Poppet Valve Optimization

An auto­mated design study was con­ducted by OMIQ srl on a poppet valve from Danfoss High Pressure Pumps, Denmark. During bench tests the valve was found to exhibit unstable dynamic behavior that was unac­cept­able. When the valve opened to its maximum dis­place­ment (27.5 mm), flow insta­bil­i­ties acted to reduce the pressure forces on the poppet, which resulted in the valve not staying com­pletely open (closing to ~6 mm). These unstable phe­nom­ena were verified by tran­sient sim­u­la­tions using Simer­ic­sMP®.

To inves­ti­gate this problem and to improve the design, OMIQ used CAESES® and Simer­ic­sMP® in an auto­mated workflow to explore the design space of the valve. Although the valve’s behavior is inher­ently tran­sient, it was decided to conduct steady-state sim­u­la­tions at the fully-open position with a flow rate of 53 m³/​h. This approach was deemed to be valid since the pressure forces are maximum in the fully-open position, and with the added benefit of much faster simulations.

Para­met­ric Model of the Poppet

CAESES® was used to create a para­met­ric model of the poppet described by 4 para­me­ters. The vari­a­tions were restricted to the under­side surfaces of the poppet since there was more freedom to modify those surfaces and high-pressure forces in that region had a large impact on the valve’s behavior. The vari­a­tions were con­strained such that the distance between the poppet and the support in the fully-open position would be maintained.

Opti­miza­tion

The CAESES® opti­miza­tion toolkit was used and a Sobol algo­rithm was selected to conduct a design of exper­i­ments (DoE) with 30 designs to make a sweep through the design space. A sub­se­quent local opti­miza­tion using the T‑search method from the best DoE design was used, however, little further improve­ment was found, thus the best DoE design was con­sid­ered to be the ​“optimal” design.

Results

The baseline tran­sient cal­cu­la­tion showed that the spring force at the fully-open position is 12.64 N, which indi­cated that the fluid forces acting on the poppet must be higher than this value for the valve to remain fully open. For struc­tural reasons, the fluid force was confined to a value lower than 17 N, which is the value of the spring force when the valve is in contact with its back support.

The DoE study with 30 designs was suc­cess­ful in uncov­er­ing a design that solved the unwanted insta­bil­ity problem. Mod­i­fi­ca­tions to the under surfaces of the poppet were effec­tive at reducing flow insta­bil­i­ties that led to the unstable dynamic behavior, causing the valve to par­tially close. The figure below shows the geo­met­ric dif­fer­ences between the baseline and best DoE design.

The fol­low­ing figures show several velocity and pressure contour and vector plots, which depict the dif­fer­ences in the flow field. 

Ver­i­fi­ca­tion of the Poppet Valve Optimization

To verify the valve’s dynamic behavior was func­tion­ing cor­rectly, a tran­sient sim­u­la­tion was run on the best DoE design. Whereas the baseline model opened to 27.5 mm and then closed back, the best DoE design opened to 29 mm in 0.039 s and stayed open. Fur­ther­more, in the fully-open position (and all posi­tions) the maximum spring force was found to be ~13 N, well below the design limit of 17 N.

Con­clu­sion

In con­clu­sion, prob­lem­atic dynamic insta­bil­i­ties in the oper­a­tion of the baseline poppet valve were effec­tively solved by setting up an auto­mated workflow with CAESES® and Simer­ic­sMP®. A DoE study of 30 designs was run quickly and effec­tively where design variants were auto­mat­i­cally gen­er­ated by CAESES® and sim­u­lated by Simer­ic­sMP® in steady-state. The results were verified by follow-up tran­sient flow sim­u­la­tions in Simer­ic­sMP® and finally by bench tests at Danfoss HPP.

Mission accom­plished! 😊

Acknowl­edge­ment

The original case study can be found on OMIQ’s website.