Leakage investigations at Cleuson-Dixence15 February 2007
H Cerjak and N Enzinger report on the use of high strength steels in hydro power plants, and share some lessons learned from the Cleuson-Dixence power plant, where leakages were observed during the project’s first period of service
The use of high strength steels in the pressurised shaft of hydro power plants allows a reduction of the wall thickness and can therefore lead to lower costs during production and/or bigger shaft diameters, which increases the efficiency of a plant. This concept was applied at the 1200MW Cleuson-Dixence hydro power plant in the canton Wallis in Switzerland. This important project began operation in July 1998, but during this first period of service several leakages began to appear in weldments of Steel S890.
The project’s pressure shaft was taken out of service and a full restesting by Non Destructive Testing (NDT) was carried out, which led to the discovery of 99 individual non acceptable Ultrasound Testing (UT) indications in 66 welds. These were subsequently repaired and the shaft was placed back into service. However, after approximately four months of service, the shaft collapsed on 12 December 2000.
Thus many questions arose concerning the properties and behaviour of Steel Grade S890, which was used in the zone of the catastrophic accident. Besides investigations performed by court experts, the project’s owner started a review on the design, fabrication and service experience history of the shaft. To assure a safe-reconstruction concept the owner began the ‘Qualification Program S890’ to define quantitative materials parameter as a necessary background for further steps. Because no original material of the shaft was available, representative material S890 was ordered and welded using original welding procedures representing the properties of the shaft. This material was the basis for the characterisation of mechanical properties and weldability reported in this paper; additionally performed fracture mechanic tests and fracture mechanics analysis based calculation of critical flaw sizes are reported in reference 1 and stress corrosion susceptibility investigations are reported in reference 2. As a short summary of these investigations it was found that steel grade S890 and its weldments show satisfactory behaviour, provided proper treatment – including preheating – is performed. In addition, diligent non-destructive testing should be applied a sufficient time after welding to enable the detection of any hydrogen induced, delayed, cold cracking.
For the penstock, material plates from high strength structural steels were used. Depending on the internal pressure three different grades of steels were selected. In the lower part of the plant the steel grade S 890 QL is used (see Figure 1). The steel plates were ordered by the consortium GSN from Öxeloesund, Thyssen Stahl, Dillinger Hüttenwerke and Sumitomo Metal Industries Ltd.
The fabrication of the shaft was divided into four work packages, which were awarded completely to the GSN consortium which included Giovanola at Monthey (leader), Sulzer Hydro at Kriens and GEC Alsthom, Neyrpic at Grenoble. The manufacturing of pipes for the parts of the shaft with thick walls (41- 72mm) was shared between Sulzer Hydro (Ravensburg) and ATB Caldereria (Brescia), acting as subcontractors. The other three sections of the shaft were manufactured entirely by Giovanola.
The assembly welding was shared between Giovanola, GEC Alsthom Neyrpic and Sulzer Hydro. Giovanola and Sulzer Hydro chose the coated electrode method for in-shaft assembly welding. GEC-Alsthom Neyrpic opted for the automatic or semi-automatic machine method using the MAG process.
For production of the shaft segments, plates were cold rolled to rings and longitudinal welded in the shop using the submerge arc welding process (SAW). Two single rings were connected with a circumferential SAW to a section transported to the site location. In front of the tunnels these single or double rings were connected to assemblies using the circumferential SAW process. The prefabricated parts were then brought into the shaft and welded in position to the final penstock. Manufacturing and prefabrication of the pipes in the workshop started in 1996. The assembly on site was carried out from 1997 to June 1998. Finishing the assembly, in July 1998, a pressure test was carried out on the whole shaft. On the basis of the concluding test results, the reception of the shaft was pronounced in July 1998.
Welding and quality assurance (QA)
For each of the single fabrication steps procedure specifications and test specifications were developed in the framework of a general QA system, with the results of the findings recorded in protocols. A certain preheating temperature and a post weld heating procedure were derived from special tests performed by Institut de Soudure. These measures are used to control an expected sensibility of high strength steel to cold cracking.
Also, in the NDT specification GSN C-3655 this fact is considered by the defined time span between the welding and the non-destructive testing. In this specification the possibility of ‘delayed cold cracking’ is considered by defining a repeat of the NDT-tests seven days after welding is finished. In addition, repair welds have to be retested seven days after welding is performed.
All welds should be tested by non-destructive methods: 100% by UT-testing and 10% by magnetic particle (MT) testing. The intersections of longitudinal and circumferential welds should be tested by applying X-ray testing, which can be replaced by UT-testing plus limited MT when the wall thickness exceeds 34mm. As a result of these NDT procedures applied, a number of repairs in the welds have been reported in the NDT protocols. Amongst them were repairs performed in the region where failure occurred.
Observed leakages and related repair works
During the plants first period of operation three leakages were detected. In July 1999, a very small but constant pressure drop was detected in the penstock by the power station’s supervisor during control work in the emptied shaft. After the systematic control of different installations it was concluded that there a leakage might exist. Using precision measurements the position of the leakage was found downstream of Peroua in the section F III (Figure 1). A systematic visual inspection of the joints led to the detection of a 60mm long crack in the longitudinal workshop weld of pipe 172-2, at the intersection of a circular workshop weld.
GSN applied a standard repair method by digging out the cracked part and welding the affected area. Because of the presence of water this repair method proved to be extremely difficult. Subsequent systematic leakage controls after the penstock was once again set under pressure showed that the leakage remained. Delayed cold cracking and the presence of hydrogen in the weld was suspected as a reason for this crack. A second repair attempt was made and the plant was put back into operation in July 1999.
In the middle of October 1999 a systematic pressure control showed another leakage. The two failed attempts to apply the standard procedure of repair welding described above was abandoned and replaced by a ‘heavier solution’, which involved manufacturing a disc, cutting the failure zone, executing an excavated niche in the concrete behind the steel penstock, draining of water occurrences, construction of a water tight metal enclosure and welding of the disc. The repair was completed on 2 February 2000.
The leakage control of the shaft after the final repair of weld 172-2 took place on 3 February 2000. Contrary to all expectations it revealed that there was still a global leakage present under maximum static pressure.
On 5 February 2000 the repair of pipe 172-2 was examined. The welding joint of the disc controlled by 100% ultrasonic testing showed no welding defects. This led to the conclusion that the pipe was perfectly tight, meaning there must be one or more other points of leakage existing in the penstock.
On 9 February 2000 a very exact control of the entire pipe section of interest led to the detection of a leak in pipe 316-1, above Dzerjonna. A 100mm long crack on the external surface of the circular workshop fabrication joint in the area of an intersection with a longitudinal weld was detected.
On 14 February small cracks, typical of cold fissuring, were discovered in the circular workshop weld of pipe 276-1, downstream of Dzerjonna. These cracks, which were perpendicular to the welding seam, showed a crack with length of about 30mm and a very small opening. The parties involved became aware that this problem could be related to the appearance of cold cracking.
Retesting of the Shaft by NDT
It has to be mentioned that the detection of cracks in the areas of leakages up to now were made only by visual inspection. Considering this fact it was necessary to proceed with a systematic control of all welded joints in the zone where the high strength steel S 890 QL was used. The application of UT was decided on 14 February 2000.
For the control operation it was decided that the same NDT specification used by GSN for the fabrication and site weldings should be used. The retesting was somewhat more restricted because of the presence of the corrosion protection inside the penstock and, of course, by the limited access.
From Table 1 it can be seen that 66 joints had to be repaired in accordance with the applied NDT-specification. That meant that on average 1.5 repairs per repaired joint had to be performed. All failures to be repaired were found exclusively in workshop welds, performed by one fabricator.
Because of the different types of failures found and local conditions, four repair procedures were selected. These methods were:
•Type 1: the failure does not penetrate to full wall-thickness - the failure zone was ground off and welded using manual welding.
•Type 2 and type 3: Wall penetrating failure. The problem of handling the water access to the welded zone became virulent. In these cases the failure zone was cut out and replaced by a circular coupon. Depending on the amount of water which had to be drained off two methods – the ‘boss’ and the ‘disc’ method – were applied, each by welding a circular disc in the wall of the penstock. In most cases these repairs were applied when the indications were found on the intersection of longitudinal and circular welds.
•Type 4: In the case of localised cold cracks perpendicular to the welding seam the repair was performed by piercing the steel and sealing it with a screwed stainless steel plug.
UT retesting and repair works were performed in a continuous working sequence over five and a half months, with the work completed at the end of July 2000.
Examination of the possible causes of leakages
After detection of the third leakage in the penstock on 2 February 2000, the first systematic investigations of the causes began. The Institut de Soudure in Paris was contacted to investigate six different samples cut from the penstock. The results of this investigation are reported in reference 4.
In one sample cold cracking perpendicular to the welding direction was found, in all other investigations longitudinal cracks, partly or fully penetrating the wall thickness, were detected. In the report4 cold cracking in the weld metal is depicted as the main reason for the occurrence of cracks and leakages in all cases. In the case of leakages, the report suggests that existing cold cracks which more or less penetrate the greater part of the wall thickness propagated step by step until they broke through to the surface. In all cases the cracks are located in the weld deposit of the multilayer weld. The appearance of the crack surface led to the conclusion that brittle separation caused the crack during or after the welding process. The crack surface was found to be oxidized, and propagated to a certain extent stepwise through the wall leading to the leakage.
The lengths of the cracks were reported to be greater than 100mm for a penetrating crack (sample 316.1) with a penetration length of 70mm and in the case of a sample 126 and 129 up to a length of 240mm, open to the surface. The crack in sample 316.1 was found in connection with a repair applied during fabrication welding.
Some examples of the results of investigations performed by Institute Soudure are given in Figure 2. Investigations of smaller types of cracks are described by Dorsch et al in reference 2.
After finalising the repair work in July 2000 the shaft was taken into service again. Tightness tests had been carried out on 15 August, 15 September and 15 October 2000. On 12 December 2000, at 20:09:17 after the planed stop of the third turbine the catastrophic accident occurred .
The shaft suffered a longitudinal crack of about 9m length and opened abound 60cm wide. According to the report of the court experts6 the crack started in the longitudinal weld of the ring 155.1, crossed the circumferential welds nr. 118 upstream, and nr. 119 downstream, penetrated the base materials of the adjacent rings, each 3m wide, and stopped by branching in the circumferential weld seams nr. 117 and nr. 120, respectively.
Qualification Program S890
To support further steps regarding reconstruction alternatives of the Cleuson-Dixence shaft after the catastrophic collapse in December 2000, the ‘Qualification Program S890’ began in July 2002.
The aim of this qualification program was the generation of material data of the steel S890, as base material, heat affected zone and weld material, especially with regard to its toughness and failure tolerance capacity.
Analysis of status
For the detailed analysis of the material S890 and its weldings, several sources were searched with the information collected in a self developed database. Out of these efforts the following results were collected: chemical analysis, plate geometry, tensile and charpy test data, suppliers material certificates, position in the shaft, fabrication and testing of the welding.
The analysis of the chemical composition shows that the involved steel supplier uses different approaches in the production of the steel S890. Significant differences can be observed concerning Ni, Cr and Cu (see table 2).
Concerning the Charpy values at -40°C a minimum of 27J is achieved. Nevertheless there are differences between the different suppliers in the width of the scatter band. Thyssen as well as SSAB show a good performance with respect to isotropy, whereas Sumitomo documents show much higher minimum values.
A similar analysis was performed concerning the values obtained from tensile tests. Again significant different scatter bands are observed. The fracture elongation shows a relative low scatter for all suppliers. No major influence of the plate thickness on the results could be observed.
For a quantification of the material used in the shaft, no original material from the shaft was available. To find representative materials for the ones used in the shaft the results of assessed material data from the self developed database were used.
Based on these results, representative materials (lower toughness, alloying concept) have been selected in cooperation between IWS and MPA-Stuttgart. At that time S890 material was only available from the manufacturers Thyssen and Dillinger.
An extensive specimen plan was designed, to enable the investigation of following welding-material-thickness combinations
•Three different welding procedures (SAW, SMAW, MAGW).
•Two different material suppliers (Dillingen, Thyssen)
•Three different plate-thickness (25mm, 35mm, 55mm)
This extensive investigation programme had the aim to quantify the toughness for the determination of the critical failure size for collapse of the material S890 used in the shaft. The investigations performed included: Tensile tests, Charpy tests, Wide Plate tests, fracture mechanics tests on small specimens (J-Integral), crack growth measurements, Gleeble simulation for the determination of the weldability, metallographic investigations, hardness tests and residual stress measurements.
Generated welded materials
In the qualification programme S890 the welding conditions applied on the shaft should be reproduced as close as possible. The following welding procedures have been used for the fabrication and assembly of this shaft:
•Submerged arc welding procedure (SAW) for all longitudinal joints and circumferential joints in the shop and in front of the tunnel
•Shielded metal arc welding (SMA) for assembling circumferential weldings in the tunnel
•Metal active gas welding (MAG) for assembling circumferential weldings in the tunnel
As a consequence of the report of the court expert , the qualification programme has been reduced to the plate-thickness 35mm (Thyssen) with the welding-procedure submerged arc welding (SAW) and shielded metal arc welding (SMAW). For the same reason a stress corrosion programme was established.
The results of the qualification program S890 are partly reported in this paper, with the results of fracture mechanic investigations and calculations reported by Roos et al in reference 1.
Investigations and results
For each microstructure of the 35mm thick joint of the XABO 890 (Thyssen) base material tensile tests were carried out in longitudinal and transversal directions at room temperature.
•The base material was confirmed to fulfil the specification and was found to be isotropic.
•The SAW weld metal matched the base metal but showed a slightly decreased rupture elongation in transverse direction. Rupture occurred in the weld metal.
•The SMAW weld metal showed a significantly lower yield strength compared to the base metal.
The hardness was measured at different distances from the surface in the cross section to find the location of the maximum hardness.
In the SAW the hardness maximum was found just underneath the final layer (approximately 4mm) in the heat affected zone in the range of about 450HV10. Over further distances the maximum hardness reached values of about 360HV10.
The SMAW showed a similar behaviour. Maximum hardness of about 430HV10 was found in the heat-affected zone near the surface. In a more distant region the hardness reached about 360HV10.
The toughness of base metal, heat affected zone and weld metal of the generated weldments was measured performing the ISO-V-Impact test at nine temperatures (-60°C up to 100°C) according to DIN EN 10045. The transition curve was derived by applying the Hofer-Hung method.
•The base material near the plate surface shows anisotropy in the upper shelf (150J in rolling direction and 125J in transverse direction) whereas in the centre of the plate uniformly 125J are measured. The lower shelf shows values of approximately 50J.
•The upper shelf value of the SAW weld metal is about 125J; the lower shelf value approximately 60J.
•The upper shelf value of the SMAW weld metal is about 125J to 150J; the lower shelf value approximately 50J.
The Chabelka method was applied at room temperature to quantify the influence of the HAZ of the distinguished weld methods (SAW vs. SMAW).
Summarising the results of the Charpy tests it can be seen that in every case the specification of 27J at -40°C is fulfilled.
All investigated microstructures, base metal, heat affected zone SAW weld metal as well as the SMAW weld metal shows the typical annealed martensitic structure.
The behaviour of the heat affected zone was also investigated by loading base metal specimens with representative thermal cycles. Two different peak temperatures and their combination were applied. The obtained samples were then tested with the following methods: tensile and charpy test, metallography and hardness.
•Simulated specimens representing the heat affected zone confirmed the high hardness which was observed in the cross section of the weldments.
•Metallographic investigations showed the expected microstructure due to different, representative thermal cycles.
•Charpy curves showed an adequate lower shelf and transition temperature.
These conclusions also include parts of the findings reported in references 1 and 2. The toughness behaviour of S890 (Thyssen and Dillinger) and its SAWelds were tested by means of:
•Standard tests (Tensiles, Charpy V-Curves) (BM, SAW, SMAW)
•Small scale Fracture Mechanics Tests KIc, Jc, da/dN, COD (BM, SAW)
•Wide Plate Tests (BM, SAW)
Steel type S890 is a high strength steel showing, in comparison to lower strength grades construction steels, a limited failure tolerance. This fact is demonstrated by the smaller distance from crack initiation point to collapse – a point which is important in case a cracked component is overloaded.
•The toughness, expressed by crack initiation limit is generally satisfactory.
•The toughness behaviour of SAW-weldings is similar, compared to that of the base material (Thyssen, Dillinger), but show somewhat different behaviour.
SAW and SMAW welds are highly sensitive to the formation of hydrogen induced cold cracking in weld material. Base material weldability is also satisfactory (Dillinger, Thyssen).
Steel grade S890 can principally be used when the following boundary conditions unrestricted can be fulfilled:
•Keeping the design stresses as low as possible
•Assuring crack free welds by strict QA measures and optimised automatic NDT
•Control of the inside maximum agreed design limits under all conditions of design, assembly and service conditions (incl. emergency)
•Strict H-control of welding processes and NDT to prevent cold cracking
O. Univ. Prof. Dr. mont. Horst Cerjak and Dr. techn. Norbert Enzinger, Graz University of Technology, Institute for Materials Science, Welding and Forming, Kopernikusgasse 24, A-8010 Graz, Austria. Email: [email protected], [email protected]
This paper is based on a presentation at the 14th International Seminar on Hydropower Plants. For more information email: [email protected]TablesTable 1 Table 2