DAY 23: CERAMICS
What properties of ceramics make them
attractive to mechanical designers?
 What properties of ceramics make them
challenging to use?
 Manufacturing Concepts
 Design Concepts

MANUFACTURING CERAMICS

The following methods are used to shape the
ceramics. Please not that (wetted) powder is key.
SINTERING

This is a process in which the small chunks of
powder loose their identity, as the whole (porous)
part is bonded. Temperature and often pressure
are needed. Shrinkage has to be understood.
DIE PRESSING (UNIAXIAL PRESSING)
Most common and rapid for small ceramic
components where speed of manufacture means
more than strength and uniformity.
 Pressure, and densification is variable through
the mold. The object will have varying
properties, and maybe differential shrinkage on
sintering.
 Simple shapes only.
 Hot pressing is a combination of sintering and
die-pressing happening at once.

ISOTACTIC PRESSING
Pressure transmitted to
the powder from a
compressed fluid.
 More uniformity, less
porosity
 An elastomer (rubber
mold) serves as the
interface.
 Slower rate of
production.
 Best for cylindrical
shapes, eg. Spark plug.

Hot isotact pressing (HIP)
combines sintering and isotactic
pressing.
EXTRUSION

We add a plasticizing agent, which is later cooked
away during sintering.
SLIP CASTING
Make a slurry by adding liquid to the powder.
 Pour into a porous mold.
 Fluid is absorbed by the mold leaving a drier
layer of powder along the walls.
 Pour off remaining slurry, slip. Opening the
mold reveals the thin-walled object.
 Ready to be sintered.

INJECTION MOLDING

This method holds the most promise for mass
production of complex shapes as evidenced by its
use in producing ceramic turbocharger rotors. A
combination of 60-70% powder mixed with an
organic binder to provide flow is injected into a
mold. Prior to sintering, burnout of the binder
must be done. Current restrictions include small
wall thickness. Because of the cost of equipment,
this is only cost-effective for large volumes, and
for simple shapes, the dry pressing methods are
more cost-effective.
REACTION BONDING
A solid powder and a gas or liquid react during
sintering to densify and bond.
 In Reaction Bonded Silicon Nitride, silicon
powder is fired in the presence of high pressure
nitrogen gas, and the reaction forms Si3N4.
 Advantage: very low shrinkage.
 Disadvantage: high porosity and lower strengths.

MORE REACTION BONDING

Reaction bonded silicon carbide, RBSC, is made
by infiltrating liquid silicon into a compact of
carbon and silicon carbide. The Si reacts with
the carbon to form SiC which then bonds with the
original SiC particles. Pores are filled with liquid
Si. Consequently, high temperature strength
falls off at silicon's melting temperature.
Dimensional changes with RBSC can be less than
1%. One interesting variation is to use carbon
fibers rather than carbon particles.
SUMMARY OF MATERIALS







Hot-pressed silicon nitride (HPSN) has the strongest specific
strength (strength/density) at 600oC of any material. It has
excellent thermal shock resistance.
Sintered silicon nitride (SSN) has high strength and can be
formed into complex shapes.
Reaction-bonded silicon nitride (RSBN) can be formed into
complex shapes with no firing shrinkage.
Hot-pressed silicon carbide (HPSC) is the strongest of the silicon
carbide family and maintains strength to very high temperatures
(1500oC).
Sintered silicon carbide (SSC) has high temperature capability
and can be formed into complex shapes
Reaction-bonded silicon carbide (RSBC) can be formed into
complex shapes and has high thermal conductivity.
Partially stabilized zirconia (PSZ) is a good insulator and has high
strength and toughness. It has thermal expansion close to iron,
facilitating shrink fit attachments.
DESIGNING WITH CERAMIC MATERIALS
Deterministic design, is where we predict a
definite “yes” or “no” as to failure based on
comparing strength data with loads through
calculations.
 Does not work for ceramics, since strength data is
so scattered.
 We use a process called probabilistic design. We
predict a probability of failure in the part given
its material, geometry and loading.

PROBABILITY OF FAILURE
We can assume that lots of testing was done on
a ductile material. Stress between 0 and 300
MPa
 We plot a histogram of the number of failures
which occur at each level of stress. Get

Bell Curve – Normal Distribution
WEIBULL DISTRIBUTION

Now try a brittle ceramic. We find that the
normal distribution or bell curve does not describe
the probability of failure. Which looks like
WEIBULL STATISTICS
The actual data shows a tilt towards failure early
on. Probability of early failures a lot more. This
is kind of expected with brittle materials.
 Weibull had the idea of describing this kind of
thing with a probability distribution function
that looked like:

f x  
0  x 



m   0 
m 1
e
 x
 
0




m
0 is a scale
parameter.
m is the Weibull
modulus
CUMULATIVE DISTRIBUTION FUNCTION -WEIBULL

Given the above, we can find the probability of
failure for all stresses below a given one. We do
this by integrating.
F   

 f  x  dx
0
 1 e
 
 
0




m
Example given in
Maple of how this
works.
SIZE IS IMPORTANT!
Suppose we have a chain of n links. For a single
link, let F() be the probability of failure at a
stress below and up to .
 Probability of survival at all stress up to  is
1- F()
 Then the probability of failure of the chain of n
links up to  is
Fn() = 1 – (1 – F() )n.
 Say we had a “chain” made out of n blocks of
ceramic characterized by m and 0.

SIZE EFFECTS
F 
  1  (1  e
  

 


 0 
m
)
n
We return to the Maple worksheet to see how the size
increase has changed our CDF.
It’s not pretty.
IT’S NOT THAT SIMPLE!
Here we have been dealing with a unit volume or
a chain of unit volumes in a state of uniaxial
tension.
 Practical designs will experience different
volumes with various distributions of multiaxial
stress.
 Computational methods are used. CARES/Life is
a program which takes finite element analysis of
the component and calculates the overall
probability of failure.

PROOF TESTING
Here the idea is to cut of the tail of the
distribution by testing all parts prior to delivery.
 This is expensive, but is sometimes done.
 Problem: damage is cumulative. Proof testing
can lead to damage that would shorten the life of
the part.

QUICK REVIEW
Design to a minimum safe value is not possible!
 We must design to an accepted probability of
failure.
 Weibull statistics is used instead of normal
statistics in fitting data and computing
probabilities.
 The Weibull modulus indicates the extent to
which the strength data is scattered. Want it as
high as possible.
 The size of the object being designed must be
accounted for. Designs that work for small
components won’t work for large.
