To Örebro University

There is a prevailing assumption that the largest inclusions in steel volumes follows mode I of the Generalized Extreme Values (GEV) distribution. In this work, the GEV distributions of non-metallic inclusions in six different high performance steels, of different grades and processing routes, were investigated by means of fractography of inclusions causing failure in ultrasonic fatigue testing to one billion cycles and all three modes of the GEV were found for the different steel grades. Values of the shape parameter ξ of the GEV distribution as high as 0.51 (standard deviation 0.11) were found in one steel grade. Thus, the present results show that the assumption of GEV-I (Gumbel, LEVD) distribution has to be substantiated before being used to estimate the size of the largest inclusions.

The presence of defects, such as oxides and other non-metallic inclusions, is an important factor in determining the properties of steels. Due to improvements in the manufacturing of high-quality steels, the amount of large defects has decreased and therefore it has become increasingly difficult to accurately determine their distribution using conventional methods. Previously, a method for estimating the distribution of large defects using a conversion from the Gumbel distribution has been presented. However, it has been shown that the Gumbel distribution is not always appropriate for modelling the sizes of the largest defects and that the Generalized Extreme Values (GEV) distribution should be used instead. In this work a more general method for the estimation of the total volume distribution of large defects isproposed, showing how the volume distribution may be calculated from the estimated parameters for the GEV distribution. The new method is applied to the results of a series of specimens made from high-quality tool steel tested in ultrasonic resonance fatigue. Possible methods for obtaining the confidence limits of thevolume distribution are also discussed.

This paper reports on results of very high cycle fatigue tests on a steel with a relatively low number of defects per unit volume. Two series of newly designed dog-bone specimens, with low stress gradients, were tested at two constant stress levels, aiming at providing fatigue life lengths of 10⁶–10⁸ cycles and 10⁷–10⁹ cycles, respectively. Fatigue-initiating defects on the fracture surfaces were measured and the corresponding size distributions were analysed to investigate the influence of the elevated stress level needed to achieve shorter life lengths. The presented results indicate a difference between the distribution of initiating defects in the long and shorter life length test series. The main conclusion is that fixed stress level fatigue testing to life lengths around 10⁷ cycles may be a suitable method to estimate the content of large defects in low defect steel materials, provided a suitable test specimen geometry with low stress gradients is used.

Fatigue strength is an important material property for many tooling applications, particularly in high performance applications. The research in Very High Cycle Fatigue (VHCF) has demonstrated that the traditional fatigue limit may not be valid for many materials subjected to 10⁷ or more load cycles. Presently, both materials data and mechanism knowledge is missing on VHCF applications, even though many components are run at these life lengths. The fatigue strength is commonly controlled by different defects initiating failure, as in well controlled laboratory experiments may be internal inclusions. In this paper VHCF experimental testing was accomplished by the use of ultrasonic fatigue testing run at 20 kHz allowing long life evaluation within reasonably short test time. Fatigue strength, failure mechanisms and inclusion content were accordingly assessed. Fatigue strength data on H13 tool steel are presented, as well as a statistical approach considering available defect distribution and load distribution in the critically stressed volume, important to both steel supplier and end-user.

The distributions of large non-metallic inclusions in two steel grades have been investigated using light optical microscopy, scanning electron microscopy and ultrasonic fatigue testing in the gigacycle range. The different methods have inherently different capabilities for finding inclusions in different size ranges. A measure of the distribution of large inclusions is proposed as the size S at which half of the fatigue specimens are expected to contain at least one inclusion of size S or larger, corresponding to 50% failure probability. Values of S are obtained using the volume distribution estimated by the three methods. Extrapolation from microscopy measurements on surfaces agree with fatigue fractography results regarding density of large inclusions, as measured by the proposed ranking variable S.