual welding is still performed on a daily basis. It is unlikely that this situation will change anytime soon, which means that manual welding operations are likely to be a part of normal manufacturing processes for at least another decade. For this reason alone, it is useful to develop reliable and automated techniques for quality inspection of welds. In the age of globalization quality control systems for weld inspection become even more important. While many large manufacturing companies have their own welding schools and routinely train their own welders to weld properly, it is becoming increasingly rare to find products that are welded from start to finish in the same factory. These days a product can contain parts that are welded across different countries and even continents. It is also becoming very common for products to be welded by more than one company, each with a slightly different quality control record. Sub-components welded by third-party suppliers are routinely integrated into finished welded assemblies. While outsourcing the welding of these sub-components may be cost efficient, the company that sells the finished product is ultimately responsible for the quality of the product. Therefore, it might be a good idea to invest in a system for automated weld integrity analysis. Even if the welds are deposited properly and have no visible structural flaws it is still useful to inspect them using an automated system. For example, suppose that the part’s blueprint calls for an 8mm weld at a certain location but the welder deposited only a 6mm weld. The 2mm difference may be quite difficult to detect with a naked eye (the typical inspection method). Furthermore, manually inspecting welded structures with multiple welds is often tedious and time consuming. The inspectors often fail to inspect all welds, especially if the parts contain thousands of welds. To counteract this, companies resort to 300% inspection levels (i.e., three different people inspect the same part one after another), which further increases the manufacturing costs. This paper describes an automated method for nondestructive evaluation of the quality of welds. The method uses a high-end stereo camera system to capture the 3D geometry of 1 Copyright c © 2010 by ASME welds. The geometrical properties of each weld are analyzed using a computer program, which extracts the weld angle and the radius of the weld at the weld toe. These weld parameters can, in turn, be used to estimate the quality of each weld as shown in the next section. Experimental results for T-fillet welds are reported. RELATED WORK Previous research has shown that the durability of a welded structure is dependent on the geometry of the weld [1]. The relationship between the stress at the weld toe and the nominal stress (the stress far away from the weld toe) is expressed with the stress concentration factor Kt . This relationship is given by the following equation [1]: σpeak = Kt ∗σn where σpeak is the stress at the weld toe, Kt is the stress concentration factor, and σn is the nominal stress. Kt is dependent on the weld angle and the radius of the weld toe [1]. There are two Kt factors, Kb t and K m t , which represent the bending stress and the membrane stress (or tension stress), respectively. The relationship between Kb t , K m t and the weld geometry is given by the following equations: Kb t = 1+0.512∗θ 0.572 ∗ ( t