Learning Forum

[henri.dolfen]

I noticed that there is a significant discrepancy between the forces outputted via a Force Report File and when computing them using a UDF (Compute_Forces_And_Moment) on a moving rigid body using dynamic mesh. Besides the Compute_Forces_And_Moments, I also used a different method, which yielded the same results. It is a 2D case where a fixed motion was applied on a cylinder in a channel.

Code snippet from UDF:

DEFINE_ON_DEMAND(fixed)
{
#if !RP_HOST
real dt=CURRENT_TIMESTEP;
x_d=0.0002*2*M_PI*18*cos(2*M_PI*18*n*dt);
#endif
}

DEFINE_CG_MOTION(cylinder,dynamic_thread,velocity,omega,time,time_step)
{
#if !RP_HOST
    NV_S(velocity,=,0.0);
    NV_S(omega,=,0.0);
    velocity[0]=x_d; /*x-component*/
#endif
}

DEFINE_ON_DEMAND(force)
{
#if !RP_HOST
Domain *domain;
Thread *face_thread;
face_t face;
real dummy_cg[3],force[3],moment[3];
real traction[ND_ND];
real area[ND_ND];

domain=Get_Domain(1);
face_thread=Lookup_Thread(domain,cylinder_id);

Compute_Force_And_Moment(domain,face_thread,dummy_cg,force,moment,FALSE);

F=force[0]*depth; /*force in x-direction*/

F_V_x=0.0;
F_P_x=0.0;

begin_f_loop(face, face_thread) {
    if (PRINCIPAL_FACE_P(face, face_thread)){
            NV_VS(traction, =, F_STORAGE_R_N3V(face, face_thread, SV_WALL_SHEAR), *, -1.0 * depth);
            F_V_x += traction[0];
            F_AREA(area, face, face_thread);
            F_P_x += area[0] * F_P(face, face_thread);
    }
} end_f_loop(face, face_thread);

F_V_x = PRF_GRSUM1(F_V_x);
F_P_x = PRF_GRSUM1(F_P_x) * depth + F_V_x;

#endif
}

And the fixed motion was applied using the following journal file:

(define (initialize)
    (ti-menu-load-string "define user-defined execute-on-demand \"initialize::libudf\"\n")
)
(define (increase-time)
    (ti-menu-load-string "solve dual-time-iterate 1 1\n")
)
(define (set-x)
    (ti-menu-load-string "define user-defined execute-on-demand \"fixed::libudf\"\n")
)
(define (calculate-flow k)
    (ti-menu-load-string "define dynamic-mesh controls implicit-update-parameters update-interval 1\n")
    (ti-menu-load-string "solve iterate 1\n")
    (ti-menu-load-string "define dynamic-mesh controls implicit-update-parameters update-interval 1000000\n")
    (ti-menu-load-string "solve iterate 1000\n")
)
(define (calculate-force)
    (ti-menu-load-string "define user-defined execute-on-demand \"force::libudf\"\n")
)
(define (finalize-time-step)
    (ti-menu-load-string "define user-defined execute-on-demand \"logging::libudf\"\n")
    (ti-menu-load-string "define user-defined execute-on-demand \"timestep::libudf\"\n")
)
(define n-end 4000)
(initialize)
(do ((n 1 (+ n 1))) ((> n n-end))
    (increase-time)
    (set-x)
    (calculate-flow 1)
    (calculate-force)
    (finalize-time-step)
)

On the figure below a large discrepancy can be seen between the force (blue) from the UDF and Fluent report (orange). When calculating the force by integrating the pressure and viscous forces over the surface in a face loop, the same value as with Compute_Force_And_Moments is obtained.

On the figure below, it can be seen that also on the longer term a discrepancy lasts:

A test without moving body and just using the different force reports showed more satisfactory results, with almost no difference. Hence, it is deduced that the error is related to the motion of the body. 

I don't know how to further investigate the problem, so if anyone had a similar experience I'd be happy to learn from you ????.

[Luca B.]

Thanks for reporting this descrepancy. 

Compute_Force_And_Moment is not a documented function, so probably it could be dependant by a different way to interpolate.

Why do you need to calculate the Force in the UDF?

[nahian.masud]

hi, 
i tried using compute_fore_and_moment while simulating an airfoil. if i use the force to get the y axis velocity, there is a large discripency in force.

can you please tell me, how I can use and hook the second code that you used,to my airfoil?