The purpose of this thesis work was to create and test a library of large scalable models in Modelica language which contains different domains: mechanical, thermal and electrical. Mechanical domain includes flexible beam and string, thermal domain includes heat conduction and heat exchanger, and electrical domain includes transmission line models. The models were implemented in the OpenModelica environment using Modelica language. OpenModelica Compiler (OMC) was used because of the convenience, different Modelica applications could also have been used. The work intends to highlight the performance of Modelica compiler with respect to scalability, which means increasing number of equations, in terms of compilation and simulation times, and it intends to contribute these key concepts. Models were implemented by discretizing their original partial differential equations (PDE) and/ or by the tools of Modelica Standard Library (MSL). All the models were discretized into N which is the number of nodes, segments or elements depending on the models. Models were function of N, therefore, discretization of the models could be enlarged and the results were compared in detail. In order to verify the models, analytical solutions or numerical formulas were implemented. In this thesis work, plots of the models are discussed and the performance of OMC are provided in terms of compilation and simulation times for increasing number of N values. Plots of the models have shown that as discretization increases, models reflect the expected results. However, compilation and simulation times grow significantly as discretization increases especially for the mechanical domain. It was observed that compiler and sequential simulation should be improved to support large models. Strategies such as sparse, multi rate solvers or parallelization become increasingly important to stay in reasonable time limits.
L’obiettivo di questa tesina è stato creare e testare in linguaggio Modelica una libreria di grandi modelli scalabili contenente diversi domini: meccanici, termici ed elettrici. Il dominio meccanico include la trave flessibile e la corda, quello termico include conduzioni e scambiatori di calore, mentre il dominio elettrico include modelli su linee di trasmissione. I modelli sono stati implementati in ambiente OpenModelica usando il linguaggio Modelica. Nonostante si sarebbero potuti utilizzare differenti applicazioni di Modelica è stato adoperato OpenModelica Compiler in quanto ritenuto più conveniente. Il progetto vuole mettere in luce la performance di Modelica Compliler rispettando la scalabilità, la quale significa un numero di equazioni in aumento, e intende contribuire a questi concetti chiave. I modelli sono stati implementati discretizzando le loro equazioni differenziali parziali originali e/ o usando gli strumenti di Modelica Standard Library. Tutti i modelli sono stati discretizzati in N che rappresenta il numero di nodi, segmenti o elementi a seconda dei modelli. I modelli, in funzione di N, permettono che la loro discretizzazione possa essere ingrandita ed è stato possibile paragonare i risultati in dettaglio. Per verificare questi modelli sono state implementate soluzioni analitiche o metodi numerici. In questa tesina sono discussi i grafici dei modelli e sono fornite le prestazioni di OpenModelica Compiler in termini di tempi di compilazione e simulazione per valori di N in aumento. I grafici dei modelli hanno mostrato che, all’aumentare della discretizzazione, essi rispecchiano i risultati attesi. D’altra parte, i tempi di simulazione e compilazione crescono significativamente con l’aumentare della discretizzazione, specialmente nel dominio meccanico. Inoltre è stato osservato che il compilatore e la simulazione sequenziale dovrebbero essere migliorati per supportare grandi modelli. Per rimanere in un limite di tempo ragionevole risultano importanti le strategie come i risolutori sparsi e multi rate o la parallelizzazione.
A test suite of large scalable models for Modelica tool evaluation
SEZGINER, KAAN
2014/2015
Abstract
The purpose of this thesis work was to create and test a library of large scalable models in Modelica language which contains different domains: mechanical, thermal and electrical. Mechanical domain includes flexible beam and string, thermal domain includes heat conduction and heat exchanger, and electrical domain includes transmission line models. The models were implemented in the OpenModelica environment using Modelica language. OpenModelica Compiler (OMC) was used because of the convenience, different Modelica applications could also have been used. The work intends to highlight the performance of Modelica compiler with respect to scalability, which means increasing number of equations, in terms of compilation and simulation times, and it intends to contribute these key concepts. Models were implemented by discretizing their original partial differential equations (PDE) and/ or by the tools of Modelica Standard Library (MSL). All the models were discretized into N which is the number of nodes, segments or elements depending on the models. Models were function of N, therefore, discretization of the models could be enlarged and the results were compared in detail. In order to verify the models, analytical solutions or numerical formulas were implemented. In this thesis work, plots of the models are discussed and the performance of OMC are provided in terms of compilation and simulation times for increasing number of N values. Plots of the models have shown that as discretization increases, models reflect the expected results. However, compilation and simulation times grow significantly as discretization increases especially for the mechanical domain. It was observed that compiler and sequential simulation should be improved to support large models. Strategies such as sparse, multi rate solvers or parallelization become increasingly important to stay in reasonable time limits.File | Dimensione | Formato | |
---|---|---|---|
2015_04_SEZGINER.pdf
solo utenti autorizzati dal 16/04/2016
Descrizione: Thesis text
Dimensione
1.6 MB
Formato
Adobe PDF
|
1.6 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/107221