Thunderhead Engineering Forum

Please login or register.

Login with username, password and session length
Advanced search  


Forum moved to

Author Topic: HVAC Vents and Ducts Configuration Appears to be Causing Numerical Instabilities  (Read 993 times)


  • New Member
  • *
  • Posts: 1
    • View Profile

In the past I have created and successfully run a number of models, varying in size, using vents with HVAC surfaces, either supplying or exhausting air from/to an ambient location utilizing an ambient HVAC node, vent endpoint HVAC node, and an HVAC duct connecting the two nodes. These ducts are then usually assigned a fan to induce the flow in the desired direction to either supply fresh air or exhaust smoke from an atrium type space.

Following the last FDS update (to FDS 6.6.0), the models I’ve run with these HVAC elements have not worked, and have instead very quickly reached numerical instabilities (within about 15-20 simulated seconds). Note that I have run very similar models on FDS 6.6.0 (that have failed) as on the previous version (FDS 6.5.0) that had worked to model completion (1200 seconds).  I used PyroSim 2017 to create the FDS files for these latest models.

The models also generally include holes that appear on a control activation, to simulate doors opening to the building exterior. The fans connected to the above-mentioned vents are also set to activate on a control, either time or detector based.

Originally I had run the models with a fire (generally 5’x5’ burner surface of about 3MW), however to attempt to debug I removed the fire and hoped the HVAC would work without any influence from pressure changes due to expanding smoke and air movement induced by the fire. Additionally, in reading some other forum postings (, I attempted an HVAC_PRES_RELAX=0.9 on the &MISC line, to no avail.

In the few seconds the model is simulated prior to crashing, a Z-plane vector slice file shows air flow appearing to come from supply vents that should not yet be activated. Then, upon doors opening (holes appearing) even greater air movement can be seen, in a space that technically should not yet have any air movement at all (seeing as how there is no fire or vents that should be operating). Images below at 3.6 and 10.8 seconds simulated time respectively; in this instance there is no fire, doors were set to open at 10 seconds and all fans begin a 60 second ramp up at 15 seconds.

The *.out files of the crashing models have very quickly indicated pressure iterations up to 10, leading me to believe even more so that somehow the HVAC solver is attempting to move air when it is not being asked to, and doing so in larger quantities and pressures than it realistically should be solving for in this instance.

Any help in determining whether this is a model issue or difference in FDS version issue would be greatly appreciated.