Professional Documents
Culture Documents
ARTICULATING ROD
submitted by
Abstract
Reliability analysis has been done on the articulating rod to ensure that it does not fail while in
operation. The articulating rod is initially designed with three independent random variables with
three different distributions. The design of the rod is then modified based on the reliability of the
rod against yielding. Two different methods to compute reliability is used. Two point function
approximation is used to come up with a function for the von misses stress acting on the rod
using which a non-linear problem specific limit state function is obtained. The rod is then
optimized to achieve a maximum practical reliability.
1. INTRODUCTION
Articulating rod is an important member (link) in a slider crank mechanism. It is the link
which converts the rotating motion of the crank to a corresponding reciprocating motion
of the piston, or vice versa. The rod is subjected to high compressive stresses during its
operation. Hence there is a necessity to design the rod to withstand these stresses based
on various criteria to ensure that the yield stress generated within the bar does not exceed
the material strength. It has to be noted that there are many other failure modes of the rod
but this is the most significant mode of failure. While most of the rods are made of
Aluminium, there are various factors affecting the stress generated within the rod. Three
factors are considered in this project (for simplicity) and in precise, they are, the diameter
of piston end, the load acting on the rod and the rod length. The rod is modeled and
simulated using FEM software to compute the stress generated for different combinations
of the variables and the one which has a greater influence on stress is obtained. A non-
linear limit state function is obtained using two point function approximation which
approximates the stress generated within the rod. The reliability is then calculated using
Hasofer-Lind Rackwitz and Fiessler (HLRF) method and Monte Carlo Simulation and
checked for agreement. The rod is then optimized to achieve a greater reliability by
modifying the defined random variables. While optimizing, it is ensured that the science
and practicality of the problem is not sacrificed for reliability.
2. TECHNICAL APPROACH
The technical approach to determine the optimized reliability is discussed in this section.
Three random variables are considered in this project and they are:
a. Diameter of the piston end of the rod (d)
b. Length of the rod (L)
c. Load acting on the rod (P)
The distributions of these random variables are assumed based on their practical
behavior. The diameter of the piston end of the rod is log-normally distributed
with a mean of 14mm and having covariance of 1%. Log normal distribution
ensures that the random variables are distributed in the positive domain and
cannot be negative. Since the mean of the diameter is small and there is a higher
probability for the diameter to be negative, if normal distribution is considered,
the rod is therefore assumed to be log normally distributed. Secondly there is a
tolerance value for the length of the rod and therefore the length of the rod has to
be within certain limits. To ensure this constraint, the length of the rod is
uniformly distributed with a lower limit of 165 and a higher limit of 191. The
mean load acting on the rod is dependent on the angular velocity of the vehicle
(which has the radial engine). Since the vehicle can have different speed range,
the load is therefore widely distributed compared to other random variables.
Hence a normal distribution has been chosen for the load (P) with a mean of
34KN having a covariance of about 2%. The random variables are now
completely defined.
The rod is now modeled in Creo Parametric 2.0. While modeling, it is ensured
that the design of the rod can be easily modified by changing either one of the
random variables. Various geometric features such as sketch, extrude, pattern and
holes are used to model the articulating rod.
1 = 3
2 = + 3
To determine the value of the nonlinearity index r, the function values are matched at
the previous design point X1.
The previous point and current point in our case is:
The gradient, i.e., the derivative of the performance function with respect to the random
variables at the current point X2 has to be determined to solve for the non-linearity index r.
The gradient is computed using Finite Forward Difference method. Firstly, for the given set
of conditions in the current point X2, the articulating rod is modeled and then simulated to
obtain the function value at the current point. The same procedure is repeated for the
previous point X1 to obtain the function value at the previous point. To calculate the gradient,
the random variable which affects the stress is perturbed by 1% and the corresponding
change in the function value is noted down. The derivative is thus calculated using the Finite
forward difference expression given as:
( + ) ()
=
The function values at X1 and X2 and also the derivatives at the current points are listed in
Table 1.
Table 1
DESCRIPTION VALUE
g(X1) 340.542
g(X2) 384.018
-0.4158
3.14e-3
0.01066
On matching the function values at the previous design point, the value of the nonlinearity index
r is found to be 0.6737.
The final approximation for the stress generated () within the articulating rod is now given as:
To compute the reliability using the two methods mentioned above, a limit state function
representing the failure mode of the rod is required. Since, the project deals with the
computation of reliability against yielding, the nonlinear limit state function is the difference
between the strength of the rod and the stress generated within the rod. The rod is modeled
using Al 6061 (an alloy of Aluminium) and thus the Yield strength of the rod (Sy) is found
to be 370 N/mm2.
The nonlinear limit state function (Z) used to compute the reliability is thus given as:
Z = Sy
The HLRF method is used to compute reliability when the random variables are all not
normal. The major principle behind this is that the normal mean and standard deviation
are computed for the corresponding non normal random variable by matching the
associated PDF and CDF. Initially, the mean and standard deviation of all the random
variables is defined along with the distribution type. Based on the type of distribution,
a corresponding normal mean and standard deviation is obtained for each variable.
Using the normal means and standard deviations, the gradients are calculated at the
current point (X*) using finite forward difference method. Once the gradients are
calculated, the safety index is computed using the following expression:
( ) =1 ( ) ()( )
=
((=1 ( ) ()) (=1 ( ) ()))
Here the expressions for X* and u* are given as follows:
X* = x + xx
u* = x
( ) ()
=
((=1 ( ) ()) (=1 ( ) ()))
The value of the safety index is thus obtained and the process is repeated to obtain
convergence on . Once the safety index is converged, the standard normal CDF of the
safety index gives the reliability of the structure. It has to be noted that the starting
point (X) in HLRF method should not necessarily be the mean. It can be any point in
the domain space. Also, after the end of each iteration, the current point X* and u* are
updated.
The Reliability is determined to be 0.8468 with the given statistics using this method.
The graphs for the convergence history and the function value for this specific case is
shown below:
Figure 1
Figure 2
The point X at the final iteration is the most probable point of failure (MPPF) which
represents the point at which the structure would fail. It is the closest point to the
failure surface.
Thus, the reliability obtained using both the methods have agreement with each
other.
2.5 OPTIMIZATION
The reliability of the rod is found to be around 0.85 in the previous section. The rod is
now optimized to decrease its total weight and also increase its reliability at the same
time. Since the load P with a wider distribution signifies various operating conditions of
the vehicle (load is dependent on ), only the diameter of the piston end of the rod and
the length of the rod is optimized to achieve the required reliability with minimum
weight. Two major functions are required to achieve optimization. They are:
The constraint function ensures that the reliability has been computed for a physically
feasible structure. The two constraints in this problem are:
a. The diameter of the piston end should be less than 80% of the diameter of the
crankshaft end.
b. The reliability of the entire structure should be greater than 0.85 to ensure that there is
an increase in reliability.
This is an iterative process and the result is an optimized value of d and L which gives a
minimum weight of the structure and also increases the reliability at the same time. After
optimization, the diameter of the piston end of the rod is found to be 17mm and the
length of the rod is 150mm with a weight of 0.149kg. Reliability of 0.8847 is achieved
after optimization.
Three random variables are considered to compute the reliability of the articulating rod.
They are:
a. Diameter of the piston end of the rod = d ~ LN (2.638,9.99e-3)
b. Length of the rod = L ~ U(165,191)
c. Load acting on the rod = P ~ N(34000,2%)
An approximation for the stress generated within the rod is derived using Two Point
Approximation and the performance function (Z) for the failure of the rod against
yielding is developed which is the difference of the Strength of the material (Sy) and the
stress generated () within the rod. Using the developed performance function, the
reliability of the rod is computed using two methods:
Table 2
d L P d L P
Figure 3
From table 2 it is clear that load has most influence on reliability of articulating rod.
Change in standard deviation of load from 680 N to 340 N (2% COV to 1% COV) will
increase reliability by about 13.5%. Also it is very clear from table 2 that reliability was
not very sensitive to change in statics of other two variables. For more understanding of
relation between reliability and statics of load two graph as shown in figure 3 and figure 4
were plotted. Graph in figure 3 is for mean load vs reliability for constant COV of 1%
while graph in figure 4 is for standard deviation of load vs reliability for constant mean
load of 34 KN
Figure 4
The rod is then optimized to achieve minimum weight and higher reliability. After
optimization, the reliability of the structure is found to be 0.8619 with a weight of 0.155
kg. Results from optimization are summarized in table 3
Table 3
a. More random variables can be included to account for the randomness of the
structure. Moreover, random variables which are not independent and are correlated
with each other can be included.
b. The approximation for the stress generated within the rod can be derived using higher
order approximations to reduce the error between the exact and approximate function
values.
c. Optimization is done using the approximate function and the optimized result is exact
only within the design points, else the optimized result is an approximate because the
performance function is exact only within the domain of the design points. This
problem can be overcome by developing a new approximation for the stress
generated, for each iteration, using the optimized values of the random variables.
REFERENCES:
1. Achintya Haldar, Sankaran Mahadevan Probability, Reliability and Statistical Methods in
Engineering Design
2. Two Point Functions Approximation, UTD-Blackboard
3. Hasofer-Lind Rackwitz and Fiessler (HLRF) method to determine Reliability, UTD-
Blackboard
4. Reliability Based Design Optimization Matlab Torsion Member Example, UTD -
Blackboard