Remote Crack Repair Welding using High Power Nd:YAG Lasers

N C Sekhar and P A Hilton, TWI, and M Tilsley, Magnox Generation BNFL, Berkeley, Gloucestershire, GL139PB,UK

Presented at ICALEO 2000, 19th International Congress on Applications of Lasers and Electro-Optics, Hyatt Regency, Dearbon, Michigan, USA, 2-5 October 2000

Abstract

The capabilities of fibre delivered laser power have been demonstrated in many successful manufacturing applications in a wide range of industry sectors. These laser systems also offer significant possibilities for remote processing applications and in particular, repair operations. This paper describes work conducted with a 4 kW continuous wave fibre delivered Nd:YAG laser, to investigate the feasibility of performing an in situ repair, on a low silicon content mild steel component used in the core region of a gas cooled Magnox reactor. Due to the high pressures and temperatures experienced in service, in addition to the environmental conditions, these components are susceptible to oxide build up. The design of the component is such that the oxide can initiate cracking in the arc welds used in the original manufacturing process. Procedures are described which allow the same laser source to both remove the oxide layer from the component and subsequently heal the cracked region. Single pass and weave welding techniques were evaluated as well as the addition of filler wire material in order to provide weld reinforcement. The potential advantages of using such a laser procedure over traditional repair methodologies are outlined.

Introduction

The uranium contained in the UK's 'Magnox' type nuclear reactor fuel cells is sealed into cans to prevent the escape of fission products, since excessive fission products in the reactor cooling gas stream would constitute a radiation and health hazard and would hinder maintenance of the boilers. Therefore, it is necessary to have a system that will provide a warning of any developing leaks in the fuel cans at an early stage, so that gross contamination of the gas circuit can be avoided.

The 'burst can detection' (BCD) system used in the Magnox reactors allows for monitoring of the coolant gas as it emerges from the top of the reactor core. It can sample gas from individual channels and in the event of a leakage of fission products into the gas, provides immediate information to the reactor controller. Inside the reactor pressure vessel (RPV), BCD gas sampling lines are suspended, allowing the gas to be sampled at various locations. These sample lines transport the gas to the measuring equipment located outside the pressure vessel. The sample lines are supported in the RPV by, amongst other things, free hangers. Figure 1 shows a photograph of four BCD hangers in situ, taken by a remote camera. The BCD hangers are constructed from two plates of thickness about 4mm, arc welded on each side of a spacer of about 6mm thickness.

As part of routine and detailed reactor inspections, the presence of cracks in some of the BCD hangers is observed. These cracks present a low probability of BCD failure and in addition any failure of the BCD system would result in a safe shutdown of the reactor. However, there would be a commercial penalty arising from BCD failure, due to loss generation and as such a programme of weld reinforcement/repair is undertaken.

The hangers are susceptible to oxide jacking at the interface between the side plates and the spacer, which can cause the welds to crack. Figure 2 shows a close up of the BCD hanger, showing weld cracking. Oxide jacking is a phenomenon caused by the build-up of oxide on mild steel in a carbon dioxide atmosphere at high pressure and temperature. As the oxide occupies more volume than the original material, stresses can be generated that force components apart. This could lead to cracking in welds or fracture in bolts. Repair of certain components may be required in-situ, so that the effective service life is prolonged.

Various methodologies ^[1-4] have been developed for the repair of components in nuclear power plants. Most of these involve arc welding methods but the use of fibre delivered laser systems for these tasks has been on the increase due to the advantages that laser power could provide. The objectives of the work reported here were to develop for the BCD hanger, procedures for both crack repair and weld reinforcement using a fibre delivered Nd:YAG laser beam.

Approach

Dummy samples were manufactured, oxidised and cracked, for the repair experiments. They were fabricated using a low silicon content mild steel and the initial welding passes were made by manual metal arc welding. A 2.4 mm ESAB OK 46.00 mild steel rutile electrode was used for this purpose. Cracking was then induced in the samples by oxidising in an autoclave. When removed from the autoclave, the samples were generally covered in an oxide layer of approximately 0.4 mm thickness.

A Multiwave Auto ^TM Nd:YAG laser, manufactured by GSI Lumonics Ltd was used for the repair processing. The laser produced infra-red light of wavelength 1.064µm, with a maximum continuous average power at the workpiece of approximately 4kW. A step index fibre optic of core diameter 0.6 mm and length 30m transmitted the beam to the fibre output housing. The output housing consisted of a collimating and focussing lens arrangement that resulted in a focussed spot of nominal diameter 0.6 mm. The output housing was mounted on a Kawasaki Heavy Industries JS-6 arc-welding robot. Welds were made by moving the output housing over the stationary workpiece.

The specimens were held vertically in a vice (akin to the position they would take in the reactor pressure vessel, which can be seen in Figure 1). The welding head was positioned in relation to the workpiece as shown in Figure 3. All welding was performed in the vertical-up position. Filler wire, when used, was fed into the leading edge of the molten pool at an angle of approximately 60° to the laser head, using a MIG wire feeder system. In the repair procedure, an A18 mild steel MIG filler wire, 1.2 mm diameter, was used. A coaxial nozzle with a 20mm diameter outlet and 5mm stand-off distance, attached to the beam delivery system, was used to protect the weld zone from oxidation during processing. In addition, the coaxial nozzle was also used for plume control. Helium gas (at flow rates of 50 lpm) was used as the processing gas.

Two possibilities for repairing the cracks and thus increasing the service life of the component were evaluated. The first involved use of the laser beam to simply re-melt the crack formed in the weld metal of the original arc welds and the second involved the use of filler wire to provide a degree of reinforcement to the original weld. Test samples had been provided with oxide and with the external oxide layer removed. The possibility of using the laser beam to spall off the oxide layer prior to repair was also investigated. In order to establish processing conditions for crack re-melting, several bead on plate melt runs were performed on samples of the low silicon content mild steel, at the laser power available.

Complete repair can only be obtained by a full re-melting of the crack and the region surrounding it. A major difficulty in achieving this was that the origin and the direction of propagation of the cracks were not known and differed from sample to sample. Based on this knowledge and the results of the bead on plate trials, five different repair possibilities were evaluated, these being: use of linear, single and double pass autogenous re-melting of the cracked region, use of linear, single and double pass re-melting with wire feed addition and use of a weave pass technique with wire feed.

Results and analysis

Most of the repaired samples were sectioned, polished and prepared for optical metallography. These cross sections were used to determine the resulting weld profile and to observe the degree of crack repair.

In the first instance a technique to remove the oxide layer from the sample surface was developed. As the objective was to remove the surface oxide and not melt the base metal, a lower laser power (1 kW at the work-piece) was used. With laser focus at the surface and scanning speeds of 2 m/min, the oxide spalled off readily but melting of the base material was also observed. This effect could be minimised by using a lower laser power, using a higher scanning speed or by using a defocussed laser beam. The last option was adopted. With the laser beam focus 20 mm above the surface, the 0.4 mm thick oxide spalled off as flakes, over a width of ~15 mm on each side of the impingement point of the laser beam, at scanning speeds of 2 m/min. No signs of melting of the base metal were observed under these conditions. The differential thermal cycling of the oxide and the substrate (base metal) and the difference in the thermal expansion coefficients between the oxide and the substrate (base metal) led to the de-cohesion of the oxide layer when the laser was scanned over the sample.

The results of using a single repair pass without wire feed addition, on a sample with oxide can be seen in Figure 4. Onto this weld cross section, for the sake of clarity, have been drawn the extremities of the original arc weld (bold line), the extremities of the laser repair weld (dotted line) and a possible position of the original crack, as indicated by the arrow. The penetration of the laser repair is typical of the deepest available using the available laser parameters (laser focus at the surface). In this particular case it can be seen that the chosen position of the re-melted track did not completely include all the cracked area. This micrograph also clearly shows the gap between the side member and spacer of the hanger, where oxide build-up has initiated the crack.

Similarly, the oxide layer on the outside edges of the sample is also visible in the cross section. The resultant 'weld metal' profile also exhibited some minor undercut in this case. In similar repairs made with the oxide layer removed, the penetration of the laser beam was in fact lower, attributed to the fact that the coupling of the laser beam was assisted by the oxide layer.

In a single pass repair made using filler wire addition on a sample with oxide removed, a highly convex and wider bead profile was obtained but with less penetration. In this case the reduced penetration was due to the use of a defocussed laser beam. The section taken ( Figure 5) indicates the crack was successfully repaired. Since welding was performed in the vertical-up position, downward progression of the filler wire was observed in the single pass weld with filler. It is evident that in repairs made with a single pass (with and without filler), the positioning of the laser beam with respect to the work piece will be critical if a complete repair of the crack is required. This is due to the size of the molten pool available and the degree of uncertainty in the position of any crack. In an attempt to overcome this problem, repairs were also made using two passes.

In two pass welds produced on oxidised samples and using filler wire addition, larger melt areas were observed. A section through one of these repairs can be seen in Figure 6. Although this shows a full healing of the crack, some minor inclusions can be seen at the root of the first pass. These could originate from the oxide that is causing the cracking. Some small degree of undercut in the resultant top bead could also be seen. As this was considered unacceptable, further work utilising a weave pattern to the laser beam was undertaken.

The aspect ratio (depth/width) of the repairs made with the laser focussed at the surface was of the order of 2. In practice it was believed that such a high aspect ratio would not be required for repair and so a defocussed laser beam was used for further processing in order to produce a wider weld.

The weave pattern chosen is shown in Figure 7. Figure 8 shows a section of a repair made on an oxidised sample using this technique and the addition of filler wire material. From the cross section obtained, it can be concluded that re-melted welds with no observable cracks were observed. However some minor, distributed inclusions and minor undercut were still features of these weave pattern welds. Because of specimen size and joint configuration, it is difficult to assess the strength of the repairs produced.

However the tensile strength of the joint can be correlated to the hardness values in the weld. As a result, hardness surveys were carried out on a sample repaired using the weave pattern technique.

Sections were made at two different positions along the length of the weld ( Figure 9) and traverses were made at the root and cap of the welds as indicated in Figure 10. The results obtained are shown in Figure 11.

An increase in hardness (~90%) was observed in the weld metal of the repaired hanger when compared to the base metal. The maximum hardness values obtained (210HV) suggest that there is no brittle phase (martensite) formation in the weld. The higher hardness values of the weld can be attributed to the composition of the weld metal and the weld thermal cycle. No significant variation was observed between the hardness values obtained at the root and cap of the weld. In addition, no remarkable differences were seen in the hardness values taken at the two chosen locations along the repair path.

The hardness values obtained were very similar to those obtained in samples repaired by the existing MIG technique. It can thus be concluded that the properties of the repaired weld (as a whole) were better than those of the base metal.

The presence or absence of inclusions in the repaired weld metal is not thought to have much significance on the hardness of the samples in the study. It is possible that their presence may effect toughness, but how the inclusion content influences crack growth characteristics can only be concluded by further experiment.

Discussion

Repairing BCD hangers with MIG welding is a multi-stage operation ^[1,3] . Most of the time is consumed in the set-up (movement of the remote welding equipment and oxide removal equipment to position). The current practice involves a procedure for removal of the oxide layer followed by MIG welding. One of the difficulties highlighted in this process is that an earth connection is required to complete the circuit for welding. Earthing each sample to be repaired is time consuming and is an operation in its own right. The size of the welding equipment is also of importance, as the spacing and location of BCD hangers dictate access. These initial experiments have demonstrated that crack repair by laser welding is feasible. Using this process there is no requirement to make an earth connection and the requirement to remove the oxide prior to welding could possibly be relaxed, although further, more detailed work would be needed to confirm the latter. Oxide removal with the laser also seems feasible and again, further work with different oxide thicknesses, on different surfaces, would fully characterise the process.

One of the major advantages of the laser approach comes in the requirement to only deploy one package in the reactor. However, the requirement for precise positioning of the laser beam and wire feed unit makes the laser system more complex than the MIG welding system. In any practical application of the laser process for this type of repair, a more compact laser head would be developed, to gain better access to the components requiring repair. As the beam quality of the type of laser required is constantly improving, this is not thought a problem for implementation of this repair technique in the reactor environment. Further trials are again recommended, concerned with minimising the size of the welding head and providing a reliable means of positioning the laser beam in relation to the in-reactor sample.

Conclusions

The following conclusions were drawn from the work:

• Using a laser power of 1kW and with the workpiece surface 20mm below the beam focus, effective oxide removal was possible at a traverse speed of 2m/min.
• The weld repairs made using a weave pattern technique showed successful re-melting of the crack. In addition, there was no appreciable change in the hardness values recorded at various locations along the weld length and from the root to the cap of the weave pass weld.
• Generally, better penetration in the repaired sections could be seen on samples covered in oxide. This was due to increased absorption of the laser energy in the oxide covered areas when compared to the base material of the hanger.
• It would appear feasible to use fibre delivered Nd:YAG laser light to repair, in situ, the BCD hangers, provided the process head could be engineered sufficiently small to secure access into the reactor vessel and that the required positioning accuracy of the head could be maintained. The use of a laser beam for this repair would have several advantages over the currently used MIG welding technique.
• Compared to a MIG repair process, reactor downtime could be shorter using the laser process, mainly due to the fact that the laser process has the potential to avoid two additional steps required in the MIG process.

Acknowledgements

Thanks are extended to BNFL Magnox Generation and to other members of the TWI project 'Exploitation of High Power Nd:YAG Laser Processing' for permission to publish this work.

References

Meet the authors

N°	Author	Title
1	Morgan-Warren, E. J.	Development and application of the MIG process for remote welding in nuclear reactors. Paper 35. Proceedings, International Conference, Advances in Joining and Cutting processes, Harrogate UK, 30 Oct - 2 Nov, 1989. Abington Publishing. pp. 543 - 552.
2	Gaudin, J. P.	Repair welding at nuclear power plants. Welding Review International, Vol. 13, No. 2, May 1994, pp. 253-254, 256, 258.
3	Morgan-Warren, E. J.	Remote repair welding in nuclear reactors. Welding and Metal Fabrication, Vol. 57, No. 3, April 1989, pp. 109, 111-112, 116.
4	Wiemer, K., Riches, S.T., Fisher, S.	Remote processing applications using Nd:YAG lasers in the nuclear power industry. Proceedings, Euromat 96, Materials and Nuclear Power, Bournemouth, UK, 21 - 23 Oct 1996. The Institute of Materials. pp 359 - 366.

N C Sekhar joined TWI (The Welding Institute) in Cambridge, UK, as Project Leader (Technical) in Nov 1998. He is pursuing a Ph.D at the University of Cambridge, Cambridge, UK. At TWI he has been involved in the development of procedures for welding and cutting of ferrous and non-ferrous alloys using Nd:YAG lasers. He has been active in the application of the developed procedures to industrial context.

Dr P A Hilton is Technology Manager - Lasers at TWI (The Welding Institute) in Cambridge UK, where he has responsibility for the strategic development of laser materials processing. As such he has been instrumental in the setting up and management of several European collaborative research projects. Dr Hilton has previously worked in the laser systems industry in the UK, and before that, was a researcher at an International Scientific Institute in France.

Dr M Tilsley obtained his PhD from the Department of Metallurgy and Materials at the University of Birmingham in UK. He worked on novel techniques for characterising Magneto resistive thin films. Since 1996 he has worked for BNFL Remote Operations Branch at Berkeley, Gloucester. The branch specialises in remote inspection and repair of components and structures in difficult to reach nuclear environments.