Opti­miza­tion of a Globe Valve with CFD in the Cloud

CAESES has an exten­sive pedigree in the opti­miza­tion of valves of any kind. The aims of this project, which was carried out in coop­er­a­tion with GEMÜ Gebr. Müller Appa­rate­bau and SimScale, a German valve man­u­fac­turer and world leader in valves for sterile appli­ca­tions, and a leading provider for engi­neer­ing sim­u­la­tion in the cloud, respec­tively, were twofold: explo­ration of the poten­tial and possible avenues for improve­ment of a valve, and inte­gra­tion of SimScale as a cloud-based CFD solver in the process.

Subject and Goal of the Optimization

The subject of the opti­miza­tion was the GEMÜ Globe Valve 534, a pneu­mat­i­cally operated globe valve with a metal body and plastic piston actuator that can be operated as a shut-off or control valve. It is used in indus­trial appli­ca­tions with non-aggres­sive media, such as water treat­ment, mechan­i­cal engi­neer­ing and pro­cess­ing, power gen­er­a­tion, and envi­ron­men­tal engineering.

As a clas­si­cally shaped globe valve, it exhibits sharp changes in the direc­tion of the flow, which create a poten­tial for opti­miza­tion by reducing the pressure loss in the valve. Within previous inves­ti­ga­tions, GEMÜ had expe­ri­enced issues with other CAD tools related to the robust­ness of geometry vari­a­tion. This had pre­vented the explo­ration of any larger design spaces and moti­vated the use of CAESES as a ded­i­cated tool for flexible and robust geometry vari­a­tion.
The specific goal of the inves­ti­ga­tion was to maximize the so-called flow coef­fi­cient Kv, which quan­ti­fies the flow rate through a valve at a pre­scribed pressure drop of 1 bar across the valve.

Geometry Vari­a­tion Setup

The geometry of the valve was provided by GEMÜ in STEP format and imported into CAESES for para­met­ric remod­el­ing. More specif­i­cally, the geometry to be remod­eled was the internal fluid volume of the valve, which was extracted from the imported data.

This fluid volume pri­mar­ily consists of an inlet and an outlet passage, which are sep­a­rated by the seat onto which the plug closes. The shape of each passage was modeled with two contours, which shall be des­ig­nated as the short and long contour (in green and blue in the picture below, respec­tively). Their shape can be con­trolled by 2 para­me­ters for each of the short contours and 5 para­me­ters for the long contours.

The cross-sections were gen­er­ated by creating a point on each the short and long contour. These points were con­nected by an ellipse. The length of the vertical axis is a result of the distance of the two points, while the length of the hor­i­zon­tal axis is defined relative to the vertical axis by an aspect ratio para­me­ter, which is pre­scribed as a function along the path of the passage. The shape of this function, and there­fore the width dis­tri­b­u­tion of the passage, are con­trolled by two para­me­ters for each the inlet and outlet passage (para­me­ters 15 to 18). In addition, two more para­me­ters allow for a vari­a­tion of the sweeping speed of the cross-section along the long contour (para­me­ters 19 and 20).

After the gen­er­a­tion of the para­met­ric passages, the geometry was com­pleted by the missing parts, such as the seat, plug, and bonnet (par­tially from imported elements). For purposes of the CFD setup, inlet and outlet were extended in up and down­stream direction.

CFD Automa­tion

The CFD inte­gra­tion in this opti­miza­tion workflow was somewhat special. CAESES is a desktop appli­ca­tion, so typ­i­cally, CAESES users connect to an on-premise CFD solver running on their local work­sta­tion or on an in-house HPC cluster. SimScale, however, is an engi­neer­ing sim­u­la­tion platform ​“in the cloud”, so a direct con­nec­tion between the local machine and the cloud envi­ron­ment had to be estab­lished. This was made possible by SimScale’s newly provided API that allows third-party appli­ca­tions to set up cases, run jobs, and retrieve results without manual inter­ac­tion in the browser-based work­bench – a pre­req­ui­site for being able to run auto­mated opti­miza­tion studies.

The com­mu­ni­ca­tion with SimScale’s API happens through a Python script that contains all the nec­es­sary commands to upload the CAD data, import it into SimScale, set up the case, run it, write the result files, and send them back to the local machine.

On the CAESES side, the con­nec­tion to SimScale was accom­plished through the usual mech­a­nism, namely CAESES’ Software Con­nec­tor inter­face. The input files gen­er­ated by CAESES for each variant are a STEP export of the geometry, the afore­men­tioned Python script, and a shell script that executes it. A CSV file with the flow coef­fi­cients is imported after the com­pu­ta­tion and parsed for the required results that are used within the opti­miza­tion process.

As a first test of the con­nec­tion, an auto­mated valve curve was run, which consists of com­put­ing the flow coef­fi­cient for a series of plug lift posi­tions, from fully closed to fully open.

Opti­miza­tion Process and Results

The opti­miza­tion process was run in several phases. The first campaign included a total of 16 para­me­ters: the first 12 contour shape para­me­ters, and the 4 para­me­ters related to the aspect ratio dis­tri­b­u­tion (15 to 18, see above). It started with a Design-of-Exper­i­ments (DoE) encom­pass­ing 150 design variants, to broadly scan the avail­able design space. The database obtained from this DoE was used to generate the sur­ro­gate model (response surface) utilized by the fol­low­ing opti­miza­tion process. During this sur­ro­gate-based opti­miza­tion, another 50 design variants were iter­a­tively computed by CFD to check the pre­dic­tion from the sur­ro­gate and augment its precision.

The result from the first opti­miza­tion yielded an improve­ment of Kv of 6.5%. It can gen­er­ally be said that the result­ing geometry exhibits a more bulgy shape in both the inlet and outlet passage, which might give the flow more space to smoothly develop when crossing the gap between seat and plug.

For the second opti­miza­tion phase, with the aim of squeez­ing out a bit more per­for­mance, 8 of the para­me­ters from the first opti­miza­tion were retained. These were the para­me­ters that proved to have the most influ­ence on the objec­tive function. In addition, two more contour para­me­ters were added (13 and 14), as well as the two sweeping speed para­me­ters (19 and 20, see above). The second opti­miza­tion was directly run as sur­ro­gate-based opti­miza­tion with a total of 160 design variants computed. In this case, the opti­miza­tion algo­rithm creates the initial database by itself and then iter­a­tively refines it with addi­tional com­pu­ta­tions, as before. This run was followed up with a final small local opti­miza­tion using only the two sweeping speed parameters.

The final opti­mized design could further improve on Kv, reaching a total improve­ment of 9%. While the shape of the (lower) inlet passaged had similar char­ac­ter­is­tics as the previous opti­mized design, the outlet passage showed an opposite devel­op­ment, by being more slender than the baseline.

This appli­ca­tion case was there­fore suc­cess­fully con­cluded, reaching all set goals: SimScale could suc­cess­fully be inte­grated in an auto­mated opti­miza­tion process as a cloud-based solution, being able to readily provide sub­stan­tial com­put­ing power when only limited capa­bil­i­ties are locally avail­able, and the valve design could be improved by sig­nif­i­cantly raising the flow coefficient.

More Infor­ma­tion

Watch the webinar record­ing about this case study here.

See this overview for the pos­si­bil­i­ties and capa­bil­i­ties that CAESES offers for valve design.