Fatigue assessment of corroded turbine blade steels

Pitting corrosion is a critical issue for steam turbine operators since localised surface degradation causes stress concentration which may lead to fatigue failure. Dual certified 403/410 martensitic 12% Cr steel – which is a standard material for steam turbine blades in the low pressure part – was tested using ultrasonic fatigue testing technique. Experiments were performed up to the very high cycle fatigue regime on both smooth and pre-pitted specimens. For the latter, corrosion pits of defined size comparable to those found in failed turbine blades were generated artificially. Test environments were air at 90 ◦C and aerated 6 ppm Cl− solution at 90 ◦C (for details see [1]). In this work, the results of extensive fatigue tests [1, 2] are evaluated using two different approaches. Fatigue assessment was performed using the √ area parameter model developed by Murakami and Endo [3] and the small-crack model by El Haddad et al. [4]. The prediction equation for the first model is expressed as w = a(Hv + 120) (√ area )1/6 · ( 1 − R 2 ) (1)


Introduction
Pitting corrosion is a critical issue for steam turbine operators since localised surface degradation causes stress concentration which may lead to fatigue failure.Dual certified 403/410 martensitic 12% Cr steel -which is a standard material for steam turbine blades in the low pressure part -was tested using ultrasonic fatigue testing technique.Experiments were performed up to the very high cycle fatigue regime on both smooth and pre-pitted specimens.For the latter, corrosion pits of defined size comparable to those found in failed turbine blades were generated artificially.Test environments were air at 90 • C and aerated 6 ppm Cl − solution at 90 • C (for details see [1]).
In this work, the results of extensive fatigue tests [1,2] are evaluated using two different approaches.Fatigue assessment was performed using the √ area parameter model developed by Murakami and Endo [3] and the small-crack model by El Haddad et al. [4].The prediction equation for the first model is expressed as where w is the fatigue limit, H v is the Vickers hardness, R is the stress ratio √ area, the square root of the projection area of a defect and a is 1.43 for surface defects and 1.56 for internal three-dimensional defects.The exponent is defined by = 0.226 + H v • 10 −4 .El Haddad et al. proposed an equation of the form where 0 is the fatigue limit for smooth specimens, l is the crack/defect length and l 0 is the fictitious crack length with the threshold stress intensity factor for long cracks K th,lc and the geometry factor Y .

Results and discussion
Figure 1a shows the fatigue life curves using the model by Murakami and Endo [3] according to Eq. (1).For smooth specimens in air, non-metallic inclusions were found at the crack initiation sites, and the a Corresponding author: bernd.schoenbauer@boku.ac.atThis is an Open Access article distributed under the terms of the Creative Commons Attribution License 4.0, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.prediction using Eq. ( 1) provided good results.For pre-pitted specimens in air, the prediction error is within ca.±20% (for pit sizes of 50 m (+14% at R = 0.05 and −13% at R = 0.5), 100 m (+2% at R = 0.05 and −14% at R = 0.5) and 250 m (−21% at R = 0.05)).Nevertheless, no acceptable results were found for 6 ppm Cl − solution where the prediction error is as high as −53%.This is not surprising since the model´s main parameter is the Vickers hardness which is a material constant and any environmental dependence is not considered.
In Figure 1b, the test data for pre-pitted specimens were evaluated using Eq. ( 2) according to El Haddad et al. [4].Half of the pit width on the surface was used as the defect size l and the geometry factor was empirically determined (Y = 0.65) as discussed in [1].0 and K th,lc were experimentally determined for different R-ratios and environments.The prediction error is significantly lower compared to the √ area model.Although there is an underestimation in air at R = 0.5 of 18%, the prediction error is mostly within ±10%.

Conclusions
Comparison of the predictive models by Murakami and Endo [3] and El Haddad et al. [4] were made.Whereas the √ area model is well applicable to failure resulting from inclusions in smooth specimens, the El Haddad et al. approach shows a higher accuracy for pre-pitted specimens.Furthermore, its main advantage is the applicability to different environmental conditions.

Figure 1 .
Figure 1.Number cycles to failure N f vs. normalised stress range / w using the model by Murakami and Endo (a) and El Haddad et al. (b).For pre-pitted specimens, the pit size is indicated.