Testing of different meta models available in the CFD-optimiser to define the best fitting model for CD-optimisation
Başlık çevirisi mevcut değil.
- Tez No: 516508
- Danışmanlar: Dr. GÜLDEN KAYNAK, Dr. ERICH JEHLE-GRAF, Prof. JOERG FLIEGE
- Tez Türü: Yüksek Lisans
- Konular: Biyomühendislik, Matematik, Mühendislik Bilimleri, Bioengineering, Mathematics, Engineering Sciences
- Anahtar Kelimeler: Belirtilmemiş.
- Yıl: 2012
- Dil: İngilizce
- Üniversite: University of Southampton
- Enstitü: Yurtdışı Enstitü
- Ana Bilim Dalı: Belirtilmemiş.
- Bilim Dalı: Belirtilmemiş.
- Sayfa Sayısı: 122
Özet
Özet yok.
Özet (Çeviri)
Response surface models rely on design of experiments and empirical modelling approaches to improve and optimise processes. Over the last few years, there has a growing interest in the use of response surface models (RSMs). The main reason is that the physical experiment is both very expensive and time consuming to perform. Generally, response surface models are defined as meta-models and they focus on either the optimisation performance or the quality of prediction to achieve a desirable model. Hence, the purpose of this study was to investigate models based on both the prediction quality and the optimisation performance by comparing different meta-models on seven dimensional Mercedes-Benz Passenger Car optimisation problems to find best fitting model. In addition, in this studying, another aim was to find out a way to get an appropriate response surface with performing different number of experiments and based on the experiment numbers the performance of meta-models was evaluated to find the best methods with respect to the number of experiments. Procedure and Methods Response surface modelling is used to generate a model of unknown response surface for optimisation. Hence, creating a good response surface is a significant piece of optimisation. In this case, Design of Experiment methods are essential to create appropriate design points and to demonstrate how design points should be varied to investigate their impact on the response surface. In this studying, Latin hypercube design method was used to obtain convenient experiments. In addition, one of the purposes of this work was to examine models in terms of the prediction quality and the optimisation performance of models. The thesis investigated four meta-models: one parametric model: Linear Model, three non-parametric approaches: Support Vector Machine (SVM), Random Forest (RF) and Gaussian Processes (GP). Each model was analysed in statistical way to assess performance and accuracy of the model. Furthermore, two approaches were generated to get an appropriate response surface with performing different number of experiments. The first approach named as Stepwise consisted of five stages. In the beginning, the importance of independent variables known as shape parameters of Mercedes-Benz Cars was checked statistically. In the second stage, the most important three parameters were selected for modelling by using four meta-models, in this case, different meta-models leaded to different response surfaces. Therefore, the quality and performance of each method was measured statistically and optimal model result was carried out to next step. In the next step, the accuracy of the model was controlled by running an experiment. When the step was finished, these parameters were kept in their values that were found in previous steps and rest parameters were taken to generate four models and to assess them by using same evaluation criteria as well as in previous stages. In the final step, the model accuracy was checked and optimum values for parameters were obtained. On the other hand, the second approach called as One-step was about the analysis of all parameters together by following the same way as in the Stepwise approach to get an optimum. These approaches were applied to two different Mercedes-Benz Cars (SL-Class and B-Class) and both results were used to evaluate accuracy of meta-models.
Benzer Tezler
- Gaz türbini kanat ucu geometrisinin aerodinamik ve ısıl optimizasyonu
Aerothermal optimization of axial gas turbine blade tip geometry
HIDIR MARAL
Doktora
Türkçe
2018
Makine Mühendisliğiİstanbul Teknik ÜniversitesiMakine Mühendisliği Ana Bilim Dalı
DR. ÖĞR. ÜYESİ LEVENT ALİ KAVURMACIOĞLU
- Devre bağlaşmalı telefon şebekesi için yönlendirme yazılımı tasarımı
Software design of routing for circuit switched telephone network
TAHİR GÜN
Yüksek Lisans
Türkçe
1992
Elektrik ve Elektronik Mühendisliğiİstanbul Teknik ÜniversitesiYRD. DOÇ. DR. ÜMİT AYGÖLÜ
- Finansal zaman serilerinin fonksiyonel yapısının genetik öğrenmeyle belirlenmesi
Determining the functional structure of financial time series by means of genetic learning
ÖZGÜR İCAN
Doktora
Türkçe
2013
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolAnadolu Üniversitesiİşletme Ana Bilim Dalı
PROF. DR. HASAN DURUCASU
- Carbon dioxide emissions prediction using meta-heuristic methods for renewable and non-renewable sources' applications
Yenilenebilir ve yenilenemeyen enerji kaynaklarının uygulamaları için meta-sezgisel yöntemler kullanılarak karbondioksit emisyonları tahmini
İNAYET ÖZGE AKSU
Doktora
İngilizce
2022
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolAdana Alparslan Türkeş Bilim Ve Teknoloji ÜniversitesiElektrik-Elektronik Mühendisliği Ana Bilim Dalı
DOÇ. DR. TUĞÇE DEMİRDELEN
- Prediction of COVID 19 disease using chest X-ray images based on deep learning
Derin öğrenmeye dayalı göğüs röntgen görüntüleri kullanarak COVID 19 hastalığının tahmini
ISMAEL ABDULLAH MOHAMMED AL-RAWE
Yüksek Lisans
İngilizce
2024
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolGazi ÜniversitesiBilgisayar Mühendisliği Ana Bilim Dalı
PROF. DR. ADEM TEKEREK