PSI - Issue 57

Malik Spahic et al. / Procedia Structural Integrity 57 (2024) 833–847 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

841

9

Laborelec developed custom-made UT inspection tools that enable to detect potential cracks in the slot bottom area without having to remove the blades (add figure of UT simulation). As exemplified further on in the risk mitigation actions, surface machining is an efficient way to extend the lifetime of a steam turbine rotor. Ideally, the surface is machined before the onset of cracking, thereby removing the surface layers that have accumulated the largest thermo-mechanical fatigue damage. While rather straightforward for a disc type rotor, this is more complicated for a drum-type rotor where the slot bottom needs to be machined, requiring removal of the blades. 4.4. Protection As mentioned in Section 1, steam turbine rotors are typically protected by a RSE (Figure 4). The RSE should protect the steam turbine rotor from excessive thermal stress during the start-up. In many cases however, including those where cracks have been found, the RSE was found to be malfunctioning. In some cases, the stator temperature probe was found plastically deformed, thereby introducing important errors in the temperature readings. Other RSE malfunctioning involves issues in the calculation itself leading to an underestimation of the real stress levels in the rotor. 5. Risk mitigation actions A steam turbine rotor is in practice only disassembled during the major overhaul given the unavailability and the associated cost. The maintenance interval between major overhauls is OEM specific, but typically determined by the equivalent operating hours (EOH) which is a weighted combination of the number of starts and operating hours. In practice, it can take 10 years in between major overhauls. It is therefore important to ensure that the steam turbine rotor can safely operate until the next major overhaul. In what follows, some actions during and in preparation of the major overhaul are listed to mitigate the risk for unexpected cracking and resulting unavailability of the unit. 5.1. Lifetime assessment As there is no established way to determine the creep-fatigue lifetime consumption by on-site inspection, calculations are still indispensable. To enable calculation over the full operational history of a unit, a simplified one dimensional model is used which is calibrated based on a more detailed two-dimensional axisymmetric finite element model covering a reference period with several cold, warm and hot starts. The different steps in the procedure are explained for three different types of rotors that are currently operational within the ENGIE fleet and denoted as Rotor A, B and C in what follows. The geometry and corresponding finite element models of the different rotors are shown on Figure 12. For Rotor A, the power plant operating the rotor found a crack during an inspection. For the two other cases, no crack was found at the moment of the analysis. The 2D rotor is split in many different surfaces (around 100). For each surface, the steam temperature and convective heat transfer coefficients are determined throughout the entire reference period (see Figure 13). This is done based on thermodynamics using real life measurement data (e.g. main and gland steam temperature/pressure, rotational speed, control valve position), along with empirical and analytical formulae found in literature [7] or estimated through internal studies. An example of the resulting rotor temperature during start-up can be found on Figure 1 and an example of the stress distribution is shown on Figure 14 for Rotor B at the instant of maximum stress over the reference period. The highest (compressive) stress is present at the first stage disc. A first estimation of the lifetime consumption can be done by extrapolating the reference period over the full operational history of the unit. As most units have many different types of starts or changed their start-up profile throughout the years, the reference period is not always representative for the full history. A 1D model has therefore been developed which is calibrated based on the 2D model and enables calculation over the full operational history.