I am simulating hydraulic hammer using the WP component in the thermal hydraulic library. The resulting transient's damping ratio seems to be incredibly sensitive to the number of segments I define within the component. Which results do I trust?

Question

My acoustic courant number is less than 1, each segment is at least 5 diameters in length (although I'm uncertain if this rule of thumb will physically affect the solution so much as it better ensures solver stability and convergence), but varying the number of segments within the WP component changes the rate at which the system achieves a steady state after the downstream valve is closed. The frequency of oscillation doesn't appear to vary, but the peak pressure surge does seem to be slightly affected. I'm concerned because increasing the number of segments doesn't cause the transient to converge to some general damping ratio, and I'm unsure as to how many segments I should use nor which solution I should trust. Any input from those with experience in modelling hydraulic hammer in Easy5 would be greatly appreciated. Thanks!

eric.plevy · Accepted Answer

Just to follow up on this, I determined that it was independent of the transient friction model used. I only found the damping characteristics to be sensitive for small numbers of total WP components. Once I built a more representative, complex system with over 20 total WP components, system damping sensitivity to spatial and temporal discretization seemed to vanish. I realized that I was supposed to keep the length of each division GREATER THAN or equal to the rule of thumb mentioned in the post, not less than or equal to. I suspect that the full scale system has a higher total fluid volume and larger cumulative WP component length which allowed me to use a larger, internal discretization on individual WP components which stabilized the system by increasing the pressure differential between internal nodes of individual WP components. I suppose only having one or two short WP components with smaller internal discretizations created a numerically unstable system with lower pressure differentials between internal nodes, which may have contributed to numerical instability caused by the overdiscretization of a single, short WP component. I also was having run time hangups that I was able to attribute to the switch state methodology built into Easy5 in reference to the orifice transition Reynolds number. I had changed the number from 100 to close to 3000 with minimal impacts to the solution with a significant impact on the run time (although it still takes a while, even with the larger transition Re number).

I am simulating hydraulic hammer using the WP component in the thermal hydraulic library. The resulting transient's damping ratio seems to be incredibly sensitive to the number of segments I define within the component. Which results do I trust?

Top Replies