Lennard-Jones cluster : BARON fails with EXECERROR, division by 0

Hi,

I am trying to solve the Lennard-Jones cluster problem with 5 atoms on the NEOS solver. Here is the GAMS model:

Scalar e /0/;
Positive variables x2,x3,y3,x4,y4,z4;
Variables x5,y5,z5;
Variable obj;
x2.up=1.2;
x3.up=1.2;
y3.up=1.2;
x4.up=1.2;
y4.up=1.2;
z4.up=1.2;
x5.lo=-1.2;
x5.up=1.2;
y5.lo=-1.2;
y5.up=1.2;
z5.lo=-1.2;
z5.up=1.2;

equations objdef;
objdef… obj =e=
(4sqr(1/(x2**6+e)-1/2)-1)+(4sqr(1/((sqr(x3)+sqr(y3)+e)**3)-1/2)-1)+(4sqr(1/((sqr(x4)+sqr(y4)+sqr(z4)+e)**3)-1/2)-1)+(4sqr(1/((sqr(x5)+sqr(y5)+sqr(z5)+e)**3)-1/2)-1)
+(4sqr(1/((sqr(x3-x2)+sqr(y3)+e)**3)-1/2)-1)+(4sqr(1/((sqr(x4-x2)+sqr(y4)+sqr(z4)+e)**3)-1/2)-1)+(4sqr(1/((sqr(x5-x2)+sqr(y5)+sqr(z5)+e)**3)-1/2)-1)
+(4
sqr(1/((sqr(x4-x3)+sqr(y4-y3)+sqr(z4)+e)**3)-1/2)-1)+(4sqr(1/((sqr(x5-x3)+sqr(y5-y3)+sqr(z5)+e)**3)-1/2)-1)
+(4
sqr(1/((sqr(x5-x4)+sqr(y5-y4)+sqr(z5-z4)+e)**3)-1/2)-1)
Model m / all /;
Solve m minimizing obj using nlp;
Display obj.l;

and options

optca 1e-3
optcr 0
MaxTime 60


BARON immediately fails with this message:

**** Exec Error at line 32: division by zero (0)

However, when the variable e is given a small value (e.g. 1e-9), BARON reaches convergence.
I have tried giving the variables an initial value, without success.
Could you help me out?
Thanks,

Charlie


To unsubscribe from this group and stop receiving emails from it, send an email to gamsworld+unsubscribe@googlegroups.com.
To post to this group, send email to gamsworld@googlegroups.com.
Visit this group at http://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/d/optout.

Hi Charly

You could set lower bounds like x2.lo = 1E-9;

Cheers

Renger



From: gamsworld@googlegroups.com [mailto:gamsworld@googlegroups.com] On Behalf Of cvanaret@gmail.com
Sent: Freitag, 17. Juli 2015 14:46
To: gamsworld@googlegroups.com
Subject: Lennard-Jones cluster : BARON fails with EXECERROR, division by 0



Hi,

I am trying to solve the Lennard-Jones cluster problem with 5 atoms on the NEOS solver. Here is the GAMS model:

Scalar e /0/;
Positive variables x2,x3,y3,x4,y4,z4;
Variables x5,y5,z5;
Variable obj;
x2.up=1.2;
x3.up=1.2;
y3.up=1.2;
x4.up=1.2;
y4.up=1.2;
z4.up=1.2;
x5.lo=-1.2;
x5.up=1.2;
y5.lo=-1.2;
y5.up=1.2;
z5.lo=-1.2;
z5.up=1.2;

equations objdef;
objdef… obj =e=
(4sqr(1/(x2**6+e)-1/2)-1)+(4sqr(1/((sqr(x3)+sqr(y3)+e)**3)-1/2)-1)+(4sqr(1/((sqr(x4)+sqr(y4)+sqr(z4)+e)**3)-1/2)-1)+(4sqr(1/((sqr(x5)+sqr(y5)+sqr(z5)+e)**3)-1/2)-1)
+(4sqr(1/((sqr(x3-x2)+sqr(y3)+e)**3)-1/2)-1)+(4sqr(1/((sqr(x4-x2)+sqr(y4)+sqr(z4)+e)**3)-1/2)-1)+(4sqr(1/((sqr(x5-x2)+sqr(y5)+sqr(z5)+e)**3)-1/2)-1)
+(4
sqr(1/((sqr(x4-x3)+sqr(y4-y3)+sqr(z4)+e)**3)-1/2)-1)+(4sqr(1/((sqr(x5-x3)+sqr(y5-y3)+sqr(z5)+e)**3)-1/2)-1)
+(4
sqr(1/((sqr(x5-x4)+sqr(y5-y4)+sqr(z5-z4)+e)**3)-1/2)-1)
Model m / all /;
Solve m minimizing obj using nlp;
Display obj.l;

and options

optca 1e-3
optcr 0
MaxTime 60


BARON immediately fails with this message:

**** Exec Error at line 32: division by zero (0)

However, when the variable e is given a small value (e.g. 1e-9), BARON reaches convergence.
I have tried giving the variables an initial value, without success.
Could you help me out?
Thanks,

Charlie


To unsubscribe from this group and stop receiving emails from it, send an email to gamsworld+unsubscribe@googlegroups.com.
To post to this group, send email to gamsworld@googlegroups.com.
Visit this group at http://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/d/optout.


To unsubscribe from this group and stop receiving emails from it, send an email to gamsworld+unsubscribe@googlegroups.com.
To post to this group, send email to gamsworld@googlegroups.com.
Visit this group at http://groups.google.com/group/gamsworld.
For more options, visit https://groups.google.com/d/optout.