advanced cutting, welding and inspection methods for

1. Fusion Engineering and Design 5152 (2000) 985991 Advanced cutting, welding and inspection methods for vacuum vessel assembly and maintenance L. Jones a, *, J.-P. Alle b , Ph. Aubert b , C. Punshon c , W. Danner a , V. Kujanpaa d , D. Maisonnier a , M. Serre e , G. Schreck f , M. Wykes g a EFDA, Garching, Germany b CEA, Arcueil, France c TWI, Cambridge, UK d VTT, Lappeenranta, Finland e CEA, Saclay f Underwater Technology Centre, Hanno6er, Germany g UKAEA, Culham Abstract ITER requires a 316 l stainless steel, double-skinned vacuum vessel (VV), each shell being 60 mm thick. EFDA (European Fusion Development Agreement) is investigating methods to be used for performing welding and NDT during VV assembly and also cutting and re-welding for remote sector replacement, including the development of an Intersector Welding Robot (IWR) [Jones et al. This conference]. To reduce the welding time, distortions and residual stresses of conventional welding, previous work concentrated on CO2 laser welding and cutting processes [Jones et al. Proc. Symp. Fusion Technol., Marseilles, 1998]. NdYAG laser now provides the focus for welding of the rearside root and for completing the weld for overhead positions with multipass lling. Electron beam (E-beam) welding with local vacuum offers a single-pass for most of the weld depth except for overhead positions. Plasma cutting has shown the capability to contain the backside dross and preliminary work with NdYAG laser cutting has shown good results. Automated ultrasonic inspection of assembly welds will be improved by the use of a phased array probe system that can focus the beam for accurate aw location and sizing. This paper describes the recent results of process investigations in this R&D programme, involving ve European sites and forming part of the overall VV/blanket research effort [W. Danner et al. This conference]. 2000 Elsevier Science B.V. All rights reserved. Keywords: NdYAG laser; Electron beam; Welding www.elsevier.com/locate/fusengdes 1. E-beam welding with reduced pressure (UKAEA/TWI, Cambridge) Since the sectors cannot be rotated to weld in the optimum position, the welding equipment must be capable of making satisfactory welds in all positions from down hand to overhead. A * Corresponding author. Present address: NET Team, Re- mote Handling, Boltzmannstrasse 2, 85748 Garching, Ger- many. Tel.: +49-89-32994278; fax: +49-89-32994198. E-mail address: [email protected] (L. Jones). 0920-3796/00/$ - see front matter 2000 Elsevier Science B.V. All rights reserved. PII: S0920-3796(00)00412-9

2. L. Jones et al. / Fusion Engineering and Design 5152 (2000) 985991986 reduced pressure vacuum chamber mounted with 100 kW electron gun on a rotary xture permitted all positional welding (Fig. 1). The best results were obtained by a partially penetrating weld, as a fully penetrating weld presents serious problems in dealing with the through-wall energy, which needs to be high for good root bead control. Further work was therefore carried out on this basis and rst results using small samples of 316-l Fig. 3. A 7 kW weld penetration versus speed top line is total penetration, middle is weld width and bottom line is weld width at 50% penetration. Fig. 1. Rotary xture for all positional e-beam welding. Fig. 4. A 7 kW welds at 19, 14 and 12 mm penetrations. Fig. 2. Welded joint incorporating a root tack weld and nishing e-beam weld. stainless steel demonstrated thickness penetration up to 60 mm easily in the at position. A joint between two blocks incorporating a root tack weld is shown in Fig. 2. Satisfactory welds were produced at a penetration depth of 40 mm for the vertical up welding position, 30 mm for the 45 to overhead position and 20 mm for the 10 to overhead position. The depth of successful weld penetration reduces as the weld path proceeds from downhand to overhead, due to the creation of large internal voids caused by the partial down- ward movements of the weld metal [1].

3. L. Jones et al. / Fusion Engineering and Design 5152 (2000) 985991 987 Fig. 5. A 3 kW lazer weld using wire lling on 6 mm plate with 1 mm gap. These preliminary results are considered to be sufciently encouraging to proceed to the second stage of the programme designed to optimise mul- tiposition welding. 2. Single-pass NdYAG welding (CEA, Arceuil) Previous results with lower powered YAG lasers indicated that welding of at least 12 mm with one pass should be possible with a new 7 kW combined laser system. Results using a single 4 kW YAG HAAS-TRUMP laser with focal length varied from 100 to 500 mm, gave penetration depths, respectively, from 4 to 8 mm with the fusion zone shapes moving from hemispheric to keyhole. To increase the available power, HAAS laser has combined the power from two separate lasers into two optical bres, assembled close together in the same connector so that, depending of the focal length, two adjacent welding spots, 300 microns to 1.5 mm apart are focussed in-line on the workpiece. The relationship between penetration depth and welding speed in 316 l stainless steel at a power of 7 kW is shown in Fig. 3 and three weld cross-sec- tions of 19, 14 and 12 mm penetration are shown in Fig. 4. The width of the weld is large enough to follow the joint, but narrow enough to minimise distortions of the structure after welding [2]. In the middle of the joint there remains a tendency to hot cracking, which will be controlled by changing the shape of the fusion zone, using an additional laser diode for pre- and post-heating, improving penetration and coupling. Fig. 6. Experimental set-up used with wire feed apparatus. Fig. 7. Optimisation trials of the gas ow shape by modifying the internal nozzel shape.

4. L. Jones et al. / Fusion Engineering and Design 5152 (2000) 985991988 Fig. 8. Cutting trials of a 60 mm thick plates at different cutting speeds. ried out for multi-pass welds with ller wire using a high power NdYAG laser HAAS 3006D [3]. To test this technique, the following different aspects need to be taken account: Filler wire feeding and its parameters (wire speed, wire angle, feeding point into the laser beam) Focus length of the focussing lens (in different passes) Form of the joints to be welded (need of air gap in different passes) Welding is performed by an accurate feed of the wire from the front edge of laser beam head. The feed rate is varied according to air gap and welding speed. Early results have successfully lled up to 1 mm gap width in 6 mm thick plate (Fig. 5). The experimental apparatus used with wire feed is shown in Fig. 6. Weld penetration is maximised to reduce the number of passes. The air gap is zero in the root pass and is increased in upper passes such that laser beam can be focused on the melt pool. In deeper passes a longer focal length is used than in the upper ones, where a weaving technique is appropriate. 4. NdYAG laser cutting (CEA, Saclay) A continuous, 4 kW NdYAG laser was used to cut 60 mm thick stainless steel plates with good cut quality (no oxidisation, parallel cross section shape, low surface roughness). A special cutting head has been developed, with an internal plane window capable of resisting the high pressure (30 bar) required for cutting trials. The material is cut by focussing the YAG laser beam at the surface of the workpiece, while removing simultaneously the molten debris with a jet of nitrogen gas aimed at the interaction point. As the speed quality and depth of the cut is strongly dependent on the aerodynamic condi- tions created by the nozzle, workpiece proximity and kerf shape, an investigation was carried out to improve the shape of the gas ow. Fig. 7 shows the visualisation of the nitrogen jet issuing from Fig. 9. Oscillating backside protection system. 3. NdYAG laser welding with ller wire (VTT, Lapeenranta) Narrow gap laser welding tests are being car-

5. L. Jones et al. / Fusion Engineering and Design 5152 (2000) 985991 989 Fig. 10. Design and characterisation of the circular phased array transducer used in immersion. two kinds of nozzle: one with a conical internal shape and on the left with an inner shape close to a Laval nozzle, which improves the gas ow jet by decreasing its diameter and increasing its effective length. A Laval nozzle is a supersonic nozzle originally developed for the study of boundary layer and shock interactions, and has a conver- gentdivergent form. Cutting trials were carried out on 20, 40 and 60 mm thick plates at a laser power of 4 kW with nitrogen assist at low and high pressure (3 and 30 bar). The 4 kW laser easily cut a thickness of 40 mm at a speed of about 80 mm min1 but could not cut through 60 mm thick plate, even though the speed was decreased to 5 mm min1 . Fig. 8 shows different kerfs produced at different cutting speeds. On the lower part of the plate, note the presence of slag, increasing with reduced cutting speed. Early cutting trials have claried the parameters to vary and the limits of achievable thickness able to be cut at various powers. The rst step is to optimise the nozzle internal shape as well as the cutting parameters. To improve the results, trials will be done with a gas mixture (nitrogen with a few percent of oxygen), as well as trials with a second gas nozzle aimed at the workpiece. To attain the goal of 60 mm cut thickness, two optical bres, each connected to its own laser (2 and 4 kW) will be combined into a new high-pressure cutting beam. 5. Plasma arc cutting (UWT, Hannover) An alternative to laser cutting of the wall section is the more conventional plasma cutting and following successful earlier work [2], the contain- ment of the dross in the narrow (25 mm) gap was investigated. A prototype reciprocating backside protection plate for vertical and horizontal cuts was procured and different high temperature resistant materials for melt ow absorption tested. The most effective tile protection material was carbon-bre-composite (CFC) and to reduce its Fig. 11. Interaction between a UT beam and a 15 mm high notch.

6. L. Jones et al. / Fusion Engineering and Design 5152 (2000) 985991990 Fig. 12. Inspection of 15 mm depth notches using adaptive delay laws. oxidation, Helium gas cooling was added. Wire brushes seal the sides and a dross reservoir placed at the bottom of the protection system collected the bigger particles. A nearly straight and plain kerf surface of one cut side was achieved by angling the cutting head. A torch protection cap of a special heat resistant, machineable material (Rescor 960) was used for successful plasma cuts with overhead workpiece position. Although the concept of an oscillating protection plate showed itself to be suitable, it will be difcult to adapt the backside protection system including a wire brush sealing, dross reservoir and pneumatically driven oscillating system as shown in Fig. 9 into a curved geometry. In view of the very good thermal resistance of CFC plates with- out an oscillating motion, the modied protection system will incorporate a xed backplate. 6. UT with phased array (Saclay) The objective is to optimise an ultrasonic test- ing method based on phased array techniques that may be applied industrially to the inter-sector welding during the vessel assembly and maintenance. Notches may appear on both walls of the vessel and the ultrasonic inspection has to detect, characterise and size these defects. Conventional ultrasonic transducer cannot achieve both detection and characterisation. For detection, the beam has to be long enough to cover the whole thickness of the plate after a beam reection from the backwall. Such a beam cannot be used to locate and accurately characterise the detected notches. Phased array probes are well adapted to shape the beam in order to optimise detection and characterisation in the same process. Initial detection tests were carried out using 45 shear waves at a xed focus depth. Subsequently the focus depth was changed to the detected aw depth so as to allow accurate sizing and location. The phased array probe (Fig. 10) used for these tests is com- posed of circular rings emphasising the focussing of the beam along the probe axis. This transducer generates 2 MHz wave beam able to focus from 30 to 120 mm depth. This probe immersed in water acted on a 60 mm thick planar mock-up containing no welds, but with breaking notches representative of lack of fusion along weld sides positioned on both walls of the specimen. The notches are 20 mm wide and their depth varies from 5 to 15 mm. All the notches were detected and characterised. A simulation model added to a segmentation process discriminates the diffraction echoes caused by the wall roughness. Fig. 11 presents simulated inspection results obtained on a 15 mm deep breaking notch. The transducer positions related to each simulated echo are identied using the beam orientation. Diffraction and corner echoes are indicated with the same shade on Figs. 11 and 12. The size of the notch was determined by measuring the distance separating the diffraction echoes from the corner

7. L. Jones et al. / Fusion Engineering and Design 5152 (2000) 985991 991 echo. The images resulting from the inspection of 15 mm deep notches positioned on the surface and on the back wall, presented in Fig. 3, show the efciency of the process. An adaptive delay law was applied in each case to focus the ultrasonic beam at the notch depth. The next step of the study will apply this phased array technique to two planar mock-ups containing realistic welds and different kinds of articial defects, both surface breaking and non- surface breaking notches, and volumetric defects. 7. Conclusions The results from the rst-stage work presented above represents an optimistic indication that power-beam methods may be successfully and economically used for VV assembly and maintenance. This work is continuing to rene and qualify these processes and recent indication is that the pace of solid state laser power and quality improvement continues to accelerate, giving possibilities in the near future of even deeper welding and cutting by laser. There remain many issues such as site working and integration and the radiation environment which an industrial study to study these issues in detail will address. References [1] L. Jones et al., Design of the intersector welding robot for vacuum vessel assembly and maintenance. This conference. [2] L. Jones et al., Development of cutting and welding methods for thick-walled stainless steel support and contain- ment structures for ITER. Proc. Symposium on Fusion Technology, Marseilles, 1998. [3] W. Danner et al., Major achievements of the European shield blanket R&D during the ITER EDA, and their relevance for future next step machines. This conference. .