A Project on Wear and its Relation to Ball
Valves
A project for MANE 6960
By Robert Sayre
Submitted 12/9/2013
Introduction
Ball valves have many possible applications; they typically have low pressure drops and a high
reliability.The tend to be more reliable when sitting idle, compared to other valves such as globe or gate
valves[1]. When ball valves are used in situations with high cycling, the valve experiences wear. This
wear of the seats due to the ball acting on it, continually wears away at the seats until leakage is
deemed unacceptable or the seat and ball tolerances are below spec. For this project 2-way, 2-posiiton
ball valves will be examined, these types of ball vales rotate 90 degrees from full open to full shut. For
simplicity a 1-inch diameter ball will be used. A cutaway of an end-loaded ball valve is shown in Figure 1
below. From left to right there is the port, a seat, a (red) sealing o-ring, the ball, the other seat, and the
opposite port. The end is also the tailpiece. Tightening this tailpiece in the body creates an interference
fit between the body, the seats and the ball. This compressing of the seats, creates the sealing force on
the ball, and ideally creates a leaktight seal.
Figure 1: Ball Valve Cutaway [2]
The seal force simply reduces to a ring on a sphere, since a ball valve is symmetric, the valve can be
modeled using symmetry and even axial symmetry to decrease calculation times. When a ball valve is
brand new, the seal force is the highest, a minimum circumferential seal force in lbf/in or N/m must be
maintained or the fluid pressure will push past the seal. On the other end of the spectrum if the seal
force is too high, the valve will be harder to operate, and assemble, and more importantly the wear rate
of the seats and possibly the coating on the ball will wear away rapidly. Archard’s classic wear equation,
of the form K=(Vw*L)/(Fn/H), where K= wear rate, Vw=worn volume, L=sliding length, Fn=normal force, H=hardness
of the softer material[3]. The Archard equation is derived on the basis of single asperities sliding past each other.
Looking at this equation one can estimate increasing the hardness of the softer material, or decreasing the normal
force should decrease the wear. Therefore increasing the normal force on the seats beyond what is required for
sealing can negatively affect wear.
In order to define the rest of the problem, one needs to identify several more parameters, through empirical
testing, it is assumed is takes the user 1 second to rotate the valve from open to shut or vice versa. The
motion is assumed to be constant velocity for the sake of modeling. The valve is assumed to turn
exactly 90 degrees. The friction between the seats and the ball also directly affects the torque required
to turn the ball valve. Nominal dimensions of the ball valve are used in the model, only the ball and
seats are modeled. Finishes within Ra < 0.5µm surface finish are commercially viable. The ball is
assumed to be made of CRES 316L and the seats are made of PEEK plastic. These are materials used in
commercial ball valves in fluid systems. These are assumed to be low pressure ball valves. The material
properties for PEEK are shown in Table 1.
Table 1: PEEK material Properties[4],[5]
Property
Specific Gravity
Value
Units
1310
Hardness, Rockwell R
126
Hardness, Shore D
100
Tensile Strength
Elongation at Break
Elastic Modulus
Flexural Yield Strength
Flexural Modulus
Compressive Strength
Compressive Modulus
Shear Strength
Izod Impact, Notched
Coefficient of Friction
K (wear) Factor
Limiting Pressure Velocity
Thermal Conductivity
Melting Point
Glass transition
temperature
Poisson ratio
Kg/m3
110
40%
4.34
172
4.14
138
3.45
55.2
0.320
0.32
755 x 10-8
0.298
0.252
MPa
%
4.34 GPa
172 MPa
4.14 MPa
138 MPa
3.45 GPa
55.2Mpa
J/cm
Unitless
mm3/N-M
MPa-m/sec
W/m-K
340
144
ºC
ºC
.39
unitless
PEEK can have different materials (such as carbon or glass to increase the strength and stiffness,
or PTFE) added for additional strength and/or lubrication property enhancement. A variation of PEEK,
PEAK can even be used in medical implants[6]. For simplicity material properties for Quadrant EPP
Keytron 1000 PEEK unfilled, extruded was used, with the gaps filled in with a general PEEK material data
sheet filling in any needed property unknowns. The CRES ball properties are shown below in Table 2.
Table 2: Ball Properties (CRES 316)[7]
Property
Hardness, Brinell
Hardness, Rockwell B
Hardness, Vickers
Tensile Strength, Ultimate
Tensile Strength, Yield
Elongation at Break
Modulus of Elasticity
Izod Impact
Charpy Impact
Specific Heat Capacity
Thermal Conductivity
Value
Units
146
79
152
560
235
55%
193
150
103
0.500
MPa
MPa
%
GPa
J
J
J/g-ºC
14.0 - 15.9
W/m-K
In a case study of excessive ball valve wear by Exxon, they investigated increasing the hardness of the
materials and increasing the seal force from 1,000 lbf/in2 to 2,000 lbf/in2[8]. In a ball valve, a minimum
seal force is required to prevent leakage, however excessive circumferential seal force can cause
accelerated wear and increase the torque requirements. This reference translates the spherical
coordinates of the ball into a wear equation to give an equation defining the wear in terms of time as:
t=(H*d)/(10*K*D*Ns), whereT=time, H=hardness, d= wear rate, N= cycles per hour (constant), s = stress
(lb/in^2), D = length in, K= wear rate. In this case study they wanted to compare increasing the effects
of the hardness of the seats and the seal force to analytically calculate the life of the valve. They
calculated the life of the valve would indeed increase to increased seat hardness, and the valve
exceeded their expected life calculation.
A study of the PEEK rubbing on CF-30 steel[9], which is relevant to the problem at hand, showed Contact
stress showed little effect on the friction coefficient. They also discovered increasing the temp to 150C
leads to a higher specific wear for all materials. Counterface roughness and sliding velocity play a role in
the wear. They’re experiment mentions Hanchi and Eiss (1997) studied the friction and wear of PEEKCF30 at elevated temperatures. The sliding experiments were carried out on a pin-on-disc tribometer at
selected temperatures in the range of 200 -225°C.As temp increased from below to above, the wear
performance decreased, and friction increased. For these experiments, the PEEK material used had a
glass transition temperature of 143ºC. As one can see in Figure 2 and Figure 3, the results of the
experiment in reference[9], clearly show and accelerated wear rate with increasing PV (product of
apparent contact pressure and sliding velocity), and increased wear rate as the material is heated,
especially when it nears the glass transition temperature. Therefore one can conclude at elevated
temperatures, the thermal effect should be considered.
-
Figure 2: PEEK pin on disk test
-
Figure 3: Weight loss in pin on disk experiment
Theory and Methodology
The configurations of a ball valve are a ring on a seat, this type of model can be reduced to aaxisymmetric model for the simplicity of modeling. Figure 4and Table 3show the cross section of the
model for axi-symmetric modeling. This model simplifies the shape of the seat to a ball since the ball to
seat interface is just a ball on a ball. The 3d representation of this is shown in Figure 7.
Figure 4: Model Dimensions
Table 3: Model Dimensions
Variable
r1
r2
a
B
Value
0.0127
.00079375
0.0210414
0.0210414
Units
M
M
M
M
Since this is an end loaded ball valve the model constrains the seat to move in the x-direction only, for
the purposes of simplicity and the lack of easily available ball valve schematics, the seat is approximated
as a ball. These material properties listed above, were input into COMSOL 4.3. In the model, the seat
was given plastic properties, while the steel ball was not, as the steel ball should not exceed yield. This
will be checked in the model. Since in an axi-symmetric model, the ball cannot rotate within the seat, a
2D model was built as well, using a flat rectangular plate, longer than the circumference of the ball, and
a half ball for the seat.Using a preload of 100 lbf/in minimum (444.82N/m). Knowing the valve turns 90
degrees in 1 second, the sliding distance L, the velocity of the seat V, and model thickness can be
calculated. The model is run as plane strain. The plastic properties were calculated using the data in
Table 1. The strain x-coordinate was calculated at a 2% yield offset for simplicity, and a tangent modulus
was calculated using the ultimate break strength. Since this was the only non-commerically free for use
data available, it was used in lieu of conditional data. The plot of these material properties are shown
in Figure 5. A summary of the model inputs are shown in Table 4. A figure of the starting geometry is
shown in Figure 6.
Table 4: 2D Model Inputs
Variable
Rectangle width, w
Rectangle height, h
Seat radius
Seat linear travel
2D Model thickness
Acceptable material loss
PEEK Elastic Modulus
PEEK Tangent modulus
PEEK Yield Strength
Friction Coefficient
Seat Velocity
Value
0.045
0.001
0. 00079375
0.019949
0.154
0.0127
4.34
52
100
0.32
0.0199
Units
M
M
M
M
M
M
GPa
MPa
MPa
M/sec
PEEK Properties
1.40E+08
1.20E+08
Stress (Pa)
1.00E+08
8.00E+07
6.00E+07
4.00E+07
2.00E+07
0.00E+00
0.000
0.100
0.200
0.300
Strain
Figure 5: PEEK Properties
0.400
0.500
Figure 6: 2D Model Layout
Figure 7: Axi-symmetric 3D Representation
A quick calculation of the heat dissipated to get the temperature ein the specimen is using Q=M*Cp*T,
where M=mass, Cp=specific heat, T if the temperature change, and Q is the heat. If we input the kinetic
energy of the sliding ring as all the energy gets turned into heat. The kinetic energy is 7.721x10-8J, then
the temperature increase is about T=0.01C. Therefore the heat effects should be negligible, and can be
disregarded.
The potential modes of wear are adhesion and transfer, corrosion film wear, cutting, plastic
deformation, surface fracture, surface reactions, tearing, melting, electromechanical, fatigue. The most
likely candidates for wear in this situation is adhesion, cutting, and tearing.
Analytically looking at the wear rate of the seat, using a wear rate found in the material specs from
[MATWEB], of K=755x10-8 mm3/N-M, a circumferential seal force of 100 lbf/in on the seat parameters in
Table 4, yields a normal force of 6.69N, and a sliding distance of 0.03989m per cycle (1 cycle is back and
forth). This yields a lost volume of material for 0.050-inch (0.00127m) of linear material lost as
5.192*10-8m3 of material lost before sealing is lost. Solving for the sliding length yields L=102.787m,
which translates into 2576 cycles before sealing is lost in the valve, and the seats will need replacement.
This assumes constant hardness, and constant normal force. In reality the normal force of the valve seat
will decrease with wear, since the force is created due to the compression of the seat between the ball
and the body/tailpiece. The model was run in multiple time steps to attempt to capture the velocity of
the seat.
Results and Discussion
Results are to include the results of analytical methods with their assumptions and comparisons to
model executed in Abaqus and/or COMSOL. The wear rates will be correlated where possible to test
data to form some sort of a basis. In the models the von mises stress in the steel plate did not exceed
the yield strength of the material (193 GPa), so the assumption of no plastic deformations in the steel
plate was correct. One optimistic goal was to capture particle wear, and use advanced FEA techniques,
such as adaptive remeshing, however computational difficulties (computer was not fast enough), made
this goal unobtainable for the timeframe. In order to work around the adaptive remeshing issue, several
geometries were made with a worn amount removed from the
The time stepping proved somewhat challenging, multiple steps must be taken with contact pairs, first a
small gap must be between the seat and ball, then they have to be pushed together slightly, then the
preload apply (calculated from the ball and seat dimensions, and circumferential seal load) applied.
Finally a velocity is applied for a time step of 1 to get the total motion. For the sake of simplicity,
The seat model with no wear exhibits a max von mises stresss 0f 6.1618 MPa. The model is shown in
Figure 8. An intermediate step of ~20% wear is shown in Figure 9 with a von mises stress of 5.097 MPa.
The seat was modeled down to 50 thousands of an inch of wear. This is typically considered acceptable
wear, but as shown in Figure 10, there may be leakage at this point depending, the max von mises stress
is 5.3048x105 Pa.
A summary of the results comparing the wear, which is the horizontal amount of material removed from
the seat, akin to sanding a cylinder, or what you would measure in the shop with a micrometer is shown
in Figure 11.
Figure 8: No Wear
Figure 9: Intermediate wear
Figure 10: Full Wear
7.00E+06
Von mises stress vs wear
Von Mises Stress (Pa)
6.00E+06
5.00E+06
4.00E+06
3.00E+06
2.00E+06
1.00E+06
0.00E+00
0
0.0005
0.001
Wear (m)
Figure 11: Von Mises Stress vs Wear
0.0015
Works Cited
[1]R.W. Zappe Peter Smith, Valve selection handbook: engineering fundamentals for selecting the right
valve design for every industrial flow application.
[2] http://flotite.com/ball-valve-experts/wp-content/uploads/2012/04/metal-seated-ball-valvecutout.png.
[3] J. F. Archard, "Wear Theory and Mechanicsms," New York,.
[4] Quadrant EPP Ketron® 1000 PEEK Polyetheretherketone, unfilled, extruded. [Online].
http://matweb.com/search/DataSheet.aspx?MatGUID=53b9159c018544a599a06726922c9b8e
[5] Overview of materials for Polyetheretherketone, Unreinforced. [Online].
http://matweb.com/search/DataSheet.aspx?MatGUID=2164cacabcde4391a596640d553b2ebe&ckck=1
[6] Steven M. Kurtz, PEEK Biomaterials Handbook.: William Andrew, 2012.
[7] Material Properties for AISI Type 316L Stainless Steel, annealed plate. [Online].
http://matweb.com/search/DataSheet.aspx?MatGUID=530144e2752b47709a58ca8fe0849969
[8] T. Sofronas, "Case 11: Excessive ball valve seat wear," Hydrocarbon Processing, 2002.
[9] "Thermo-mechanical model to predict the tribological behaviour of the composite PEEK-CF30/steel pair
in dry sliding using multiple regression analysis".