[Tingyun YIN]

Hi, everyone.

I want to define Signum function in CFX, which is used to export -1, 0 or 1. For example, sgn(C) is equal to 1 when C>0; sgn(C) is equal to 0 when C=0; sgn(C) is equal to -1 when C<0.

Thanks in advance, Any comments and suggestions are appreciated.

[AtoHM]

Depending on what values your are looking at, the if-Statement might be sufficient for this case. https://www.sharcnet.ca/Software/Ans...statement.html
Lets assume you look at a pressure field and do this kind of stuff, giving some threshold value, I doubt there would be a point in the field where the value will be exactly as the one you defined. Lets say you use 1 Pa. Double precision will give you maybe 1.0000008 Pa and 0.99999998 Pa as closest values, resulting in -1 or 1 eventually?

[Tingyun YIN]

Originally Posted by AtoHM

Depending on what values your are looking at, the if-Statement might be sufficient for this case. https://www.sharcnet.ca/Software/Ans...statement.html
Lets assume you look at a pressure field and do this kind of stuff, giving some threshold value, I doubt there would be a point in the field where the value will be exactly as the one you defined. Lets say you use 1 Pa. Double precision will give you maybe 1.0000008 Pa and 0.99999998 Pa as closest values, resulting in -1 or 1 eventually?

Thanks for your reply.
Exactly, this function is used to distinguish positive or negative or zero in CFX. I am modifying the cavitation model. Flow field pressure is over than saturated vapor pressure, then evaporation occurs. Flow field pressure is lower than saturated vapor pressure, then condensation occurs. Otherwise, returns to 0, no changes.

[AtoHM]

What I was trying to say is, that I dont think it is possible using the simple if-Statement in the link to distinguish into -1,0,1. However, I doubt there is an actual need for having "0", because the solutions are numerical and the chance that it hits exactly the value you set are very small. So I think you can go with if C<0 use -1 and if C=>0 use 1 which allows to use this kind of if-Statement.

if( cond_expr, true_expr, false_expr )

if( Pressure < Threshold, -1, 1 )

[Tingyun YIN]

Originally Posted by AtoHM

What I was trying to say is, that I dont think it is possible using the simple if-Statement in the link to distinguish into -1,0,1. However, I doubt there is an actual need for having "0", because the solutions are numerical and the chance that it hits exactly the value you set are very small. So I think you can go with if C<0 use -1 and if C=>0 use 1 which allows to use this kind of if-Statement.

if( cond_expr, true_expr, false_expr )

if( Pressure < Threshold, -1, 1 )

Yes, you are right. 0 is unable to be avoided, which is a problem. IF statement seems possible, but it still exits defect.

[Opaque]

Besides the IF statement alternative discussed, there is also the "step" function.

step(C) = 1 if C>0
step(C) = -1 if C<0
step(0) = 0.5

The argument C must be dimensionless.

Be careful when working with floating points numbers, it is not a good practice to compare to 0. Perhaps, you want to return 0 whenever C is within some tolerance (machine/precision dependent) around 0.

In addition, this type of function generates considerable numerical instabilities because of their abrupt/sudden changes around 0. It is sometimes better to create a smooth "blending" function to transition from one regime/range to another.

Hope the above helps

[Tingyun YIN]

Originally Posted by Opaque

Besides the IF statement alternative discussed, there is also the "step" function.

step(C) = 1 if C>0
step(C) = -1 if C<0
step(0) = 0.5

The argument C must be dimensionless.

Be careful when working with floating points numbers, it is not a good practice to compare to 0. Perhaps, you want to return 0 whenever C is within some tolerance (machine/precision dependent) around 0.

In addition, this type of function generates considerable numerical instabilities because of their abrupt/sudden changes around 0. It is sometimes better to create a smooth "blending" function to transition from one regime/range to another.

Hope the above helps

Thanks for your suggestions. For current situation, STEP looks better.