Experimental stress analysis on the emergency power diesel engine of a nuclear power station

Nuclear power stations usually rely on diesel generator sets for emergency power because they are highly reliable following many years of development as ship propulsion units. As emergency generator sets, they must be ready for operation within seconds, and then run at a constant speed, in order to drive a directly coupled generator.

The problem

Nuclear power stations usually rely on diesel generator sets for emergency power because they are highly reliable following many years of development as ship propulsion units. As emergency generator sets, they must be ready for operation within seconds, and then run at a constant speed, in order to drive a directly coupled generator.

When one such emergency diesel engine, (20-cylinder, 5 MW output) was undergoing its annual overhaul, a defective push-rod was discovered in both exhaust and intake of the valve actuators. A similar fault had occurred some years previously and had been judged to be a single failure.

The push-rod is a cylindrical piece of pipe 750 mm long, with both end pieces in a spherical head design. It is flexibly mounted in half shells between the valve lever and the rocker arm. To enable the push-rod to move smoothly in the half shells, permanent lubrication is ensured by the valve lever.

The question of why a component that was only subjected to compressive loading should be damaged by bending vibrations, arose the first time the fault occurred. The action taken at the time to prevent the fault re-occurring was to test the tappets for radial run-out, as buckling stress was seen as a possible cause with a rod that was narrow and not perfectly straight. In future, only tappets with a concentric run-out tolerance of less than 0.1 mm were used. The re-occurrence of the fault meant that now an explicit explanation was required.
The fault (tappet fracture) occurred more or less exactly in the center of the tappet. Metallurgical investigations pointed to a high number of vibration load cycles being the cause.

In a four-stroke engine, the camshaft rotates at half speed. At 1500 rpm, this means a dynamic excitation of the tappet at a repetition frequency of 12.5 Hz. The manufacturer stated a natural bending frequency of more than 100 Hz, that is, too far removed, to emanate from resonance excitati

The measurement program

The problem led to the following experimental procedure:

Establishing the natural frequency of the tappet. The natural frequency was defined by means of pulse-shaped excitation at 115 Hz. This value was a good match to the analytical derivation of the natural bending frequency.
Preparing the two test specimens with electrical strain gages (SGs) and measuring the loading when the diesel engine starts to come on line and at various power levels.

The following illustration shows a section through the valve gear and the positions of the individual strain gages:

First the important measurement technology information is listed below:

Mounting the tappets prepared with strain gages during the overhaul
Running the signal cables through prepared valve covers
Supplementing individual strain gages outside the valve gear to half bridges
Connecting to a carrier frequency amplifier
Including the operating parameters of power and speed
Digitizing and data acquisition in the adjoining diesel control station

To measure the loading, two tappets were fitted with HBM type K-LY41-3/120 strain gages. Four strain gages distributed lengthways around 90 °, were applied to the center of each tappet with Z70 adhesive. Type K-LY41 strain gages already have connecting cables insulated by Teflon, and these were long enough to bring out of a prepared valve cover. The measuring points were protected by sealing shrinkdown plastic tubing. As only the dynamic strains were of interest, a simple 2-wire circuit was used, with the addition of dummy resistors. Using the individual strain signals, it is possible to define the axial stresses and forces by summation and to define the bending strains by differentiation, and derived from this, to define the bending stresses and moments, including the direction of maximum bending vibration. The required calculations are shown below.

1. Normal force F_z

2. Bending moment M_x

3. Bending moment M_y

Calibration

The measuring chains are calibrated by applying an accurately known individual load (a weight of 10 kg) in the center of the tappet, which was flexibly mounted at both ends. The horizontal tappet was thus stressed to bending by the individual load in the center plus the line load from its own weight. The tappet spinning on its supports caused alternating bending strain in the principal axes. Comparing the measured values with the expected values in accordance with the laws of geometry and materials, produced a deviation of only 0.5 % (see Figure 2, V_target, V_actual).

Fig. 2: Diagram of measuring point calibration

Measurement and analysis

The two tappets prepared in the laboratory were installed on different cylinders on the emergency power diesel engine. Two replacement valve covers had been fitted with holes, through which the four core pairs were brought out and the measuring points supplemented to half bridges.

The valve tappets can rotate in their ball cups. Although this is desirable, there is no mechanism forcing it to rotate. Telemetric signal transmission would have had to be installed for continuous measurement or even for longer duration measurement. This amount of expenditure did not seem necessary, as with a shorter runtime, there were only a few tappet rotations to consider. The connection wires were wound in a spiral and of an adequate length.

Around 25 m away, the measuring points were connected to a carrier frequency amplifier and the signals were recorded by a data acquisition system (sampling rate 2 kHz). At the beginning of the start-up phase, the diesel engine was started by a turnover mechanism triggered by compressed air. Measurement at this slow rotation shows the pure loading on the tappets from the valve springs (Figure 3).

Fig. 3: The averaged axial strain over time for the initial turnover of the engine with compressed air in so-called turning mode

The measurement could be re-checked by comparing the axial forces with the static spring forces that were previously measured with a dial gage. Figure 4 below shows the actual loading curve of the diesel engine and the range of fluctuation of the maximum strains.

Fig. 4: The varying range of fluctuation of the individual strains

During operation, it could easily be seen at the cable pull-in that the tappets were rotating slowly with one moving a little faster than the other. This is shown in the alternating rise and fall of the bending strains, see Figure 4. This showed that the bending vibration had no fixed direction with regard to the tappet. However, it could be inferred that it did, with regard to the machine. This could only be regarded as a possibility, because there no reference marking generator was installed for tappet angle. It was also clear that buckling stress from tappet inhomogeneity was not the cause of the bending vibration.

As expected, the tappet forces showed a certain dependency on engine power, because pressures in the cylinder were acting on the valves, as well as the spring forces.

Figure 5 below shows a zoomed section of the high-pass-filtered bending strain signals during stationary partial load running.

Fig. 5: Bending strains on both tappets during stationary running

The bending vibration shows that at around 115 Hz, the natural frequency is intensely excited.

The engine deceleration spectrogram (Figure 6) shows that higher harmonics excite this frequency subject to the speed.

What was very interesting was high bending strain at higher harmonics. This was particularly noticeable at the ninth harmonic.

Fig. 6: Spectrogram of bending strain during engine deceleration

Figure 7 looks at a slow working cycle in engine deceleration.

Fig. 7: Analysis of a working cycle with regard to axial and bending strain

A cycle of duration T is approximately divided into T/4 for opening, T/4 for closing and T/2 for a passive phase. In T/2 of the active phase, the axial force rises and falls in accordance with the actuation of different springs. With the bending strain, there is a marked change in the prevailing bending stress at the time of maximum force and reversal of motion. The explanation for this is as follows. The motion changes the angle between the tappet and the rocker arm, causing friction torque in the ball cups. The change from upward to downward changes the direction of the friction torque.

In both phases, a higher-frequency oscillation is superimposed on the bending strain. This is the resonant vibration at 115 Hz, excited by the force pulse and unsteady friction in the bearings.

The change from upward to downward changes the direction of the friction torque very quickly. From the shape of the cam at the camshaft , it would be possible to estimate that of the 360 ° of a rotation, the change occurs in part of the angle, for example, within 30 °, which results in the dominance of the ninth harmonic.

In normal running at 1500 rpm, that is, for a vastly shorter period of time than that shown in Figure 7, the steep edge of the change in bending moment synchronizes the energy for resonant vibrati

Summary and perspective

The measurements made it possible to clarify the loading mechanism. The machine with rocker arms prescribes the direction of bending vibration, not the tappet with an assumed discontinuity.

The evaluation of vibration loading by level and frequency showed that with the original material, the limits of sustained loading can be reached and that tappets could occasionally fail.

The remedy was to make the tappets from a higher quality material.

<- retour: Articles techniques

Published Auteur : Burkhard Kempf, Thomas Marschner, AREVA NP GmbH, Erlangen

18.05.2009