Engineering Failure Analysis 18 (2011) 256–270

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

Failure analysis of a Pelton turbine manufactured in soft martensitic

stainless steel casting
D. Ferreño a,⇑, J.A. Álvarez a, E. Ruiz a, D. Méndez a, L. Rodríguez b, D. Hernández b
a
LADICIM, University of Cantabria, ETS Ingenieros de Caminos, Av/Los Castros s/n, 39005 Santander, Spain
b
Technical & Scientific Research Services, University of Cantabria, R&D Research Building, Plaza de la Ciencia s/n, 39005 Santander, Spain

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, the failure of a large Pelton turbine, arising immediately after receiving a
Received 4 March 2010 quenching and tempering heat treatment, is analysed. The turbine was manufactured in
Received in revised form 9 September 2010 ASTM CA-6NM soft martensitic stainless steel. A thermo-mechanical finite elements model
Accepted 9 September 2010
was developed to evaluate the homogeneity of the heat treatment; for this purpose, the
Available online 17 September 2010
thermal histories of internal and external points of the component were compared, and
the possible existence of temper embrittlement was assessed. Moreover, the thermal stres-
Keywords:
ses during heat treatment were obtained in order to perform a failure analysis. After ana-
Pelton turbine
CA-6NM stainless steel
lysing the fracture surface through visual inspection and scanning electron microscope
Finite elements fractography, the material properties were characterised all around the fracture surface
Temper embrittlement including optical and transmission electron microscopy, chemical composition and Vickers
and Charpy impact tests. All this experimental information allowed a satisfactory under-
standing of the phenomenon to be obtained and a failure analysis to be performed in order
to justify the fracture of the component.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction and objectives

The Pelton wheel is among the most efficient types of water turbines. This kind of impulse machine uses Newton’s second
law to extract energy from a jet of fluid. Many variations of impulse turbines existed prior to Pelton’s design, but they were
highly inefficient as the water leaving these wheels typically had high speed, and carried away much of the energy. Fig. 1
shows a scheme of the Pelton wheel whose failure is analysed in this paper. A Pelton turbine consists of a Pelton wheel
mounted on a rotating shaft or rotor. The wheel (or runner) is composed of a circular disc and a set of cup-shaped blades,
called buckets, placed at equal spacing around its circumference (see Fig. 1).
The high speed water jets running the Pelton wheel turbine are obtained by expanding the high pressure water through
nozzles to the atmospheric pressure. Nozzles are arranged around the disc such that the water jet emerging from a nozzle is
tangential to the circumference of the wheel (see Fig. 2). According to the available pressure of water and the operating
requirements, the shape and number of nozzles placed around the Pelton wheel can vary, usually in the range 8–20; in this
case, the wheel includes 18 buckets (see Fig. 1).
The high pressure water can be obtained from any water body situated at some height or streams of water flowing down
the hills. As schematically depicted in Fig. 2, the water jets emerging form the nozzles strike the buckets at splitters, placed at
the middle of a bucket, from where jets are divided into two equal streams. These streams flow along the inner curve of the
bucket and leave it in the direction opposite that of the incoming jet; therefore, the change in momentum of the water

⇑ Corresponding author.
E-mail address: ferrenod@unican.es (D. Ferreño).

1350-6307/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfailanal.2010.09.005
 D. Ferreño et al. / Engineering Failure Analysis 18 (2011) 256–270 257

Fig. 1. Schematic perspective of the Pelton wheel whose failure is analysed in this research.

Outoming jet

Nozzle

Incoming water jet

Nozzle Outoming jet

Water jet
(a) (b)
Fig. 2. Sketch showing the process of momentum transfer from the water stream to the bucket.

stream – and, as a consequence, the impulse received by the blades – is the maximum possible. This impulse generates the
torque and rotation in the shaft of a Pelton turbine.
The runner is the most relevant component in a complete Pelton turbine device; the other elements (as the manifold, the
housing, the nozzles or the turbine shaft) play a complementary role in the process for energy production. In order to facil-
itate the manufacture and maintenance, the disc and the buckets composing a Pelton wheel were traditionally manufactured
separately and then joined through rivets (as in the original design by Lester Allan Pelton), screws or welding. However, in
recent designs, where higher specific speeds are required, the complete Pelton wheel is fabricated in one part, using casting
techniques, thus leading to an improvement in stiffness and simplicity in the assembly.
In this paper, the failure of a large Pelton turbine (sketched in Fig. 1), immediately after receiving a quenching and tem-
pering heat treatment, is analysed. The external diameter of the wheel (including the disc and the blades) is approximately
1400 mm whereas the total weight is about 5200 kg. The turbine was manufactured in ASTM CA-6NM soft martensitic cast
stainless steel (G-X4 Cr Ni 13-4, according to the German DIN designation). Quenched and tempered CA-6NM steel is widely
used for hydraulic turbine runner castings, since it possesses high toughness together with excellent resistance to cavitation
and erosion [1–4]. A complete revision of the microstructure, mechanical and fracture properties is presented in [5]. In Fig. 3,
a set of photographs of the broken component and the fracture surfaces can be appreciated.
258 D. Ferreño et al. / Engineering Failure Analysis 18 (2011) 256–270

Fig. 3. Several photographs showing the broken component and the fracture surface.

2. Material properties

Special care must be taken to select an adequate material to manufacture a Pelton wheel, as it must resist erosion, cav-
itation and fatigue during its operative life. Soft martensitic stainless ASTM CA-6NM–13Cr–4Ni cast steels are widely used
for hydraulic turbine runner castings, since they possess excellent resistance to cavitation, erosion and have high toughness
[1–5]. They have lower carbon and 3–6% higher nickel content compared with normal martensitic stainless steels. Type CA-
6NM is an iron–chromium–nickel–molybdenum alloy that is hardenable by heat treatment being characterised by good cor-
rosion resistance and weldability, together with a superior combination of high tensile strength and impact properties. The
chemical composition of this steel, according to the ASTM specifications A352, A487 and A743 is presented in Table 1 (there
are slight differences with the limits imposed by other specifications as DIN, VSM or JIS).
The alloy is normally used in the quenched and tempered condition in which the microstructure is essentially 100% mar-
tensite (with or without delta ferrite). CA-6NM can contain appreciable amounts of retained austenite because this structure
provides the optimum combination of strength, ductility, hardness, and toughness. Variations in heat treatment can be se-
lected to enhance one or more of these properties. Improved corrosion resistance, particularly resistance to sulfide stress cor-
rosion, can be obtained with a lower carbon as in grade CA-6NM Class B (ASTM A487).
The alloy is hardened by heating between 1038 and 1066 °C (austenising stage) followed by cooling in either air or oil.
After the castings have cooled below the martensite finish temperature, which varies with the compositional balance, they
should be tempered as soon as possible. Depending on strength requirements, the alloy is tempered at 316 °C or more com-
monly in the range of 593–621 °C. Tempering in the vicinity of 482 °C should be avoided because lower toughness will result.
Some reaustenitisation may occur if tempering temperatures above 649 °C are employed, and upon cooling, the microstruc-
ture may contain untempered martensite. Double tempers are employed to achieve hardness values below 22 HRC for cast-
ings intended for wet H2S environments. A typical double temper heat treatment would consist of a 677 °C temper followed
by a 607 °C temper. The highest strength and hardness are obtained by tempering at 316 °C; however, impact strength is

Table 1
Chemical composition, according to ASTM specifications (% wt.).

C Cr Ni Mo Mn Si P S Cu + W + V
Min 11.5 3.5 0.4
Max 0.06 14.00 4.50 1.00 1.00 1.00 0.04 0.03 0.50
 D. Ferreño et al. / Engineering Failure Analysis 18 (2011) 256–270 259

reduced by 50% and ductility is reduced to about 12%. Holding times for austenitising and tempering will vary with the thick-
ness of the casting sections involved, but should be sufficiently long to heat all sections to a uniform temperature
throughout.
In Table 2, the typical and required mechanical properties at room temperature are included [6]. The first column corre-
sponds to the typical mechanical properties (air cooled from >1038 °C and tempered at 593–621 °C) while the second rep-
resents the minimum tensile and toughness requirements according to ASTM A743 and 757. As can be appreciated, this steel
shows high strength and ductility. Moreover, for the sake of completeness and convenience for future discussions, Fig. 4 [6]
shows the Charpy-V notch impact energy data band for CA-6NM steel. From this figure, the typical range of Charpy energies
at room temperature 90–140 J is obtained. Moreover, the DIN specification demands (for the quenched and tempered con-
dition) the mean value of the three tests performed at 0 °C to be higher than 50 J and the smallest individual value not to fall
below 35 J.

3. Failure description: Hypothesis to explain the failure

As can be appreciated in Fig. 3, this analysis is focused on assessing the fracture of a large component (as mentioned be-
fore, the total weight of the wheel is of about 5200 kg). According to the available information, provided by the manufacturer
of the wheel, the fracture process took place immediately after finishing the quenching and tempering heat treatment. The
fracture surface (see Fig. 3) is completely plain without any noticeable plastic deformation; thus, a brittle failure occurred.
The field of fracture mechanics consists of quantifying the relationship between flaw size, stress and material toughness so
that the severity of flaws can be assessed in components and structures. In this sense, several (not necessarily independent)
hypotheses were considered to explain the reasons leading to this brittle fracture, which are summarised below:

 The existence of defects in the part was analysed. It must be pointed out that only a set of small sized samples were pro-
vided by the owner of the wheel, and, therefore, the scope of this study was restricted to this available material. In Fig. 5, a
scheme showing the regions around the fracture surface where the coupons were extracted is shown. Nine coupons of
about 60  40  40 mm3 were available to perform the experimental work; they were obtained both from the interior
(coupons 3, 6, 7 and, specially, coupon 9) and the contour of the part (coupons 1, 2, 4, 5 and 8), in order to detect any
heterogeneity. The fracture surface of these coupons was studied and, after that, small samples were taken to analyse
the microstructure in order to detect differences related to the influence of the heat treatment.
 Several material properties (Vickers and Charpy impact energy) were characterised all around the fracture surface: it is
well known that, the larger the size of the part, the greater the problems encountered in manufacturing and the higher the
possibility of finding material heterogeneities.

As the failure took place before the wheel became operational, and immediately after applying the quenching and tem-
pering heat treatment, it is important to obtain the thermal stresses in the part caused by this treatment. For this purpose, a
thermo-mechanical finite elements (FE) model was developed to simulate the two stages composing the heat treatment.
Moreover, this model makes it possible to determine the differences in heat treatment experienced by an internal and an
external point of the part which could eventually lead to microstructural differences. This is a relevant point because, as
underlined by Iwabuchi [1,7], the difficulty in obtaining a uniform temperature during heat treatment may create a critical
condition for cracking and slow cooling down from the tempering causing temper embrittlement, which results in an inter-
granular fracture. Several authors have pointed out that the impact notch toughness deteriorates in steels slowly cooled from
tempering temperatures [8–10] Such toughness deterioration, known as temper embrittlement, imposes many practical
problems.

4. Experimental

As mentioned above, nine coupons taken from different regions all around the fracture surface of the wheel (see Fig. 4)
were provided by the manufacturer and available for the experimental work. The following actions were performed:

Table 2
Typical and required (ASTM A743 and 757) room temperature mechanical and fracture
properties of the CA-6NM steel [6].

Typical Required (A743, A757)

Tensile strength (MPa) 826 757
Yield strength (MPa) 688 550
Elongation, in 2 in. (%) 24 15
Reduction of area (%) 60 35
Brinell hardness 268
Charpy-V notch ( 73 °C) 27.1/16.2a
a
27.1 J average for three specimens, 16.2 minimum allowed for one specimen only.
260 D. Ferreño et al. / Engineering Failure Analysis 18 (2011) 256–270

Fig. 4. Charpy-V notch impact energy data band for CA-6NM steel (obtained from [6]).

Fig. 5. Fracture surface showing the regions of the coupons available for the analysis.

 First, the fracture surfaces were inspected, visually and through scanning electron microscopy (SEM) (see Section 4.1).
 Second, small samples were taken from the coupons and the chemical composition was determined by Optical Emission
Spectroscopy (OES) (see Section 4.2).
 The same samples were used to analyse the microstructure in the different regions around the surface through optical and
transmission electron microscopy (TEM) (see Section 4.3).
 Concerning the material properties, a large number of micro hardness Vickers tests were performed and, moreover, two
Charpy impact specimens were obtained from several of the coupons available (it was not possible to obtain complete
Charpy specimens from the rest of them) (see Section 4.4).

4.1. Inspection of the fracture surface

The photographs provided by the owner of the wheel, Fig. 3, were visually examined. The main relevant feature is that no
plastic deformation was involved and therefore, brittle fracture took place. The main purpose of this stage is to identify the
fracture mechanisms developed in the material during the fracture and to analyse the possibility of pre-existing defects. It is
worth noting that for a brittle fracture to occur in a normally ductile material like CA6-NM steel, a stress concentrator must
be present which must be large and sharp enough to be a critical flaw in the presence of an existing tensile stress.
The fractographic study performed by SEM has revealed two interesting features. First, the dominant failure mechanism
consists of microvoid coalescence (see Fig. 6c), which is a ductile mode of propagation. No intergranularity was observed and
only small amounts of transgranular decohesion (cleavages) were appreciated. Second, the fracture surface is highly irregular
(see Fig. 6a) and several regions with material discontinuities were detected. It is worth emphasising that, although only the
fracture surfaces of the coupons provided by the owner were available for the study, this kind of features were repeatedly
 D. Ferreño et al. / Engineering Failure Analysis 18 (2011) 256–270 261

Fig. 6. SEM analysis of the fracture surface of the Pelton wheel.

observed in all the surfaces analysed. On the other hand, only small size irregularities were detected: this is a consequence of
the reduced area which is possible to analyse with the SEM, and, in this sense, further research must be done about this issue
(see Sections 4.3 and 4.4).
The irregular shape of these discontinuities suggests that they consist of shrinkage cavities or hot tears, rather than poros-
ity or gas holes because porosity implies the existence of small smooth-faced cavities. Shrinkage cavities appear during solid-
ification as a result of the reduction in volume when metal changes from the liquid to the solid state. They occur in situations
where molten metal is not available to compensate for the volume decrease during solidification. The hot tears are jagged
crack type defects resulting from stresses imposed on the cast metal when it is just below the solidification temperature and
thus is in a weak condition. The stresses usually arise when the casting is restrained during contraction by the mould, or by
an already solid thinner section (as in the case here analysed). This description is coherent with Fig. 6b and d. At the same
time, the chemical composition of several inclusions detected in the fracture surface was obtained through Energy Dispersive
Spectroscopy, (EDS) thus determining that they correspond (see Fig. 7) to alumina inclusions.

4.2. Chemical composition

The chemical composition of coupons 1–4 and 9 was determined by Optical Emission Spectroscopy (OES). Table 3 in-
cludes the results obtained. As several tests were performed on each of the coupons, the average and the standard deviation
(SD) are included in all cases. The comparison between the results in Table 3 and the limits of the material specification (dark
cells in Table 3 indicate composition out of range) shows that the chemical elements analysed satisfy the material require-
ments, with the exception of molybdenum: values in the range 1.20–1.40% wt. are obtained whereas the limit in the spec-
ification is 1:00% wt. Moreover, the contents of silicon, which is the most embrittling element, are well below the maximum
allowed. Concerning the molybdenum, it must be considered that CA-6NM is CA6N modified with molybdenum to improve
corrosion resistance and castability. Molybdenum additions improve resistance to pitting and crevice corrosion in chloride
containing environments and corrosion by sulfuric, phosphoric, and hydrochloric acids. The elevated temperature mechan-
ical properties of austenitic stainless steels and the strength and tempering resistance of martensitic stainless steels are im-
proved by molybdenum. Iwabuchi [7] has made an in-depth study of the influence of several factors, including the
chemistry, on the tempering embrittlement. As indicated above, this phenomenon takes place during slow cooling from
262 D. Ferreño et al. / Engineering Failure Analysis 18 (2011) 256–270

Fig. 7. Photograph of an alumina inclusion detected in the fracture surface.

Table 3
Experimental chemical composition, obtained from OES (%wt.).

Coupon C Cr Ni Mo Mn Si P S Cu + V + W
1 Average 0.055 11.90 4.11 1.25 0.72 0.53 <0.015 <0.015 0.16
SD 0.005 0.26 0.17 0.15 0.07 0.04 – –
2 Average 0.053 11.70 4.06 1.40 0.74 0.50 <0.015 <0.015 0.17
SD 0.005 0.26 0.17 0.15 0.07 0.04 – –
3 Average 0.052 11.74 4.08 1.37 0.72 0.49 <0.015 <0.015 0.17
SD 0.005 0.26 0.17 0.15 0.07 0.04 – –
4 Average 0.054 11.74 4.03 1.40 0.75 0.48 <0.015 <0.015 0.18
SD 0.005 0.26 0.17 0.15 0.07 0.04 – –
9 Average 0.050 11.86 4.13 1.20 0.70 0.49 <0.015 <0.015 0.16
SD 0.005 0.26 0.17 0.15 0.07 0.04 – –

the tempering temperature; nevertheless, Iwabuchi concludes that the addition of molybdenum reduces or suppresses this
toughness degradation. This evidence would suggest that not too great an importance should be placed on an excess of
molybdenum in the material fracture properties.

4.3. Metallographic examination through optical and transmission electron microscopy

4.3.1. Optical microscopy

The microstructure of the material was examined through optical microscopy and TEM. Small samples were taken from
the coupons available and, after grinding and polishing up to mirror finish, they were etched with Vilella’s reagent in order to
reveal the microstructure. The general microstructure consists of tempered martensite with the presence of disperse small

Fig. 8. Material microstructure as obtained through optical microscopy.

 D. Ferreño et al. / Engineering Failure Analysis 18 (2011) 256–270 263

elongated islands of ferrite (see Fig. 8). This result was to be expected, taking into consideration the Schaeffler diagram
(Fig. 9). This is one of the most convenient ways of representing the effect of various elements on the basic structure of chro-
mium–nickel stainless steels. It plots the compositional limits at room temperature of austenite, ferrite and martensite, in
terms of nickel and chromium equivalents (defined in Fig. 9), Creq and Nieq, respectively. According to the chemical compo-
sition of CA-6NM steel (Table 1), Creq = (11.6–16.5) and Nieq = (3.5–6.8). This information is superimposed in Fig. 9.
During the inspection, several cracks were detected. A representative example is shown in Fig. 10. Notice the general as-
pect consisting of a sharp crack following the length of a ferrite island. Therefore, the ferrites act as elements of decohesion of
the martensite matrix and, as a consequence of their sharpness, they are susceptible to initiate a subsequent fracture process.
This result is in agreement with what exposed in Section 4.1; moreover, as again only small defects were detected, further
insight about this matter is present in Section 4.4.

4.3.2. TEM
In previous works [1], Iwabuchi has suggested that the temper embrittlement in 13Cr–4Ni cast steels is related to car-
bides precipitated at prior austenite grain boundary. This type of carbide precipitates is, in principle, observed by immersing
the as-polished specimen in a dilute nitric acid aq. solution. This kind of etch provided negative results and, as a conse-
quence, the specimens were etched with Vilella’s reagent, thus enabling the microstructure to be revealed. The importance
of carbide precipitates on temper embrittlement is associated with the groove along the austenite grain boundary.
Fig. 11 shows the TEM micrographs of carbon replicas extracted from the available specimens. The grain boundaries are
highlighted by means of discontinuous lines superimposed on the micrograph, to facilitate the analysis. As can be appreci-
ated, coarse carbides are observed densely on the matrix but they are completely absent on prior austenite grain boundaries.
According to this result, the possibility of a material temper embrittlement must be ruled out.

30

25
Nieq = %Ni + 30%C + 0.5%Mn

A
20

15

10
M
A+M+F
5 CA6NM
F
F+M M+F
0
0 5 10 15 20 25 30 35 40
Creq = %Cr + %Mo + 1.5%Si + 0.5%Nb + 2%Ti

Fig. 9. Schaeffler diagram including the region corresponding to the CA-6NM stainless steel.

Fig. 10. Example of a cracked ferrite island found during the metallographic examination.
264 D. Ferreño et al. / Engineering Failure Analysis 18 (2011) 256–270

Fig. 11. TEM micrograph of a carbon replica extracted from the available specimens (the grain boundaries are highlighted by means of discontinuous lines).

4.4. Mechanical properties

Vickers tests were performed on coupons 1–4 and 9 (1, 2 and 4 for taken from the contour of the wheel whereas 3 and 9
are internal coupons). In each case, 25 Vickers test were performed (200 g, 25 s) after carefully polishing the surfaces. The
results, including the average and the standard deviation, are represented in Fig. 12. As can be appreciated, the results (tak-
ing the deviations into consideration) are included in the range 300–420 HV. In all cases, a noticeable scatter was obtained
and, as a consequence, the slight differences between the average values from different regions are included in the experi-
mental scatter. Therefore, from the point of view of Vickers tests, no heterogeneity was detected.
As mentioned before, it was possible to obtain two Charpy specimens from each one of some selected coupons (the rest of
specimens were too small to machine a complete Charpy sample). Specifically, the specimens were extracted from external
coupons 4 and 8 and internal coupons 6, 7 and 9. The experimental results, expressed in terms of the absorbed energy at
room temperature, are included in Fig. 13; they belong to the interval 40–64 J. Concerning the inhomogeneity of the wheel,

500

450

400
HV

350

300

250

200
1 2 3 4 9

Fig. 12. Results of the Vickers tests performed on selected coupons.

70

65

60

55
E (J)

50

45

40

35
4 6 7 8 9
30

Fig. 13. Results of the Charpy impact tests (absorbed energy) performed on coupons 4, 6, 7, 8 and 9 (two Charpy specimens were obtained from each of the
coupons).
 D. Ferreño et al. / Engineering Failure Analysis 18 (2011) 256–270 265

no definitive conclusions can be established, as the scatter is larger than the absolute differences between regions. In Fig. 4
[6], the Charpy-V notch impact energy data band for CA-6NM steel was represented. From this figure, a typical range of Char-
py energies at room temperature of 90–140 J was obtained. Therefore, the material characterised here offers very poor re-
sults in toughness. Moreover, even though the tests were performed at room temperature, the results hardly fulfil the
DIN requirement for tests performed at 0 °C (the mean value of the three tests performed at 0 °C must be higher than
50 J and the smallest individual value must not fall below 35 J).
In [7] an exhaustive experimental work was developed to correlate the hardness and room temperature energy (corre-
sponding to shelf energy) for quenched and tempered CA-6NM stainless steel. According to this research, the upper shelf
energy, Eupp, is linearly related with the hardness Vickers, as represented in Fig. 14 (this figure is reproduced from [7]).
The experimental points obtained in the present research were also included in the figure, therefore highlighting the limited
toughness of the material studied.
After performing the Charpy tests, the broken surfaces were carefully examined in order to identify the mechanisms that
lead to the fracture. Several of the photographs can be appreciated in Fig. 15, where the code of the specimen is also included.
The first relevant feature is that a noticeable plastic deformation took place during fracture. The photographs show a large
amount of lateral expansion in all cases, together with a fibrous appearance. Another relevant feature, also evident in the
photographs, but extensible to the rest of the tests performed, is that the morphology of the fracture surfaces shows strong
discontinuities and deep notches; in some cases, the propagation of the crack was clearly out of the plane of symmetry of the
specimen, even being able to tear the notch (see for example the specimen (9.2)). These characteristics are in agreement with
the results of the fracture surface of the wheel, analysed in Section 4.1 and with the metallographic examination present in
Section 4.3. Moreover, whereas in the previous analysis only small irregularities or defects were determined, here large dis-
continuities in the material were detected.

4.5. Conclusion concerning the experimental results

In Section 3, several hypotheses were proposed to explain the failure of the Pelton wheel here analysed. The experimental
results obtained allow the following conclusions to be drawn.

 The existence of defects in the part was proved. The SEM inspection of the fracture surface (Section 4.1) has revealed the
presence of material discontinuities interpreted as shrinkage cavities or hot tears. The metallographic examination
through optical microscopy made it possible to detect several sharp cracks following the length of the elongated ferrite
islands, which are thus susceptible to initiate a subsequent fracture process. Finally, the study of the fracture surfaces of
the Charpy tested specimens showed a morphology consisting of strong discontinuities and deep notches; in some cases,
the propagation of the crack was clearly out of the plane of symmetry of the specimen, even being able to tear the notch.
 Concerning the chemical composition of the material, slight discrepancies were detected between the experimental
results and the requirement of the specification. These are far from being disturbing as the excessive amount of molyb-
denum cannot be responsible for any material embrittlement.
 None of the analyses performed (microstructure, chemistry, hardness Vickers, Charpy impact tests) showed any signifi-
cant difference between the material extracted from different regions of the Pelton wheel. As a consequence, the hypoth-
esis of material heterogeneities must be ruled out. Nevertheless, the results of the FE model developed will provide
further insight concerning this topic (Section 5).

The above evidence suggests that an inadequate manufacturing process of the Pelton wheel is the main cause of failure.
None of the results indicated that the quenching and tempering heat treatment had any influence in the fracture process. In
particular, no proofs of temper embrittlement were detected.
Impact Energy at 300 K, Eupp (J)

180
160
140
Eupp= 299 - 0.715·HV
120
100
80
9
60
40 4
20
0
200 250 300 350 400 450
Hardness Vickers, HV

Fig. 14. Relationship between hardness and upper shelf impact energy [7]. The experimental results obtained in the present research are also represented.
266 D. Ferreño et al. / Engineering Failure Analysis 18 (2011) 256–270

Fig. 15. Macrographs of the fracture surface after performing Charpy tests (selected coupons).

5. Finite elements thermo-mechanical simulation of the heat treatment

It is evident that the heat treatment applied to the wheel played a role in the fracture process, as this took place imme-
diately after the treatment. The purpose of the thermo-mechanical FE model developed is twofold. First, in order to confirm
the homogeneity of the heat treatment, the evolution of temperature with time in three different points of the wheel was
analysed. With this information, it is possible to elucidate whether any change in microstructure might have occurred. This
analysis is presented in Section 5.1. Second, to perform a failure analysis, it is necessary to know the stress state in the part;
in this sense, in Section 5.2, the thermal stresses during the application of the heat treatment were obtained.
The FE simulation consists of a three dimensional (3-D) model of the part. A sequential thermo-mechanical simulation
was performed: first, the temperature fields were obtained and from them, in a second stage, the thermal stresses were cal-
culated. Sequential analysis requires noncoupled conditions to be considered, thus assuming that the stresses do not affect
the thermal distribution in the part. The SOLID70 element [11] was chosen for the thermal stage of the simulation. This ele-
ment has eight nodes with a single degree of freedom, temperature, at each node. The element is applicable to a 3-D, steady-
state or transient thermal analysis, as in this case. The SOLID45 was used to model the mechanical process obtaining the
stress fields as a function of time. This element is defined by eight nodes having three degrees of freedom at each node:
translations in the nodal x, y, and z directions. A full integration method was chosen in all cases.
The thermal boundary conditions consist of a convection process applied to the external surface of the wheel (controlled
by the film coefficient, h) following the temperature profile shown in Fig. 16, which corresponds to the furnace temperature.
No detailed experimental information about the thermal properties was available and, for this reason, generic values were
considered. The thermal, physical and mechanical properties used for the FE calculations are included in Table 4. As can be
 D. Ferreño et al. / Engineering Failure Analysis 18 (2011) 256–270 267

1200
1040 ºC, 8 hours
1000

Temperature, T (ºC)
800
610 ºC, 8.5 hours
600
Air quenching
Air
(forced)
quenching
400 (natural
convection)
200

0
0 5 10 15 20 25 30 35
Time, t (h)

Fig. 16. Temperature profile corresponding to the applied heat treatment (furnace temperature).

Table 4
Physical, thermal and mechanical properties used for the FE calculations.

Density, q (kg m 3) 7858

Thermal conductivity (J m 1 s 1 K 1) 43.6
Film coefficient (free), h (J m 2 s 1 K 1) 30
Film coefficient (forced), h (J m 2 s 1 K 1) 450
Specific heat, c (J kg 1 K 1) 536
Coefficient of thermal expansion, a (K 1) 1.31  10 5

Elastic modulus, E (GPa) 200

Tangent modulus, E (MPa) 600
Yield strength (MPa) 826
Poisson ratio, m 0.30

appreciated, two different values of film (convection) coefficient were considered, one for free convection (corresponding to
the quenching of the tempering process) and the other for forced convection (quenching after the austenite stage). Concern-
ing the mechanical properties, a bilinear kinematic hardening material model was used with the typical properties provided
in Table 2 [12]. In Table 4, the elastic modulus, the tangent modulus (after yielding), the yield strength and the Poisson’s ratio
are included.

5.1. Study of homogeneity during the heat treatment

The FE model allowed the evolution of temperatures in the wheel to be determined as a consequence of the quenching
and tempering heat treatment applied. For the purpose of comparing the efficiency of the treatment, three representative
nodes were selected, as described in Fig. 17. Two of them belong to the disc of the wheel; the first one is an internal node
whereas the second is placed in the contour of the part (in contact with the surrounding convective medium). The last node
is also an external one, but taken from the contour of one of the blades.
The results of the comparison are shown in Fig. 18 where the temperature of each of the three points is represented as a
function of time. For the calculations, it was considered that the process starts with the quenching performed after the
austenising. It is worth noting several evident features. Firstly, there is a noticeable difference between the temperature

11

2
1
3
Y
X Z
1: Internal node (disc)
2: External node (disc)
3: External node (blade)

Fig. 17. Scheme showing the nodes selected for the comparison of temperature profiles during the heat treatments.
268 D. Ferreño et al. / Engineering Failure Analysis 18 (2011) 256–270

1200
Internal node (disc)

1000 External node (disc)

Temperature (ºC)
External node (blade)
800 TTT curve

600

400

200

0
0 5 10 15 20 25 30
Time (h)

Fig. 18. Comparison of temperature profiles in different parts of the wheel, as obtained from the FE model.

profiles of the points belonging to the disc with the one in the blade of the wheel. In fact, the curves corresponding to the
internal and the external nodes in the disc are almost superimposed. This is due to the large thermal inertia of the disc
(which represents the majority of the mass of the part). In contrast, the node taken from the boundary of one of the blades
shows a temperature profile close to the applied heat treatment (see Fig. 16).
The TTT (time–temperature-transformation) curve (also called isothermal transformation diagram or ITT) of the CA-6NM
alloy is also represented in the figure. Although this kind of plots are limited to isothermal transformations (meaning that the
curves are determined by very quickly reducing the temperature from the region of stable austenite to a lower value and
then holding the new temperature constant until the transformation has completed), they can provide estimated informa-
tion in cases where a fast cooling took place, as in this case. Assuming the validity of the TTT curve, it is evident that in all
cases, the martensitic transformation is complete during the quenching following the austenising. This result is in complete
agreement with the metallographic examination described in Section 4.3.
The second stage of the heat treatment, the tempering, also plays a role in the final microstructure of the material. As
stated above, several authors [1,7–10] have pointed out the possibility of a material tempering embrittlement depending
on the parameters chosen for this stage of the treatment. The good combination of strength and toughness of CA-6NM steels
is a consequence of the highly dispersed austenite that forms as a result of segregation when these alloys are tempered above
500 °C [1,7]. The amount of austenite depends on the temperature and time of the tempering process (and also on some
alloying contents). With increasing tempering temperature, tensile and yield strength gradually decreased, exhibiting a min-
imum at around 627 °C, and increased again at slightly higher temperatures. Also, fracture appearance transition tempera-
ture (FATT) reaches a minimum after tempering around 627 °C; this minimum is associated with the complete softening of
the matrix and reverted austenite formed during tempering. Moreover, temper embrittlement increases the fraction of inter-
granular fracture after tempering at lower temperatures. The embrittlement is found to be related to the intergranular frac-
ture and precipitation of carbides along the prior austenite grain boundary. The precipitation of intergranular carbides
results from the decrease in carbon solubility in the reverted austenite during tempering. Susceptibility to temper embrit-
tlement in this steel is suppressed as tempering temperatures are raised. For these reasons, the alloy is commonly tempered
in the range of 593–621 °C. According to the information in Fig. 18, this requirement was fulfilled in all the representative
nodes of the part. Therefore, no temper embrittlement is expected for this material after receiving the heat treatment shown
in Fig. 16. This result is in complete agreement with the fracture examination, where no intergranularity was detected.

5.2. Estimation of thermal stresses

As previously pointed out, a structural failure is a consequence of an existing stress state in the presence of cracks or de-
fects in the component. In Sections 4.1 and 4.3–4.4, the existence of a large number of discontinuities, cracks and notches,
randomly dispersed in the material was proved. Nevertheless, as only a limited number of small coupons was provided by
the owner of the wheel, it was not possible to completely examine the fracture surface in detail, in order to define the se-
quence of phenomena leading to the failure. For this reason, the extent of this analysis is restricted to the available
information.
The tensile stress must be of a magnitude high enough to provide microscopic plastic deformation at the tip of the stress
concentration. The tensile stress need not be an applied stress on the structure, but may be a residual stress inside the struc-
ture, i.e., from welding or uneven cooling. In order to justify the fracture of the Pelton wheel analysed in this paper, the ther-
mal stresses during the two stages of heat treatment were obtained. Taking into consideration the orientation of the fracture
surface (see Fig. 3) only the circumferential stresses were determined. Fig. 19 shows the two radial lines (numbered as 1 and
2, respectively) where the thermal stresses were calculated through FE. The results are represented in Figs. 20 and 21,
respectively. In both cases, only six representative profiles of (circumferential) stresses were represented: three of them
correspond to the first cooling down of the heat treatment (time 50, 100 and 250 s) whereas the other three belong to
the second cooling down (time 72,100, 72,200 and 72,500 s).
 D. Ferreño et al. / Engineering Failure Analysis 18 (2011) 256–270 269

1
2

Y
X Z

Fig. 19. Sketch of the lines 1 and 2 where the circumferential stresses were calculated.

600

500
t = 100 s
Stress (MPa)

400
t = 250 s
t = 50 s
300 t = 72100 s

t = 72200 s
200
t = 72500 s
100

0
0 0.05 0.1 0.15 0.2 0.25 0.3
Position in the thickness (m)

Fig. 20. Circumferential stresses along line 1.

300

200

100
Stress (MPa)

t = 50 s
0
0 0.05 0.1 0.15 0.2 0.25 0.3
-100
t = 72500 s
-200 t = 72100 s
t = 72200 s
t = 100 s
-300
t = 250 s
-400
Position in the thickness (m)

Fig. 21. Circumferential stresses along line 2.

According to Fig. 20, it is evident that high stresses are present along line 1 during each of the cooling downs; this con-
clusion can obviously be extended to the rest of the contour of the part, as it is a consequence of the uneven cooling (the
boundary of the wheel is cooling faster than its interior). During the first cooling down, tensile stresses of 400–500 MPa
are reached whereas in the second, the stress state is up to 200–300 MPa. This tensile stress state would be of no importance
if no defects were present in the part, as the stresses are relatively low in comparison with the yield stress (see Table 4);
nevertheless, in the presence of defects, they can provide enough crack driving force to promote the fracture of the compo-
nent. In contrast, in the case of applying the heat treatment to a correctly manufactured wheel, without internal defects
(shrinkage cavities or hot tears), as those detected in the Charpy specimens tested in this study, no failure would occur,
as the stresses are well below the yield stress.
The results in Fig. 21 show that the internal region of the wheel is subjected to a compressive stress state during both
cooling down operations (this point is evident as the thermal stresses are secondary stresses which must be self balanced
270 D. Ferreño et al. / Engineering Failure Analysis 18 (2011) 256–270

in any section of the part). Nonetheless, in the case of a crack propagating fast in the component, the stress state must be
variable with time.

6. Summary and conclusions

A large experimental and analytical research was performed to assess the failure of a CA-6NM steel Pelton turbine imme-
diately after receiving a quenching and tempering heat treatment. The experimental part included:

 Optical and SEM fracture surface inspection.

 Chemical composition determination through OES.
 Optical and TEM microstructural analysis in different regions around the fracture surface.
 Characterisation of mechanical properties by means of micro hardness Vickers and Charpy impact tests.

These experimental activities allowed the following conclusions to be drawn:

 The existence of defects in the part was proved, including shrinkage cavities or hot tears detected in the fracture surface
(obtained through SEM inspection) and several sharp cracks following the length of the elongated ferrite islands (optical
microscopy).
 The study of the fracture surfaces of the Charpy tested specimens showed a morphology consisting of strong discontinu-
ities and deep notches; in agreement with the results described in the above point.
 Concerning the chemical composition of the material, slight discrepancies were detected between the experimental
results and the requirement of the specification. The experimental and theoretical information available shows that these
small discrepancies are of no importance concerning the fracture and mechanical material response.
 None of the typical features associated with a temper embrittlement was found in the TEM analysis. In particular, no evi-
dences of intergranular fracture were detected.
 None of the analyses performed (microstructure, chemistry, hardness Vickers, Charpy impact tests) showed any signifi-
cant difference between the material extracted from different regions of the Pelton wheel. As a consequence, the hypoth-
esis of material heterogeneities must be ruled out.

Moreover, a thermo-mechanical FE model was developed to analyse the homogeneity of the heat treatment and to obtain
the thermal stresses during the application of the heat treatment; the following conclusions were obtained:

 By comparison between the evolution of temperature in representative points of the wheel and the material TTT curve, it
is evident that in all cases, the martensitic transformation is complete during the quenching following the austenising.
 The efficiency of the heat treatment in any region of the volume was proved. As a consequence, no temper embrittlement
is expected to take place.
 The tensile stress state in the part would be of no importance if no defects were present, as the stresses are relatively low
in comparison with the yield stress.

The above evidence suggests that an inadequate manufacturing process of the Pelton wheel, responsible of the presence
of defects and discontinuities in the part, is the main cause of failure. None of the results indicated that the quenching and
tempering heat treatment had any influence on the fracture. In particular, no proofs of temper embrittlement were detected.

References

[1] Iwabuchi Y. Temper embrittlement of type l3Cr–4Ni cast steel. Trans Iron Steel InstJpn 1987;27(3):211–7.
[2] Tokuda A, Kumada Y, Nakagawa K. Development and results of a trail production of 13% Cr base cast stainless steel propeller for ships. JSW Tech Rev
1970;27:3142–52.
[3] Iwabuchi Y, Sawada S. Metallurgical characteristics of a large hydraulic runner casting of type 13Cr–Ni stainless steel. Stainless steel casting, ASTM STP
756; 1982. p. 332–54.
[4] Larson JA, Fisher R. The effect of heat treatment and melt practice on the impact properties of CA-6NM steel. AFS Trans 1979;63:113–26.
[5] Gisel W, Gerber E, Trautweisen A. CA6NM: new developments based on 20 years’ experience. Stainless steel casting, ASTM STP 756; 1982. p. 403–35.
[6] Steel castings handbook supplement 8: high alloy data sheets. corrosion series, steel founders’ society of America; 2004.
[7] Iwabuchi Y. Factors affecting on mechanical properties of soft martensitic stainless steel castings. JSME Int J Ser A. Special Issue on Recent Adv Mater
Process 2003;46(3):441–6.
[8] Leymonie C, Ottmann MC, Risache R. Structural transformation and mechanical properties of 13% Cr–4% Ni steels. Power Ind Res; 1982. p. 17–33.
[9] Iwabuchi Y. Toughness deterioration of 13Cr–3.8Ni cast steel in the process of tempering. J Iron Steel Inst Jpn 1984;70(7):701–8.
[10] Iwabuchi Y. Effect of tempering condition on toughness degradation in 13Cr-3.8Ni cast steel. J Iron Steel Inst Jpn 1984;70(7):1437–44.
[11] ANSYS user’s manual. Release 11; 2008.
[12] ASM handbook. Fundamentals of modeling for metals processing, vol. 22A. Materials park, Ohio: ASM international; 2009.