I used a program many years ago that I think was called ctrend. If this was the same named program, it is probably much different now in features and design after such a long time. From what I remember, I had an existing multistage compressor that I wanted to rate for significantly changed feed composition. I did not quite trust the simplified centrifugal compressor models in process engineering software where the unit op is a compressor stage. The specialized compressor program allowed simulation in smaller calculation chunks than compressor stages. I was able to model down to the level of the wheels inside each stage! I only needed to use a portion of the software's capability for my purpose, but I remember being impressed by the claims of the software for mechanical engineer/ rotating equipment specialist usage. The software was not really made for process engineers and I have found since then that my needs could be satisfied with Pro/II or AspenPlus tools. But years ago, the specialized compressor software supplied me with information that I could not find anywhere else.
From compressor-ctrend's forum name, perhaps there is a vested interest in selling something. But there is value in process engineers being made aware of tools available. Ctrend software is not common among process engineers and many likely have never heard of it. I appreciated the link to see what the software claims to do. In this case under discussion where composition is unchanged, modelling in the usual steady-state process engineering software should be good enough. But the calcs can get complicated with significant change in molecular weight and it is good to know there is help for those calcs.
Hi there,
Thanks for your reply. I do not know if the "old" program you are referring to is the same. I will try to explain the background behind it from a technical standpoint and some milestones..
Originally the program (~ 2019) was written to model the performance of multi-stage compressors by modelling the performance of each stage individually and then stage-stacking the stages to determine the performance envelope of the entire phase or process section.
The problem encountered was that the performance model is dependant upon a high number of variables (diameter, flow coefficient, peripheral mach number, material roughness, etc). So the initial model was based on an artificial neural network concept that was trained over "typical" but very scattered performance database across 2D with and without spliters and 3D range (pipeline, normal and high Mach No. application). Then a proprietary algorithm was developed to stack thermodynamically the performance. Typically the end result is that for a given process duty, the program as developed allowed to obtain a "descent" estimate of the stage performance and subsequently estimate shaft span, rotating speed well inline with industry benchmark. The bottom line being versatility so that in essence, any 2D/3D stage database (including stage data generated from CFD simulation) can be learned by the neural net and adapted to generate performance. It is to be noted that the inlet and outlet section are no loss at present development stage.
Most practical feature for the user is the ability to iterate design and simulate process cases with several option of iterative variables (namely speed, inlet pressure, outlet pressure and flow). The caveat is that the performance is bound to the data being fed into the machine learning system. The beauty of it on the other hand is that the impact on which impeller triggers choke and surge is captured with this kind of tool, and at the end of the day you would get an operating map similar exhibiting OEM-like features, especially when compressibility effects are predominant.
Some time later (~2021), CTrend has forked from the main program (which is still available as a module called SIZER) with aim to focus on existing units. So, the same concept originally applied in order to model a single stage performance has been derived/adapted to model the performance of a complete section (flange to flange). In other words, when the reference data/map of the compressor unit is available (usually curves "as expected" from OEM databook or better "as tested" maps), the tool will model a thermodynamic signature/characterization of the compressor (that is to say based upon non-dimensional coefficients/parameters) and based upon that generates new performance output at a given off-design conditions (that is, condition of operation at process parameters / gas composition shift vs. design and/or degraded state vs. pristine/clean state), PROVIDED the off-design conditions are within the Mach No. variability range of the characterizer (conditioned by the OEM data available). The process of regression analysis on non-dimensional data is not new, for instance I am aware of commercial software performing most probably on the similar concept, for example CMAP (https://www.compressormap.com/). CTrend has added the capability to leverage on artificial neural networks or operate on more conventional regression algorithms at user's convenience. With such modeling "fire-power" the tool can build characterization based on reference map PLUS alternative cases - if these data happen to be known/available from OEM. For example, if are dealing with an existing compressor operating on natural gas and you happen to have data for startup case operating on N2 - all from OEM-, then you may add this information to the learning algorithm so that the spectrum of predictability is increased further. Next step has been to complement the tool with DCS fetching module.
I hope sharing this info and like will benefit the technical community.
I myself will be happy to know about other programs and how they tackle simulation at off-design and stay tuned.
Thanks and kind regards.