Time Step significantly affects Model Output

Question

When I run my pump station failure scenario with no protection (Pump shutdown of all 4 pumps after 5 seconds, control valve closes over 300 seconds) at a time step of 0.1s the model results in negative transient pressures (min of -9.71 psi at Pump suction). When I run the same scenario and only change the time step to 0.01 seconds, the model results in positive pressures (min of 3.79 psi at Pump suction). I am not sure which results to believe...Please Help! Thanks! 
 Time Step 0.1s

Time Step of 0.01s

Scott Kampa · Accepted Answer

Hello Venessa, 
 The smaller time step is more likely to have the more accurate results. The reason for this is because of one of the ways that the transient solver operates. A pressure wave will need to move from one end of the pipe to other in even increments with the time step. For that reason, HAMMER adjusts the lengths (or wave speed) of the pipes in the model. If the length adjustment is greater than the tolerance for this, a user notification is generated that says ""WARNING: The wave speed or length approximations deviate excessively from the entered values. Lengthen short pipes and/or subdivide longer pipes." 
 This user notification is likely occurring for the run with the larger time step. If the adjustment is too large, this may have a negative impact on the results. The following link has more information on length adjustments and why they occur: Understanding length/wave speed adjustments and their impact on results . 
 Please let us know if you have any questions on this. 
 Regards, 
 Scott

Jesse Dringoli · Answer

I concur with Scott's answer. I'll also add that wave propagation and reflection tends to be chaotic in nature in some situations, especially on the suction side of a pump, at dead ends, in closed off places with zero flow (no friction to help dampen the transient energy) and in systems that experience vapor pocket formation and collapse (such pockets can form and collapse rapidly). 
 In the case of a pump suction side, the waves travel so fast and reflect back and forth so rapidly that a small change like the one in question can have a relatively large impact, since it impacts the timing of the wave reflection and interaction. I would trust the results with the smaller timestep more, especially if there are other fast changes or vapor/ air pockets occurring in this system. A timestep of 0.1 seconds is large for a transient simulation and 0.01 seconds is more typical. 
 If you're not doing so already, I would highly recommend animating profiles in the Transient Results Viewer to understand what happens during the transient event. You may need to limit the number of report points to ensure that the profile is showing enough detail and not missing out on things that are happening between timesteps.