A retrospective review of some troublesome turbine blade failures

H.B. Carrick    Process Industry Machinery Expertise Ltd (PrIME), UK

2 INTRODUCTION

Turbine blades are not normally a problem. However when troubles do occur they can be difficult to cure. This is partly because the rotating blades are exposed to quite high steady stresses. A typical blade root with a mean stress of 250 MPa will be likely to yield locally at overspeed conditions due to stress concentrations. Yet fatigue is the failure mechanism of almost all turbine blade failures. This is because the flow field into a rotating turbine blade is by definition unsteady. Wakes from the preceding stationary blade row (the nozzles) impose a strong excitation equal to approximately 100% of the steady bending load at nozzle passing frequency. Partial arc admission on control stages imposes excitations of similar magnitude but lower frequency. Since this partial arc loading has a typical ‘square wave’ form the resulting excitation of the rotor blade covers a broad range of frequencies. More subtle excitations come from the non-uniformity of nozzle spacing, obstacles upstream and downstream of the blade row and non-uniformities due to connections into or out of the machine. Flow instability can also provide an excitation for compressors and for longer turbine blades (typically in the last stages of condensing machines).

Turbine blades of any size have many natural frequencies in a frequency range which can be brought to resonance by these excitations. Thus variable speed machines have to be designed to endure resonance. This requires conservative blade design (low bending stresses and careful blade detailing), control over the excitation from the flow path, and control over the response of the blading, including damping. Fixed speed machines are sometimes designed to avoid specific dangerous resonances, and may have specially tuned blades for this purpose.

3 PLANT NO. 1 - A VARIABLE SPEED MACHINE

This plant had a 10 MW steam turbine driving a recycle gas compressor. The machine was commissioned in late 1979.

In April 1982 after about 16,000 hours operation on rotor no. 2 the machine suddenly stopped. When a restart was attempted very heavy unbalance was found on the turbine at low speed so the turbine was opened. One blade was found to be missing from the last stage of this 4 stage turbine (see figure 1). It was concluded that the vibration following the blade failure had been so violent that it caused the turbine mechanical overspeed trip to operate.

The blade fracture surface was dominantly fatigue. On disassembly another 3 blades in stage 4 were found to be cracked. The relative position of the cracked blades and the broken blade are shown in figure 2. The 70 blades were grouped into 10 packets (or groups) of 7 blades by a rivetted cover-band. It can be seen that the cracked blades were all at the end of a packet. Also the cover-band had interlocking 'tabs' at the ends of each packet, and these tabs had fretted, indicating relative motion between the blade packets. It seemed clear that the blade failure was due to blade packet vibration. Extensive investigations of blade, packet and disc natural frequencies were carried out by the owner.

The blades had cracked at the upper platform of the double hammerhead root (see fig. 3).

The crack initiation was at the suction side corner. On further investigation it was found that the blade airfoil was stacked so that centrifugal loads imposed significantly higher stress on this side of the root. In addition, the blade root platforms were flat, while the disc groove is curved, hence all the centrifugal load was concentrated on the corners of the blades.

The failed blade was examined under the scanning electron microscope (SEM) by the vendor who concluded that the fracture surface indicated fatigue 'with the influence of corrosion'. Recommendations from the vendor included changing to a material of somewhat improved strength and corrosion resistance, re-stacking the blade profile to distribute the centrifugal load more equally, and increasing the radius between root and blade shank to reduce the stress concentration in the blade root.

This solution was not accepted technically or commercially by the owner, mainly because it was felt that the vendor was not investigating the failure seriously and because there was no guarantee available against a further failure. The turbine was re-bladed by a third party blade manufacturing specialist, supported by their consultant engineer. The blades were re-engineered with the following changes: improved material similar to that proposed by the turbine vendor, rolling radius for the root to eliminate corner loading, longer cover bands to damp out per rev excitation of the lower packet modes, better radial stacking of the blade profile and shot peened roots. The re-bladed 4th stage had four packets, two of 18 blades and two of 17 blades.

In 1983 after about 15,000 hours operation, turbine rotor no. 1 (the original spare) was removed from service and the 4th stage blades removed for examination. A number of blades were found to be cracked, but this time in different locations in the packets (see figure 4).

One of these cracks was over 75% through the critical root section of the blade (see figure 5). Once again the cracks were caused by fatigue. Obviously we had been very lucky to avoid another failure in service.

In Oct 1984 after about 12,000 hours operation the first modified (Mk2) rotor experienced a blade failure, of the single (unique) closing blade. No other cracks were found. This rotor had operated at significantly higher speed and load than either of the Mk1 rotors. The blade manufacturer's consultant suggested that the failure was due to vibration in the 1st tangential out of phase mode. However it was subsequently discovered that there were also some manufacturing anomalies with this blade (the serrated teeth had been re-machined during manufacture).

At this point it was decided that the original vendor should once again be involved to see if a more robust design could be achieved. After extensive discussion in which all the resources of the vendor were involved, the following package of changes was proposed:

- retain better radial stacking from Mk 2 design

- retain rounded root land from Mk 2 design

- retain shot peening from Mk 2 design

- diffuser plate between 1st and 2nd stage to reduce any partial arc excitation

- cut back the 18 piers on the stage 4 diaphragm to try to reduce the excitation from this source

- alter the stage 4 nozzle exit angle slightly to reduce the blade loading on the 4th stage rotor blade (at the expense of the 3rd stage)

- remove the two blank nozzles from the diaphragms on stages 3 and 4, and replace them with special nozzles to reduce the flow disturbance

- reduce the 4th stage nozzle length slightly, and hence the 4th stage blade length (to match the new nozzles). This produced a slight reduction in tensile load on the blade which compensated for the next change

- install a rotor blade with an integral cover plus a cover band, to increase damping in the blade assembly. The cover band was to be longer than the original, but shorter than Mk2 design (14 blades per packet)

- increase the fillet radius at the root to shank transition to reduce the stress concentration. This involved machining the rotor disc.

- install a loose lacing wire to provide friction damping of the higher (out of phase) packet modes. A key requirement from the owner was a 3 year guarantee for the redesigned blading. The lacing wire was the vendor’s technical insurance for this guarantee.

Meanwhile a limit was put on the speed to which the turbine could be run, to try to avoid predicted resonances and hence a repetition of the Mk2 rotor failure. This was unsuccessful, and on 1st August 1985, after only 7000 hours operation, and before the Mk 3 rotor was ready, the second Mk 2 rotor failed. Once again only one blade had cracked, though in this case some additional cracking was seen in the blade lugs and disc seal rim. Examination of the blade showed that the blade root platform of the failed blade had been machined at an slight angle to the blade, increasing the stress at the crack initiation site.

The turbine was operated with only three stages for two months (at some cost in efficiency). The Mk 3 design was installed in Oct 1985.

An independent study was commissioned into the failures and the redesign to see if a clean bill of health could be given to the Mk 3 rotor. The study managed...