Supervised deep neural networks have been successfully applied to many real-world measurement applications. However, their success relies on labeled data which is expensive and time-consuming to obtain, especially when domain expertise is required. For this reason, researchers have turned to semi-supervised learning for image classification tasks. Semi-supervised learning uses structural assumptions to automatically leverage unlabeled data, dramatically reducing manual labeling efforts. We conduct our research based on images from Enclosures Direct Inc. (EDI) which is a manufacturer of enclosures used to house and protect electronic devices. Their industrial robotics utilizes a computer vision system to guide a robot in a welding application employing a laser and a camera. The laser is combined with an optical line generator to cast a line of structured light across a joint to be welded. An image of the structured light is captured by the camera which needs to be located in the image in order to find the desired coordinate for the weld seam. The existing system failed due to the fact that the traditional machine vision algorithm cannot analyze the image correctly in unexpected imaging conditions or during variations in the manufacturing process. In this thesis, we purpose a novel algorithm for semi-supervised key-point detection for seam placement by a welding robot. Our deep learning based algorithm overcomes unfavorable imaging conditions providing faster and more precise predictions. Moreover, we demonstrate that our approach can work with as few as ten labeled images accepting a reduction of detection accuracy. In addition, we also purpose a method that can utilize full image resolution to enhance the accuracy of the key-point detection.

Em geral, um problema importante em um sistema de soldagem robotizado, ou com manipuladores automáticos, é a necessidade de uma trajetória ser reprogramada quando a forma ou a posição das peças mudam. Como solução a esse problema é proposto um método que utiliza técnicas de visão computacional, para assim detectar a trajetória de soldagem em juntas de topo. O método deve ser dinâmico permitindo a identificação de juntas de topo com cantos não próximos, linhas não retas e tamanhos diferentes entre as peças, assim como apresentar robustez contra variáveis desfavoráveis do ambiente industrial como riscos nas peças ou no fundo e mudanças de luminosidade. A trajetória é detectada baseando-se na linha de bordas da imagem global e adicionalmente trabalha-se sobre duas características da linha de solda: distância quase constante entre linhas paralelas e pixels de linha de solda com menor intensidade de luminosidade que as bordas. Uma etapa é proposta para robustez do método sobre linha de bordas descontinuas e possíveis erros em ponto inicial e final de solda. O trabalho proposto é validado com distintas configurações de junta de topo, como com cantos de peças não coincidentes, linha de solda não reta e diferentes orientações de peças. Os pontos da trajetória de solda obtidos são comparados com uma linha de solda considerada ideal, resultando em uma média e desvio padrão geral inferiores à um valor de 0,5 mm. Uma validação experimental é realizada com uma solda executada por um robô industrial seguindo os pontos de solda detectados, com os resultados demostrando que o método efetivamente pode definir uma trajetória de solda para aplicações industriais robotizadas. / In general, one of the most important problems in a robotic welding system, or in automatic manipulations, is the requirement of a path trajectory to be reprogramed when the shape or the position of the welding pieces are changed. In order to detect the welding seam in butt joints, a method that makes use of computational vision techniques is proposed. The method needs to be dynamic against variation in the configuration of the pieces, so as curved or not straight seam lines, not coincident corners; beside of that, it must to present robustness against unfavorable variables of the industrial environment, so as scratches in the pieces or illumination changes. Two features of the welding line are taken into account to develop the method: almost constant distance between parallel seam lines and darker pixels in the center of the seam than in the borders. Moreover a robustness step is proposed over two weaknesses of the method: discontinuities in the edges of the welding line and possible errors in the location of the start and end welding points. The validation step of the method involves different configuration of butt joints, as pieces without corner being coincident, not straight welding line and different orientations. The points of the welding seam detected by the method, are compared against a set of welding points considered as an ideal set of points, getting as results a mean and standard deviation lower than ± 0.5 mm. An experimental test is carried out by an industrial robot that welds two pieces following the welding line points, with the results showing that the method can effectively define a welding trajectory for industrial robotics applications.

The investigation of reactor pressure vessel (RPV) material from the decommissioned Greifswald nuclear power plant representing the first generation of Russian-type WWER-440/V-230 reactors offers the opportunity to evaluate the real toughness response. The Greifswald RPVs of 4 units represent different material conditions as follows: • Irradiated (Unit 4), • irradiated and recovery annealed (Units 2 and 3), and • irradiated, recovery annealed and re-irradiated (Unit1). The recovery annealing of the RPV was performed at a temperature of 475° for about 152 hours and included a region covering ±0.70 m above and below the core beltline welding seam. Material samples of a diameter of 119 mm called trepans were extracted from the RPV walls. The research program is focused on the characterisation of the RPV steels (base and weld metal) across the thickness of the RPV wall. This report presents test results measured on the trepans from the beltline welding seam No. SN0.1.4. and forged base metal ring No. 0.3.1. of the Units 1 2 and 4 RPVs. The key part of the testing is focussed on the determination of the reference temperature T0 of the Master Curve (MC) approach following the ASTM standard E1921 to determine the facture toughness, and how it degrades under neutron irradiation and is recovered by thermal annealing. Other than that the mentioned test results include Charpy-V and tensile test results. Following results have been determined: • The mitigation of the neutron embrittlement of the weld and base metal by recovery annealing could be confirmed. • KJc values of the weld metals generally followed the course of the MC though with a large scatter. • There was a large variation in the T0 values evaluated across the thickness of the multilayered welding seams. • The T0 measured on T-S oriented SE(B) specimens from different thickness locations of the welding seams strongly depended on the intrinsic structure along the crack front. • The reference temperature RT0 determined according to the “Unified Procedure for Lifetime Assessment of Components and Piping in WWER NPPs - VERLIFE” and the fracture toughness lower bound curve based thereon are applicable on the investigated weld metals. • A strong scatter of the fracture toughness KJc values of the recovery annealed and re-irradiated and the irradiated base metal of Unit 1 and 4, respectively is observed with clearly more than 2% of the values below the MC for 2% fracture probability. The application of the multimodal MC-based approach was more suitable and described the temperature dependence of the KJc values in a satisfactory manner. • It was demonstrated that T0 evaluated according to the SINTAP MC extension represented the brittle fraction of the data sets and is therefore suitable for the nonhomogeneous base metal. • The efficiency of the large-scale thermal annealing of the Greifswald WWER 440/V230 Unit 1 and 2 RPVs could be confirmed.

Em geral, um problema importante em um sistema de soldagem robotizado, ou com manipuladores automáticos, é a necessidade de uma trajetória ser reprogramada quando a forma ou a posição das peças mudam. Como solução a esse problema é proposto um método que utiliza técnicas de visão computacional, para assim detectar a trajetória de soldagem em juntas de topo. O método deve ser dinâmico permitindo a identificação de juntas de topo com cantos não próximos, linhas não retas e tamanhos diferentes entre as peças, assim como apresentar robustez contra variáveis desfavoráveis do ambiente industrial como riscos nas peças ou no fundo e mudanças de luminosidade. A trajetória é detectada baseando-se na linha de bordas da imagem global e adicionalmente trabalha-se sobre duas características da linha de solda: distância quase constante entre linhas paralelas e pixels de linha de solda com menor intensidade de luminosidade que as bordas. Uma etapa é proposta para robustez do método sobre linha de bordas descontinuas e possíveis erros em ponto inicial e final de solda. O trabalho proposto é validado com distintas configurações de junta de topo, como com cantos de peças não coincidentes, linha de solda não reta e diferentes orientações de peças. Os pontos da trajetória de solda obtidos são comparados com uma linha de solda considerada ideal, resultando em uma média e desvio padrão geral inferiores à um valor de 0,5 mm. Uma validação experimental é realizada com uma solda executada por um robô industrial seguindo os pontos de solda detectados, com os resultados demostrando que o método efetivamente pode definir uma trajetória de solda para aplicações industriais robotizadas. / In general, one of the most important problems in a robotic welding system, or in automatic manipulations, is the requirement of a path trajectory to be reprogramed when the shape or the position of the welding pieces are changed. In order to detect the welding seam in butt joints, a method that makes use of computational vision techniques is proposed. The method needs to be dynamic against variation in the configuration of the pieces, so as curved or not straight seam lines, not coincident corners; beside of that, it must to present robustness against unfavorable variables of the industrial environment, so as scratches in the pieces or illumination changes. Two features of the welding line are taken into account to develop the method: almost constant distance between parallel seam lines and darker pixels in the center of the seam than in the borders. Moreover a robustness step is proposed over two weaknesses of the method: discontinuities in the edges of the welding line and possible errors in the location of the start and end welding points. The validation step of the method involves different configuration of butt joints, as pieces without corner being coincident, not straight welding line and different orientations. The points of the welding seam detected by the method, are compared against a set of welding points considered as an ideal set of points, getting as results a mean and standard deviation lower than ± 0.5 mm. An experimental test is carried out by an industrial robot that welds two pieces following the welding line points, with the results showing that the method can effectively define a welding trajectory for industrial robotics applications.

Em geral, um problema importante em um sistema de soldagem robotizado, ou com manipuladores automáticos, é a necessidade de uma trajetória ser reprogramada quando a forma ou a posição das peças mudam. Como solução a esse problema é proposto um método que utiliza técnicas de visão computacional, para assim detectar a trajetória de soldagem em juntas de topo. O método deve ser dinâmico permitindo a identificação de juntas de topo com cantos não próximos, linhas não retas e tamanhos diferentes entre as peças, assim como apresentar robustez contra variáveis desfavoráveis do ambiente industrial como riscos nas peças ou no fundo e mudanças de luminosidade. A trajetória é detectada baseando-se na linha de bordas da imagem global e adicionalmente trabalha-se sobre duas características da linha de solda: distância quase constante entre linhas paralelas e pixels de linha de solda com menor intensidade de luminosidade que as bordas. Uma etapa é proposta para robustez do método sobre linha de bordas descontinuas e possíveis erros em ponto inicial e final de solda. O trabalho proposto é validado com distintas configurações de junta de topo, como com cantos de peças não coincidentes, linha de solda não reta e diferentes orientações de peças. Os pontos da trajetória de solda obtidos são comparados com uma linha de solda considerada ideal, resultando em uma média e desvio padrão geral inferiores à um valor de 0,5 mm. Uma validação experimental é realizada com uma solda executada por um robô industrial seguindo os pontos de solda detectados, com os resultados demostrando que o método efetivamente pode definir uma trajetória de solda para aplicações industriais robotizadas. / In general, one of the most important problems in a robotic welding system, or in automatic manipulations, is the requirement of a path trajectory to be reprogramed when the shape or the position of the welding pieces are changed. In order to detect the welding seam in butt joints, a method that makes use of computational vision techniques is proposed. The method needs to be dynamic against variation in the configuration of the pieces, so as curved or not straight seam lines, not coincident corners; beside of that, it must to present robustness against unfavorable variables of the industrial environment, so as scratches in the pieces or illumination changes. Two features of the welding line are taken into account to develop the method: almost constant distance between parallel seam lines and darker pixels in the center of the seam than in the borders. Moreover a robustness step is proposed over two weaknesses of the method: discontinuities in the edges of the welding line and possible errors in the location of the start and end welding points. The validation step of the method involves different configuration of butt joints, as pieces without corner being coincident, not straight welding line and different orientations. The points of the welding seam detected by the method, are compared against a set of welding points considered as an ideal set of points, getting as results a mean and standard deviation lower than ± 0.5 mm. An experimental test is carried out by an industrial robot that welds two pieces following the welding line points, with the results showing that the method can effectively define a welding trajectory for industrial robotics applications.

The investigation of reactor pressure vessel (RPV) material from the decommissioned Greifswald nuclear power plant representing the first generation of Russian-type WWER-440/V-230 reactors offers the opportunity to evaluate the real toughness response. The Greifswald RPVs of 4 units represent different material conditions as follows: • Irradiated (Unit 4), • irradiated and recovery annealed (Units 2 and 3), and • irradiated, recovery annealed and re-irradiated (Unit1). The recovery annealing of the RPV was performed at a temperature of 475° for about 152 hours and included a region covering ±0.70 m above and below the core beltline welding seam. Material samples of a diameter of 119 mm called trepans were extracted from the RPV walls. The research program is focused on the characterisation of the RPV steels (base and weld metal) across the thickness of the RPV wall. This report presents test results measured on the trepans from the beltline welding seam No. SN0.1.4. and forged base metal ring No. 0.3.1. of the Units 1 2 and 4 RPVs. The key part of the testing is focussed on the determination of the reference temperature T0 of the Master Curve (MC) approach following the ASTM standard E1921 to determine the facture toughness, and how it degrades under neutron irradiation and is recovered by thermal annealing. Other than that the mentioned test results include Charpy-V and tensile test results. Following results have been determined: • The mitigation of the neutron embrittlement of the weld and base metal by recovery annealing could be confirmed. • KJc values of the weld metals generally followed the course of the MC though with a large scatter. • There was a large variation in the T0 values evaluated across the thickness of the multilayered welding seams. • The T0 measured on T-S oriented SE(B) specimens from different thickness locations of the welding seams strongly depended on the intrinsic structure along the crack front. • The reference temperature RT0 determined according to the “Unified Procedure for Lifetime Assessment of Components and Piping in WWER NPPs - VERLIFE” and the fracture toughness lower bound curve based thereon are applicable on the investigated weld metals. • A strong scatter of the fracture toughness KJc values of the recovery annealed and re-irradiated and the irradiated base metal of Unit 1 and 4, respectively is observed with clearly more than 2% of the values below the MC for 2% fracture probability. The application of the multimodal MC-based approach was more suitable and described the temperature dependence of the KJc values in a satisfactory manner. • It was demonstrated that T0 evaluated according to the SINTAP MC extension represented the brittle fraction of the data sets and is therefore suitable for the nonhomogeneous base metal. • The efficiency of the large-scale thermal annealing of the Greifswald WWER 440/V230 Unit 1 and 2 RPVs could be confirmed.

Darbe tiriami mažaanglio plieno S235J2 (LST EN 10025:2004) suvirintų siūlių pasikeitimai, plieną apvirinant skirtingais glaistytaisiais elektrodais ir suvirinimo režimais. Tyrimų metu buvo naudojami AV-23 ir AV-66 glaistytieji elektrodai bei keičiamas trumpojo jungimo srovės stipris. Nustatytas ryšys tarp metalo įvirinimo gylio, rumbelės aukščio, išlydytų lašų dydžio, išlydytų lašų kiekio, trumpojo jungimo trukmės ir trumpojo jungimo srovės dydžio. Gauti tyrimų rezultatai, nubraižyti priklausomybių grafikai ir pateiktos išvados. Darbą sudaro: įvadas, literatūros apžvalga, eksperimentuose naudojama įranga ir medžiagos, tyrimo metodika, tyrimų metu gauti rezultatai, grafikai, išvados, literatūra. Darbo apimtis: 83 p. teksto, 73 iliustr., 18 lent., 23 bibliografiniai šaltiniai. / In final work was analyzed the changes of seams properties of steel S235J2 (LST EN 10025:2004), produced by welding with different technology are discussed. Samples were welded by hand arc welding, using AV-23 and AV-66 welding electrodes. The welded seams were made by changing short-circuit current. The duration of weld depth, ridge height, metal drop size, metal drop frequency, short-circuit time and short-circuit current value are determinated. The results of the research are given in graphs, final conclusions are also written. Structure: the introduction, survey of the related literature, equipment and materials, description of the methodology, results of research, graphs of the results, conclutions, references. Thesis consist of: 83 p. text, 73 pictures, 18 tables, 23 bibliographical entries.