14 replies to this topic

[StealthProg]

Does anyone know the name of the algorithm that is used in VMGSim/Hysys for allowing bi-directional transfer of properties?  Or have any references to published papers on the subject?

[Bobby Strain]

You should find the answer on the program licensor's support site. Or, contact them.

 

Bobby

[StealthProg]

It's really not the sort of information they like to publicize.

[serra]

what do you mean with bi-directional transfer of properties ?

solving recycles ? all simulators are able to do that with different algorithms,

selection depending from specific application.

[StealthProg]

Simulators are typically either sequential modular and all properties flow uni-directionally through the flow sheet, including recycles, or they are EO based in which case all equations are solved simultaneously.

 

However with Hysys, Unisim, Petrosim, VMGSim, which all have similar origins, information, e.g. properties, can flow in two directions both upstream and downstream at the same time, i.e. they are bi-directional.

 

It is this bi-directional algorithm i am interested in getting information on.

[serra]

no, there are hybrid solvers and there are papers discussing details (see Lucia etc.)

for example you can detect the sections which must be recalculated and solve only a subset,

some units (flash operations etc.) can be solved in two ways,

others (columns, reactors etc.) not,

etc. etc.

[StealthProg]

Thanks, yes i found a couple of his papers now. Interesting. Im not totally convinced this answers my question, for example, in some simulators the problem can be solved sequentially backwards, i realise this could be done in a large matrix and solved simulataneously, but it appears that they are actually solved sequentially with some variables being passed forward and some backward. In fact the sim42 source code seems to support this interpretation. Admittedly i havent spent a huge amount of time going through it, but each unitop is popped off a stack and solved and connected port values reset, forwards and backwards and inconsistencies flagged up. Is that part of the hybrid solver?

[serra]

it's a complex problem and different products (from Hyprotech, Simsci, GProms, Prode etc. etc.) proposed different solutions,

an efficient solution requires proper initialization of variables, identification of sequence and units to solve, local / global convergence etc.

many simulators include different solvers, allow to initialize variables etc. etc.

for simple problems I have tested global search with multistart solvers (Matlab toolbox), ok but slow (not efficient, in general).

[StealthProg]

Does anybody have any actual information on the bi-directional algorithms employed by several of these simulators?

[DarrenO]

Each object has its own self-contained solution code, and the solver is nothing more really than an event handler that builds and runs through stacks of objects and tells them to solve themselves over and over for as much as they can each time.  The system is built so that whenever new information is calculated into an object (a stream or a unit operation) then that object and its attached objects get pushed onto the solver stack again and the process repeats until no more new information can be calculated.  Roughly speaking.

[serra]

DarrenO,

thanks for the contribute,

possibly the solver in the sim42 code mentioned by StealthProg works in that way,

that would confirm there was some misinterpretation about the meaning of bi-directional algorithm,

Edited by serra, 28 August 2017 - 11:58 AM.

[DarrenO]

Bi-directional in terms of the heritage Hyprotech simulators means that information can be specified anywhere in the flowsheet and will eventually work its way up or downstream as needed to satisfy individual unit operation degrees of freedom until they all solve.  One sample case that ships with all the HYSYS flavours involves a piping network where composition is specified at the very downstream end of the pipe network, the non-sequential modular solver allows this composition to be calculated backwards all the way back to the most upstream pipe, where it then fills out the degrees of freedom for that pipe to start pressure calculations going forward along the flowsheet until the flowsheet is fully solved.  It isn't a true EO solution (as it isn't solved simultaneously), but it is bi-directional in terms of information flow allowing the solver to fill in bits as necessary.

 

I don't know anything about VMG-SIM, but I would assume it is similar as it is all ex-Hyprotech developers that wrote it.

[serra]

Stealthprog,

can you explain the red flag to post #8,

where I mention Hyprotech, Simsci, GProms, Prode etc. including solvers with similar features (meaning, for example, the capability to solve a pipeline from end, etc.) ?

As said in post #6 hybrid solutions allows to switch from local to global solvers in order to  improve efficiency (time savings),

you can solve sequentially (in two ways) some units (for example a series of H-P flash to model a pipeline with heat transfer)  but for other units (columns, reactors etc.) as far as I know hybrid solutions are more efficient.

[muhammad.naeem]

Dear All,

 

can some body tell me how I can read VMG simulation file in Hysys, if thats possible.

 

What is the procedure..

 

Need urgent support.

 

Muhammad Naeem

MEng (Specialization Energy & Environment)

Schulich School of Engineering

University of Calgary

 

 

[imeagor]

Dear All,

 

can some body tell me how I can read VMG simulation file in Hysys, if thats possible.

 

What is the procedure..

 

Need urgent support.

 

Muhammad Naeem

MEng (Specialization Energy & Environment)

Schulich School of Engineering

University of Calgary

 

perhaps you will be dissapointed