A survey of robotized seam-tracking techniques was conducted in preparation for a project consisting of using an infrared camera on a robot for on one side collect continuous weld images for NDT inspection and on the other one track the weld joint. The tracking system can be used to discover the position of the weld bead without previous information. The robust system is outlined, along with its merits and disadvantages.

The implementation of robotic technology into welding processes is made difficult by the inherent process variables of part location, fit up, orientation and repeatability. Considering these aspects, to ensure weld reproducibility consistency and quality, advanced adaptive control techniques are essential. These involve not only the development of adequate sensors for seam tracking and joint recognition but also developments of overall machines with a level of artificial intelligence sufficient for automated welding. The development of such a prototype system which utilizes a manipulator arm, ultrasonic sensors and a transistorised welding power source is outlined. This system incorporates three essential aspects. It locates and tracks the welding seam ensuring correct positioning of the welding head relatively to the joint preparation. Additionally, it monitors the joint profile of the molten weld pool and modifies the relevant heat input parameters ensuring consistent penetration, joint filling and acceptable weld bead shape. Finally, it makes use of both the above information to reconstruct three-dimensional images of the weld pool silhouettes providing in-process inspection capabilities of the welded joints. Welding process control strategies have been incorporated into the system based on quantitative relationships between input parameters and weld bead shape configuration allowing real-time decisions to be made during the process of welding, without the need for operation intervention.

A recent method in aero engine production is to fabricate components from smaller pieces, rather than machining them from large castings. This has made laser beam welding popular, offering high precision with low heat input and distortion, but also high productivity. At the same time, the demand for automation of production has increased, to ensure high quality and consistent results. In turn, the need for sensors to monitor and control the laser welding process is increasing. In laser beam welding without filler material, the gap between the parts to be joined must be narrow. Optical sensors are often used to measure the gap, but with precise machining, it may become so narrow that it is difficult to detect, with the risk of welding in the wrong position. This kind of problems can cause severe welding defects, where the parts are only partially joined without any visible indication. This thesis proposes the use of an inductive sensor with coils on either side of the gap. Inducing currents into the metal, such a sensor can detect even gaps that are not visible. The new feature of the proposal is based on using the complex response of each coil separately to measure the distance and height on both sides of the gap, rather than an imbalance from the absolute voltage of each coil related to gap position. This extra information allows measurement of gap width and misalignment as well as position, and decreases the influence from gap misalignment to the position measurement. The sensor needs to be calibrated with a certain gap width and height alignment. In real use,these will vary, causing the sensor to be less accurate. Using initial estimates ofthe gap parameters from the basic sensor, a model of the response can be used to estimate the measurement error of each coil, which in turn can be used for compensation to improve the measurement of the gap properties.The properties of the new method have been examined experimentally, using a precise traverse mechanism to record single coil responses in a working range around a variable dimension gap, and then using these responses to simulate a two coil probe. In most cases errors in the measurement of weld gap position and dimensions are within 0.1 mm.The probe is designed to be mounted close to the parts to be welded, and will work in a range of about 1 mm to each side and height above the plates. This is an improvement over previous inductive sensors, that needed to be guided to the mid of the gap by a servo mechanism.

Syftet med denna studie är att jämföra två olika lösningar för fogföljning med svetsrobot. I nuläget används en extern sjunde axel utgiven av Permanova, denna jämförs med programvara från ABB vid namn EGM. Främst skall dödtid och tidskonstant fastställas och jämföras för systemen. Problematiken med detta var främst osäkerheten i den okulära inspektion av graferna som gav värdena. Ytterligare ett problem, som dessutom försvårade avläsningarna, var att det förekom störningar i processen som orsakade förvrängda kurvor. Experimenten utfördes på en modell av den riktiga processen. I den modellerade processen anslöts ingen svetsanordning till roboten, istället monterades en laseravståndsgivare som mätte mot ett parallellt metallblock. Den sjunde axeln var ansluten vid alla test för att säkerställa att resultaten var jämförbara med hänseende till vikten som belastar robotarmen. Varje testfall hade olika parametrar, så som hastighet och avstånd från robotbas. Flera experiment gjordes med samma parametrar för att säkerställa bra resultat. Analys av insamlade data gjordes bland annat genom att liknande stegsvar av första ordningen jämfördes med experimentets stegsvar. Det estimerade stegsvarets värden finjusterades tills dess att de båda stegsvaren var någorlunda lika, dock är mänsklig faktor mycket relevant i denna analys. Resultatet visade en klar fördel med Permanovas sjunde axel i både minskad dödtid och tidskonstant. Exempelvis var det filtrerade medelvärdet för dödtiden med EGM 266 ms och en standardavvikelse på 20,8 ms, medan det för Permanova är nere på 11,5 ms och standardavvikelse på 2,0 ms. De filtrerade tidskonstanterna följde samma trend och uppgick till 68 ms, standardavvikelse 13,5 ms, för EGM respektive 33 ms och standardavvikelse 4,8 ms för Permanova. Det var överraskande att skillnaden mellan EGM och Permanovas sjunde axel var så pass stor, men sjunde axeln verkar helt klart vara bäst lämpad för en sådan här process på grund av dess mycket snabbare dödtid och tidskonstant. Då EGM presterade så mycket sämre anses den inte lämplig för applikationer i lasersvetsning på PTC. / The purpose of this study is to compare solutions for seam tracking with a welding robot. At the moment the used solution is an external seventh axis made by Permanova which is going to be compared to software made by ABB called EGM. The primary comparisons shall be done regarding the dead time and the time constant of the two processes. A primary problem with this was the uncertainty that comes with ocular inspection of the graphs in order to extract relevant values. Another problem, that also made construing the curves harder, was that there were some disturbances that caused distorted curves. Experiments were done on a model of the actual process, meaning that no welding equipment was connected to the robot in this model process. Instead a laser sensor was mounted onto the robot that measured against a parallel metal block. The seventh axis was connected for all tests to make sure that the results were comparable when it comes to the weight that the robot arm is burdened with. Each experiment had different parameters, such as speed and distance from the robot's base. Several tests were made with the same parameters to make sure that good results were had. One method used for analysis of gathered data was comparing a similar first order system to the step response of different experiments. The values of this approximate step response were tweaked until both step responses were fairly similar, however, the human factor is very relevant in this analysis. The results showed a clear advantage towards Permanova's seventh axis in both decreased dead time and time constant. For example, the filtered mean value for the dead time with EGM was 266 ms and a standard deviation of 20,8 ms, while Permanova clocked in at 11,5 ms with a standard deviation of 2,0 ms. The filtered time constants followed the same trend and were at 68 ms, standard deviation of 13,5 ms, for EGM and 33 ms, standard deviation of 4,8 ms, for Permanova. It was quite surprising how big the difference was between EGM and Permanova's seventh axis, but the seventh axis was without a doubt the solution best suited to this kind of process. Due to the lacking performance from EGM it is determined to be unfit for laser welding applications at PTW.

With the growth in pipeline installations all over the world, there is a great demand for highly productive and robust welding systems. Mechanised pipe welding has been developed over the last 50 years and the present focus is towards development of automated pipeline welding systems. Pipeline welding automation is aimed at reducing costs and improving the installation quality. To attain fully automated pipe welding systems there is a need to rely on sensors and controls systems to mimic human like capabilities, such as visual inspection, in real time. The key aim of this work is to develop and evaluate methods of automatic assessment of weld bead shape and quality during narrow gap GMAW of transmission pipelines. This implies that the measured bead profile will be assessed to determine whether the bead shape will cause defects when the subsequent pass is deposited. Different approaches have been used to conquer the challenge that is emulating human reasoning, all with different objectives in mind. In spite of extensive literature research performed, very little information was found concerning the real time determination and assessment of bead shape quality and none of it was reported to be applied successfully to the pipeline industry. Despite the continuous development of laboratory laser vision systems commercial ones have been on the market for decades, some specifically developed for the welding application. Laser vision sensor systems provide surface profile information, and are the only sensors which can satisfactorily measure bead profile on a narrow groove. In order to be able to use them to automatically assess weld bead shape and quality, a deep understanding of their characteristics and limitations needs to be achieved. Once that knowledge was attained it was then applied to determine the best sensor configuration for this purpose. After that the development of human like judgment algorithms were developed to accomplish the aim that was set. Empirical rules were obtained from an experienced welder regarding the acceptability of bead shapes and were then applied in the developed system with good results. To scientifically evaluate and determine the rules to use in this system, further experiments would be required. The output of the system developed showed very accurate, reliable and consistent results that were true to the external measurements and comparisons performed. The developed system has numerous applications in the pipeline industry and it could easily be implemented on commercial systems.

Tungsten Inert Gas (TIG) welding is extensively used in aerospace applications, due to its unique ability to produce higher quality welds compared to other shielded arc welding types. However, most TIG welding is performed manually and has not achieved the levels of automation that other welding techniques have. This is mostly attributed to the lack of process knowledge and adaptability to complexities, such as mismatches due to part fit-up. Recent advances in automation have enabled the use of industrial robots for complex tasks that require intelligent decision making, predominantly through sensors. Applications such as TIG welding of aerospace components require tight tolerances and need intelligent decision making capability to accommodate any unexpected variation and to carry out welding of complex geometries. Such decision making procedures must be based on the feedback about the weld profile geometry. In this thesis, a real-time position based closed loop system was developed with a six axis industrial robot (KUKA KR 16) and a laser triangulation based sensor (Micro-Epsilon Scan control 2900-25).