Prediction of Turbulence Statistics in a Model Dump Combustor Using Artificial Neural Networks

الموضوعات

علوم تطبيقية وتكنولوجية

الملخص

The flowfield characteristics downstream of an axisymmetric suddenexpansion dump combustor model are important to designers of gas turbines and liquid-fuel ramjets ducted rockets. Many experimental techniques such as Laser Doppler Velocimetry (LDV) measurements provide only limited discrete information at given points; especially, for the cases of complex flows such as swirling flows of a dump combustor. For these types of flows, usual numerical interpolating schemes appear to be unsuitable. Artificial Neural Networks (ANN) methods are thus proposed as an alternative and the flow predictions obtained are tested and presented in this thesis. To predict the velocity components and the turbulence statistics obtained experimentally under a variety of swirl numbers, the use of a variety of ANN architectures is investigated. In each case, the predictions obtained are compared with published experimental data to determine the ANN structure that predicts the flow parameters most accurately. Moreover, the generated data is used to provide contour and surface plots to show the detailed flow characteristics throughout the model. The examined turbulence statistics are fluid flow velocity components in axial, radial, and tangential directions, in addition to Reynolds shear and normal stresses. Also triple velocity correlations are scrutinized. The investigation of ANN architecture variation shows that generalized feedforward network (GFF) with one hidden layer is the most efficient network, in which most turbulence statistics are predicted accurately. Moreover, the study shows that GFF network performs better when its built architecture uses Levenberg-Marquardt learning rule and Tanhaxon transfer function for weights update. The obtained results look promising, thus, ANN is utilized to enhance the understanding of the behavior of swirling, recalculating, axisymmetric, and turbulent flow inside dump combustors. ANN is employed to compute kinetic energy terms (production, diffusion, convection, and viscous dissipation as well as estimating stream function to recognize the recirculation regions at combustor's corners and centerline.

روابط وملفات

التعريف والنوع

رقم الوثيقة

0d129133-7a54-4af5-aa7f-999122f13e69

رقم العقد

0

نوع الوسائط

Crawler

نوع المحتوى

الرسائل العلمية

صيغة المصدر

رسائل ماجيستير

نوع الملف

pdf text

أسماء الملفات

46117_1.pdf

بيانات النشر

ألقاب المؤلفين

[{"name_ar":" Amin Moh'd AlSharif","title_ar":"اعداد","title_en":"Preparation"},{"name_ar":"Saad Ahmed ","title_ar":"اشراف","title_en":"Supervision"}]

اللغة

English

المصدر والدورية

اسم المصدر

Prediction of Turbulence Statistics in a Model Dump Combustor Using Artificial Neural Networks

المحتوى والصفحات

عدد الصفحات

0

إشراف وإعداد

الإشراف

Saad Ahmed

الإعداد

Amin Moh'd AlSharif

الاقتباسات الببليوغرافية