Dealing with turbine blade failures in India30 September 1999
Turbine blade failures have been a continuing challenge at India’s nuclear power units. Now NPC has tackled the problem with stricter operating regimes and new inspection techniques.
The operating performance of India’s nuclear plants has improved considerably in the last few years: the average capacity factor, for example, has grown from 60% in 1995-96 to 75% in 1998-99.
In achieving this improvement the reliability of the turbine generator system has been very important. Consistent efforts have been made to improve the performance of the turbine generator (TG), and in particular to arrest blade failures, and these efforts have resulted in good operating performances in the recent past.
Blade failures had been a major cause of non-availability of the stations, causing outages ranging from one month to 22 months. Stabilising the affected units has often taken a long time: if the modern analysis tools available now could have been used, it would have taken less time to bring them back on line.
Turbines at Raj-asthan and Madras encountered several problems during the commissioning period. Many HP blade failures led to long shutdowns at all the units.
Year by year, the history of blade failures is described below.
RAJASTHAN 1 & 2
At Rajasthan 1 & 2 the TGs were supplied by John Inglish of Canada during the mid 1960s, which selected a turbine from English Electric of the UK and a generator from GEC in Canada. The LP rotors were built in the UK. The high pressure (HP) turbine was a new design, while the low pressure (LP) turbine was based on units in other thermal and nuclear power stations.
The Rajasthan machine was designed for 40 bar steam pressure, 250°C steam temperature and 0.26% wetness. The machine supplied is a double cylinder HP turbine with five stages of impulse reaction blades, and two double-flow LP rotors each with six stages of blading. The blade height of the LP turbine last stage is 914.4 mm (36 in).
1973: First synchronisation.
1974: The first blade failure was observed, on 4 July . Three blades were found broken in stage 4, one at the root and the other two at the lacing wire hole. The blade roots were an inverted ‘T’ design with side grip. Investigation showed that the failure was due to excessive clearances at the side grip. It caused the natural frequency to vary from that intended in the design, and resulted in resonance with the diaphragm impulse frequency. A new blade design was adopted, with a modified profile and without lacing wire. The root width was increased and the number of blades reduced from 137 to 125.
1975: In August a second HP blade failure was observed in unit 1. This time it was on HP stage 3, where one blade was broken at the ‘T’ root and many others had cracks, also at the ‘T’ root. Three cracks were also observed on the disc at the side grip. Considering the disc crack, the rotor was salvaged by changing the root to single fork type pin root, using modified blades.
1976: In February one blade of LP stage 5 failed near the lacing rod hole. Static and dynamic tests on the unit 2 rotor, then at GEC in the UK, were conducted to re-establish the natural frequency of these blades. In view of the adequate margin available the failed blade was replaced and the unit was re-commissioned.
1989: In December one blade in stage 4 of the unit 1 HP turbine was found to be broken at the root. The rotor was replaced with one of a new design procured from GEC which had improved blade root design in stages 2, 3 & 4. Since then, the HP turbines have operated satisfactorily.
1997: In July one blade failed from the root in LP stage 5, in flow path 2. Later, in November, there was a second failure in the same flow path. In April 1998 one blade in flow path 1 also failed. All these failures were high cycle fatigue failures. To deal with the problem the blade height in the affected stages was reduced, increasing the margin on high frequency.
1981: First commissioning.
1982: In January a blade failure was observed in HP stage two. Two blades were found fractured and two blades had cracks at the ‘T’ root head. Disc cracks were also observed in the side grip. The unit was restarted after the rotor was replaced with one from unit 1. The failure of the blade was attributed to corrosion-assisted high cycle fatigue with excitation predominantly in the tangential direction. The failed rotor was re-bladed with new blades shot peened at root and spot welded at packer. Lacing wire was introduced and blade packets were connected by under-strapping, converting the stage into a continuous blade assembly. This changed the natural frequency of the blades and provided additional margin in the operating frequency.
1983: In January the new HP rotor from unit 1 failed again – as in the earlier failure, at the HP stage 2. This stage was also re-bladed with the modified design.
MADRAS 1 & 2
The TG sets for Madras, and for subsequent plants, were supplied by Bharat Heavy Electricals of India, with HP and LP turbine designs from English Electric AEI of the UK and generator design from Russia. The first LP rotors were built in the UK.
The turbine comprises one HP single cylinder with five stages of impulse reaction blades, and one double flow LP rotor with five stages. All five stages of HP blades are of inverted firtree design. The last stage blade selected is 945 mm (37 in) long with curved inverted firtree side entry. The LP rotor basic requirements of strength and stiffness were met by adopting large chordal widths, coupled to adequate camber profiles, over the inner part of the blade. Torsional restraining zig-zag rods were employed to connect all the blades.
1983: Madras 1 was synchronised to the grid in July.
1987: After 23 000 hours of operation, in October, HP rotor blades failed in stages 2 and 4. Several cracks were also observed on the stage 4 rotor disc. The unit was re-commissioned by cannibalising the HP rotor from Kakrapar 1. The blade failure at the root was classified as high cycle fatigue failure, and preliminary investigations revealed inadequate water drainage from the HP cylinder. The existing orifices from the cylinder drains were removed and the unit was restarted.
1988: In December HP stage 4 blade failure was again observed on the new rotor. Several cracks on the HP stage 4 disc were reported. It was decided to operate the unit without the stage 4 blades at reduced power level, since no spare rotor was available. All the disc cracks were removed; the stage 4 diaphragm was replaced with a pressure plate and the blade roots were used to protect the disc profile. The unit was restarted, after modification of the drains similar to that undertaken for unit 2 (see below).
1993: Following the Narora fire (see below) Madras 1 was shut down to examine the last stage blades. Fine crack initiations at the roots were observed in few blades during wet fluorescent magnetic particle test (WFMT). The blade could not be successfully removed at site, as the blade fit-up was tight. The rotor was replaced with that of Rajasthan 3: the old rotor was modified and used at Madras 2.
1985: Madras 2 was synchronised to the grid on 12 September.
1987: In December, after 12 900 hours of operation, blade failures were observed on HP stages 3 & 4. Several cracks were seen on the HP stage 4 disc. The failures observed were due to high cycle fatigue, similar to those at unit 1. Blades were also fractured at the LP turbine stage 1 due to hot reheat bellow sleeve failure. In response, the HP cylinder drainage after stages 1 & 2 was modified, steam traps from the drain lines of extraction were replaced by orifices and the drain by-pass valve was kept open to 110 MWe load for effective drainage. The unit was restarted after the rotor had been replaced by one from Kakrapur 2.
1992: In July, following 39 500 hours of operation (around 23 300 with the new rotor), cracks were observed on one blade in HP stage 3 and eight blades in HP stage 4. All the blade roots of stages 3 & 4 were shot peened to improve their fatigue properties. Additional modifications on stages 2, 3 & 4 diaphragms were carried out to improve the water separation and removal, while within the cylinder the drain was enlarged with a drain plate to improve water drainage. The unit was restarted with these modifications, and no further failure in has been reported in the HP turbine.
1993: Madras 2 was inspected following the Narora 1 failure. Examination did not reveal any crack initiation and the unit was allowed to operate for a further 5000 hours, before being modified. During the following annual shutdown the rotor was removed and sent for modifications. The unit was restarted using the modified rotor from Madras 1.
NARORA 1 & 2
Turbine generator sets on these two units are similar to those of Madras 1 and 2.
1989: The Narora 1 turbine was synchronised to the grid on 29 July.
1991: In July, inspection revealed that one zig-zag lacing rod on the last LP stage blade assembly was missing. The lacing rod was replaced and the unit restarted.
1993: In the early morning on 3 March there was an oil and hydrogen fire in the unit 1 turbine generator set. The turbine had operated for around 16 000 hours. Subsequent investigation established that the failure of two LP stage 5 blades was the initiating event for the fire and consequent extensive damage to the turbine generator assembly. There are 78 stage 5 blades 945 mm long. The two blades failed at the root, which in turn led to failure of 14 adjacent last stage blades. These failed blades caused very severe unbalance, leading to failure and uprooting of the bearings of the HP, LP and generator shafts. The rear end bearing of the LP turbine broke in pieces. The generator front bearing had moved out from its position and fell near the pedestal. The generator rear bearing, mounted in a separate pedestal, was dislocated and thrown some 3-4 m from the pedestal. The bearing pedestal itself shifted around 250 mm from its location, after shearing off the entire hold-down bolts. The entire unit was damaged extensively and all the three rotor shafts were found to be beyond repair. The complete HP, LP turbine and generator units were replaced, using equipment from Rajasthan 3, and the damaged machine was returned to the manufacturer for rehabilitation or replacement. The unit was put back in service on 5 January 1995, 22 months after the failure.
Detailed investigation and root cause analysis indicated the failure was due to high cycle fatigue. A number of lacing rods were found missing on the undamaged stage 5 blade. Blades without a lacing rod tend to resonate in the first mode at operating speeds. Dynamic frequency tests were conducted on blades without a lacing rod at GEC’s works in the UK and it was established that there was substantial shift in natural frequencies of the blade, from 112 Hz with all the zig-zag lacing rods in position, to 99–102 Hz without a lacing rod (free standing blade). Prior to this incident GEC had observed very fine crack initiation under WFMT at the blade root in other utilities. The following modifications were carried out:
• Material in the form of scallops was removed from the blade platform above the top serration in the regions near the inlet and exhaust of the blade, to increase fatigue endurance.
• The corners at the inlet and exhaust side of the blade root were re radiised.
• The blade roots were shot-peened.
• The sharp edges at the intersection of the concave edge of the root, which had plain side faces, were chamfered.
• The radii of the fillets of the lacing rod were increased to increase fatigue strength.
• Ripple springs were provided at the base of the root to avoid fretting.
• High tensile alloy materials were used to replace bearings holding down studs.
These modifications were carried out at the first available opportunity at Madras 1 & 2, Narora 1 & 2, and Kakrapur 1 & 2.
1991: First power was achieved in October.
1992: In December, after 5000 hours of operation, examination of unit 2 found 30 lacing rods to be missing or failed, 13 of them in front flow and 17 in rear flow. The rotor was replaced by the one from Kaiga 1 and the original was sent to the supplier to be modified. All the blades were examined by WFMT. Crack initiation was observed in a few blades; these were replaced with new blades, and after modifications the rotor was used at Narora 1.
1994: In September Narora 2 suffered HP stage 3 shroud and tennon failure. The unit was re-commissioned using a spare HP rotor with pin-type blade root design. The rotors had used straddle root fixings.
1995: In January the blade failed from the root in HP stage 3 of the new rotor . This stage comprised 94 blades fixed with pins. The rotor had been in service for only 1500 hours. A thorough examination of the machine was conducted. The stage 3 diaphragm was sent to the works for detailed examination. A large number of deviations were observed in the diaphragm, such as bulging, and throat redimensions. These deviations had caused resonance due to asymmetric excitation leading to high cycle fatigue.
After critical examination a new diaphragm from Rajasthan 3 was installed in Narora 2, and the axial gap between the diaphragm blade and the rotor blade was increased to reduce excitation forces on the blades.
The machine was started with these modifications and after installation of the repaired rotor (the original rotor had inverted firtree roots). The unit is currently in use and operating satisfactorily.
Kakrapur 1 & 2
Kakrapur 1 & 2 were shut down temporarily in the mid 1990s following the Narora 1 failure and underwent modifications similar to those at Narora and Madras. Other turbine failures are as follows.
1992: In November Kakrapur 1 was synchronised to the grid.
1997: In April, after 23 000 hours of operation, full width cracks were observed at five places in the shroud of the LP stage 3 blades. Stage 3 is composed of 11 packets of 11 or 12 blades. Four adjacent packets were affected by the cracking. Similar cracking was found at Madras.
1995: In March Kakrapur 2 was synchronised to the grid.
1997: During an inspection in October five full-width under-strap cracks were observed in the LP stage 3 rotor. On the last stage blade a 25 mm crack was observed, initiated at the stellited strip around 75 mm from the tip of the blade. A shroud crack around 13 mm long was also observed in the HP stage 3 rotor. The unit was re-commissioned after replacing the cracked blade in stage 5.
1998: An inspection in December revealed a crack about 25 mm long on two blades of the LP stage 4 at the lacing wire hole, and another crack around 40 mm long from the stellite strip on an LP stage 5 blade. The unit was restarted after repair and replacement.
OPERATION & DEVELOPMENT
Strict quality controls on water chemistry, in particular control of impurities, were imposed after the initial HP turbine failures at Rajasthan and Madras. Strict controls were observed on boiler levels and on operation of the steam and turbine drain systems.
The last stage failure at Narora and subsequent modification and replacement at other units has caused enormous loss of generation and cost. One result of the failure was a requirement for stricter control of off-frequency operation. In view of the narrow band of allowed operating frequencies, revised guidelines were issued for operation beyond this band: this required WFMT inspection of the rotor blades. The guidelines include continuous operation within 48–51 Hz. The unit is tripped after 20 s at a grid frequency beyond 47.5 Hz and 51.5 Hz. The HP stage 3, LP stage 3 and LP stage 5 blades are prone to resonance and close attention must be paid to fabrication and operating speeds. Recommendations for off-frequency operation are shown above.
With limited expertise available in the area of rotor dynamics and blade design, indigenous efforts were focused on establishing blade frequencies for the assembled HP & LP rotors by conducting telemetry tests in the balancing and overspeed tunnel. FEM methods were used to carry out theoretical analysis of blade frequency and stress levels. Base data was generated on each assembled HP & LP rotor by conducting static vibration tests.
A high level task force has been constituted, with representation from the manufacturer, utility and outside experts, to address problems related to turbine performance.
Although the Indian power programme has seen many failures of turbine blades, constant improvement in design and manufacturing has improved availability. However, the wide frequency variation observed in the grid is still a cause of concern. With modern design tools and improved materials the problems may be overcome in future machines.
Related ArticlesKakrapur Unit 1 back in operation