Refprop Execution time optimization

[rigojea-bh]

Hello NIST team,

Context

We have a series of design and selector tools in the field of turbomachinery that has leveraged Refprop for more than 2 years now, embedding it as a Fortran static library serving other Fortran code when calling thermodynamic properties of various mixtures involved, in particular industrial, in Centrifugal Compressors treating natural gas, refrigerants (similitude testing), or other compounds such as N2 or ammonia in petrochemical or CO2 in CCUS applications.

Description

We have observed, in several cases involving intense iterative convergence search, such as resolution of tooth-by-tooth leakages at end seals, an inflation of execution time by at least 2 orders of magnitude with respect to older code implementation of thermodynamic property calculations leveraging legacy equation of states (BWRS, LKP, PR, SRK).
That is creating issues in our bid and job execution process, since calculations may sometimes take hours to complete a compressor plant design where they can take just a few minutes with the traditional EoS.

A first examination of the Refprop code, we noted that one of the causes may be some "wasted" time reading the FLD files, while other implementations that we have use pre-compiled code (modules) that contain the gas properties database.

We would like to understand if there are specific options in the code to optimize execution time.

Thanks in advance,
Jean-Christophe Rigo
Staff Software Engineer
Baker Hughes

Versions

REFPROP Version: 10.0
Operating System and Version: RHEL Linux 8.10
Access Method: compiled as a refprop.a static library through "make" and a tailored makefile.

[ianhbell]

What sort of thermodynamic inputs cause these troubles? In general more accurate EOS = slower. See for instance the benchmarks in https://pubs.acs.org/doi/full/10.1021/acs.iecr.2c00237