Discussion Forum

Hi, I am also very interested in EHD simulation. But I cannot understand your model. Are you trying to build the boundary condition equation.

Hi, I just make a EHD using weak form boundary PDE to represent Reynolds Equations. In this model Dirichlet boundary conditions are adopted. A global equation is to guarantee the load balance. Variables are dimensionless. But it does not converge. Have you any experience in this field?

Dear all,

I have built a 1D EHD in COMSOL. But it does not work. There always exists an error. Can anybody give me any suggestion? Please help me! My *.mph file is in the attachment. The error is as follows

The following feature has encountered a problem:
Failed to find a solution.
Singular matrix.
There are 249511 void equations (empty rows in matrix) for the variable comp1.P.
at coordinates: (2.5949,0), (2.5948,0), (2.59482,0), (2.59484,0), (2.59486,0), ...
There are 1 degrees of freedom that do not occur in any equation (empty columns in matrix) for the variable comp1.H0.
at coordinates: (0,0), ...
Returned solution is not converged.
- Feature: Stationary Solver 1 (sol1/s1)

Thanks a lot!

hi, I also met this problem. Have you figured it out?

Hi,

Have you read the papers

Computational Approaches for Modelling Elastohydrodynamic
Lubrication Using Multiphysics Software from Tan, Goodyear, Jimack
www.scs.leeds.ac.uk/pkj/Papers/Journals/TGJTW11.pdf

and

Engineering Software Solution for Thermal Elastohydrodynamic
Lubrication Using Multiphysics Software from
Lohner, Ziegltrum, Stemplinger, Stahl
www.hindawi.com/journals/at/2016/6507203/

Both are very useful (and open acess) to and provide good examples how to approach the
modelling of EHD.

Hi,

Yes, I have read both of them, the problem is when I try to use a global equation to couple the problem, COMSOL said "one degree of freedom does not occur", any idea about how to figure this out?

Thanks.

Ok after looking at the attached model there are several mistakes in the implementation.

1. partial derivatives

You use the weak form on boundary and need the derivative dp/dx.
If you use d(P, x) , the result is d(P,x) = 0 . So its obviously wrong.
read COMSOL Reference Manual page 233 regarding
"Differentiation operators: D, PD, and DTANG"

You can either use dtang(P, x) or PTx.

1.2 Use x instead of X

for instance
- dtang(P, X) ( won't work)
- dtang(P,x) -> this work

the dimensionless coordinate X within the formula (Uppercase) is
expressed within COMSOL by a x (Lowercase).

2. Film Thickness H

your code : H = H0 + 0.5*X^2, with X = x/b
better : H = H0 + 0.5*x^2

do not use X ! all your equations are already dimensionless.
Use only x in all expressions.

3. Pressure p

for eta and rho you use p (dimensionally valued). so you have first to define it
with dimensionless pressure P and the maximum hertzian pressure pH
p = P*pH

4. Initial value for H0
You should set the initial value H0 to 1.0 . It does not converge for 0.1

5. Initial value for P
it should be >0 ! so set it to 1

6. Solver settings
at Fully Coupled
set max number of iterations to 100 or 200. 25 is too low.

7. To prescribe P=0 at the boundary points I would suggest
using the Diriclet Boundary Condition only. I think Identity pair + DBC is
too complicated and not necessary.

8. naming convention ( suggestion)
Use Lowercase for dimensionally valued expression
and Uppercase for dimensionless expression
e.g. for density
- rho , dimensionally valued expression
- RHO, dimensionless expression

weak form. You can use one of these two formulations:
1. dtang(test(P),x)*qq
2. test(PTx)*qq

for the variable qq I would define:

RHO*H^3/ETA*dtang(P,x)-A1*RHO*H
or
RHO*H^3/ETA*PTx-A1*RHO*H

I've attached two models
- first is with your naming convention: 1D_EHD_5_4_fixed.mph
- second is with my suggested naming convention: 1D_EHD_5_4_v1.mph
Also the second modell only converged if I use a constant viscosity eta = eta0

Best Regards