Makine Mühendisliği Bölümü Yayın Koleksiyonu
Permanent URI for this collectionhttps://hdl.handle.net/20.500.12416/263
Browse
9 results
Search Results
Article Citation - WoS: 5Citation - Scopus: 5Ann and Anfis Performance Prediction Models for Francis Type Turbines(Turkish Soc thermal Sciences Technology, 2020) Aylı, Ülkü Ece; Ayli, Ece; Ulucak, Oguzhan; Makine MühendisliğiTurbines can be operated under partial loading conditions due to the seasonal precipitation fluctuations and due to the needed electrical demand over time. According to this partial working need, designers generate hill chart diagrams to observe the system behavior under different flow rates and head values. In order to generate a hill chart, several numerical or experimental studies have been performed at different guide vane openings and head values which are very time consuming and expensive. In this study, the efficiency prediction of Francis turbines has been performed with ANN and ANFIS methods under different operating conditions and compared with simulation results. The obtained results indicate that it is possible to obtain a hill chart using ANFIS method instead of a costly experimental or numerical tests. ANN and ANFIS parameters which effect the output, have been optimized with trying 100 different cases. 75% of the numerical data set is used for training and 25 % is used for validation as testing data. To asses and compare the performance of multiple ANN and ANFIS models several statistical indicators have been used. Insight to the performance evaluation, it is seen that ANFIS can predict the efficiency distribution with higher accuracy than the ANN model. The developed ANFIS model predicts the efficiency with 1.41% mean average percentage error and 0.999 R-2 value. To the best of the author's knowledge, this is the first study in the literature that ANN and ANFIS are used in order to predict the efficiency distribution of the turbines at different loading conditions.Article Citation - WoS: 10Citation - Scopus: 11Analysis of Heat Transfer Enhancement of Passive Methods in Tubes With Machine Learning(Sage Publications Ltd, 2024) Ayli, Ece; Turkoglu, Hasmet; Yapici, Ekin Ozgirgin; Özgirgin Yapıcı, EkinThis study investigates the efficacy of machine learning techniques and correlation methods for predicting heat transfer performance in a dimpled tube under varying flow conditions, including the presence of nanoparticles. A comprehensive numerical analysis involving 120 cases was conducted to obtain Nusselt numbers and friction factors, considering different dimple depths and velocities for both pure water and water-Al2O3 nanofluid at 1%, 2%, and 3% volume concentrations. Utilizing the data acquired from the numerical simulations, a correlation equation, SVM ANN architectures were developed. The predictive capabilities of the statistical approach, ANN, and SVM models for Nusselt number distribution and friction factor were meticulously assessed through mean average percentage error (MAPE) and correlation coefficients (R2). The research findings reveal that machine learning techniques offer a highly effective approach for accurately predicting heat transfer performance in a dimpled tube, with results closely aligned with Computational Fluid Dynamics (CFD) simulations. Particularly noteworthy is the superior performance of the ANN model, demonstrating the most precise predictions with an error rate of 2.54% and an impressive R2 value of 0.9978 for Nusselt number prediction. In comparison, the regression model achieved an average error rate of 6.14% with an R2 value of 0.8623, and the SVM model yielded an RMSE value of 2.984% with an R2 value of 0.9154 for Nusselt number prediction. These outcomes underscore the ANN model's ability to effectively capture complex patterns within the data, resulting in highly accurate predictions. In conclusion, this research showcases the promising potential of machine learning techniques in accurately forecasting heat transfer performance in dimpled tubes. The developed ANN model exhibits notable superiority in predicting Nusselt numbers, making it a valuable tool for enhancing thermal system analyses and engineering design optimization.Article Citation - WoS: 13Citation - Scopus: 14A Novel Cfd-Ann Approach for Plunger Valve Optimization: Cost-Effective Performance Enhancement(Elsevier Sci Ltd, 2024) Kaak, Abdul Rahman Sabra; Celebiog, Kutay; Bozkus, Zafer; Ulucak, Oguzhan; Ayli, Ece; Çelebioğlu, KutayThis paper introduces a novel computational fluid dynamics-artificial neural network (CFD-ANN) approach that has been devised to enhance the efficiency of plunger valves. The primary emphasis of this research is to achieve an optimal equilibrium between hydraulic flow and geometric configuration. This study is a novel contribution to the field as it explores the flow dynamics of plunger valves using Computational Fluid Dynamics (CFD) and proposes a unique methodology by incorporating Machine Learning (ML) for performance forecasting. An artificial neural network (ANN) architecture was developed using a thorough comprehension of flow physics and the impact of geometric parameters acquired through computational fluid dynamics (CFD). Using optimization, the primary aspects of the Artificial Neural Network (ANN), including the learning algorithm and the number of hidden layers, have been modified. This refinement has resulted in the development of an architecture exhibiting a remarkably high R2 value of 0.987. This architectural design was employed to optimize the plunger valve. By utilizing Artificial Neural Networks (ANN), a comprehensive analysis comprising 1000 distinct configurations was effectively performed, resulting in a significant reduction in time expenditure compared to relying on Computational Fluid Dynamics (CFD). The result was a refined arrangement that achieved maximum head loss, subsequently verified using computational fluid dynamics (CFD) simulations, resulting in a minimal discrepancy of 2.66%. The efficacy of artificial neural networks (ANN) becomes apparent due to their notable cost-efficiency, along with their capacity to produce outcomes that are arduous and expensive to get through conventional optimization research utilizing computational fluid dynamics (CFD).Article Citation - WoS: 5Citation - Scopus: 6Prediction of the Heat Transfer Performance of Twisted Tape Inserts by Using Artificial Neural Networks(Korean Soc Mechanical Engineers, 2022) Kocak, Eyup; Ayli, EceA numerical study is undertaken to investigate the effect of twisted tape inserts on heat transfer. Twisted tapes with various aspect ratios and single, double, and triple inserts are placed inside a tube for Reynolds numbers ranging from 8000 to 12000. Numerical results show that the tube with a twisted tape and different numbers of tape is more effective than the smooth tube in terms of thermo-hydraulic performance. The highest heat transfer is achieved with the triple insert, with the highest turning number and an increment of 15 %. Then, an artificial neural network (ANN) model with a three-layer feedforward neural network is adopted to obtain the Nusselt number on the basis of four inputs for a heated tube with a twisted insert. Several configurations of the neural network are examined to optimize the number of neurons and to identify the most appropriate training algorithm. Finally, the best model is determined with one hidden layer and thirteen neurons in the layer. Bayesian regulation is chosen as the training algorithm. With the optimized algorithm, excellent precision for measuring the output is provided, with R2 = 0.97043. In addition, the optimized ANN architecture is applied to similar studies in the literature to predict the heat transfer performance of twisted tapes. The developed ANN architecture can predict the heat transfer enhancement performance of similar problems with R2 values higher than 0.93.Article Citation - WoS: 8Citation - Scopus: 8Supervised Learning Method for Prediction of Heat Transfer Characteristics of Nanofluids(Korean Soc Mechanical Engineers, 2023) Kocak, Eyup; Ayli, EceThis study focuses on the alication and investigation of the predictive ability of artificial intelligence in the numerical modelling of nanofluid flows. Numerical and experimental methods are powerful tools from an accuracy point of view, but they are also time- and cost-consuming methods. Therefore, using soft-computing techniques can improve such CFD drawbacks by patterning the CFD data. After obtaining the aropriate ANN and ANFIS architecture using the CFD data, many new data can be created without requiring numerical and experimental methods. In the scope of this research, the FCM-ANFIS and ANN methods are used to predict the thermal behaviour of the turbulent flow in a heated pipe with several nanoparticles. A parametric CFD study is carried out for water-TiO2, water-CuO, and water-SiO2 nanofluid through a pipe. The Reynolds number is varied between 7000 and 15000, and the nanofluid concentration is varied between 0.25 % and 4 %. The effects of using nanofluid on local values of Nusselt number and shear stress distribution were investigated. Numerical results indicate that with the increasing nanoparticle volume fraction of nanofluid, the average Nusselt number increases, but the required pumping power also increases. The obtained soft computing results demonstrate that the FCM clustering ANFIS has given better results both in training and testing when it is compared to the ANN architecture with an R-2 of 0.9983. Regarding this, the FCM-ANFIS is an excellent candidate for calculating the Nusselt number in heat transfer problems.Article Citation - WoS: 1Citation - Scopus: 1Prediction of the Onset of Shear Localization Based on Machine Learning(Cambridge Univ Press, 2023) Ayli, Ece; Ulucak, Oguzhan; Ugurer, Doruk; Akar, Samet; Ayll, EcePredicting the onset of shear localization is among the most challenging problems in machining. This phenomenon affects the process outputs, such as machining forces, surface quality, and machined part tolerances. To predict this phenomenon, analytical, experimental, and numerical methods (especially finite element analysis) are widely used. However, the limitations of each method hinder their industrial applications, demanding a reliable and time-saving approach to predict shear localization onset. Additionally, since this phenomenon largely depends on the type and parameters of the constitutive material model, any change in these parameters requires a new set of simulations, which puts further restrictions on the application of finite element modeling. This study aims to overcome the computational efficiency of the finite element method to predict the onset of shear localization when machining Ti6Al4V using machine learning methods. The obtained results demonstrate that the FCM (fuzzy c-means) clustering ANFIS (adaptive network-based fuzzy inference system) has given better results in both training and testing when it is compared to the ANN (artificial neural network) architecture with an R-2 of 0.9981. Regarding this, the FCM-ANFIS is a good candidate to calculate the critical cutting speed. To the best of the authors' knowledge, this is the first study in the literature that uses a machine learning tool to predict shear localization.Article Citation - WoS: 6Citation - Scopus: 7Machine Learning Based Developing Flow Control Technique Over Circular Cylinders(Asme, 2023) Turkoglu, Hasmet; Ayli, Ece; Kocak, EyupThis paper demonstrates the feasibility of blowing and suction for flow control based on the computational fluid dynamics (CFD) simulations at a low Reynolds number flows. The effects of blowing and suction position, and the blowing and suction mass flowrate, and on the flow control are presented in this paper. The optimal conditions for suppressing the wake of the cylinder are investigated by examining the flow separation and the near wake region; analyzing the aerodynamic force (lift and drag) fluctuations using the fast Fourier transform (FFT) to separate the effects of small-scale turbulent structures in the wake region. A method for stochastic analysis using machine learning techniques is proposed. Three different novel machine learning methods were applied to CFD results to predict the variation in drag coefficient due to the vortex shedding. Although, the prediction power of all the methods utilized is in the acceptable accuracy range, the Gaussian process regression (GPR) method is more accurate with an R-2(coefficient of determination) > 0.95. The results indicate that by optimizing the blowing and suction parameters like mass flowrate, slot location, and the slot configuration, up to 20% reduction can be achieved in the drag coefficient.Article Citation - WoS: 16Citation - Scopus: 17A Comparative Study of Multiple Regression and Machine Learning Techniques for Prediction of Nanofluid Heat Transfer(Asme, 2022) Ayli, Ece; Turkoglu, Hasmet; Kocak, EyupThe aim of this article is to introduce and discuss prediction power of the multiple regression technique, artificial neural network (ANN), and adaptive neuro-fuzzy interface system (ANFIS) methods for predicting the forced convection heat transfer characteristics of a turbulent nanofluid flow in a pipe. Water and Al2O3 mixture is used as the nanofluid. Utilizing fluent software, numerical computations were performed with volume fraction ranging between 0.3% and 5%, particle diameter ranging between 20 and 140 nm, and Reynolds number ranging between 7000 and 21,000. Based on the computationally obtained results, a correlation is developed for the Nusselt number using the multiple regression method. Also, based on the computational fluid dynamics results, different ANN architectures with different number of neurons in the hidden layers and several training algorithms (Levenberg-Marquardt, Bayesian regularization, scaled conjugate gradient) are tested to find the best ANN architecture. In addition, ANFIS is also used to predict the Nusselt number. In the ANFIS, number of clusters, exponential factor, and membership function (MF) type are optimized. The results obtained from multiple regression correlation, ANN, and ANFIS were compared. According to the obtained results, ANFIS is a powerful tool with a R-2 of 0.9987 for predictions.Article Citation - WoS: 7Citation - Scopus: 7A Study on the Μwire-Edm of Ni55.8ti Shape Memory Superalloy: an Experimental Investigation and a Hybrid Ann/Pso Approach for Optimization(Springer Heidelberg, 2023) Seyedzavvar, Mirsadegh; Boga, Cem; Akar, SametThe unique properties of high hardness, toughness, strain hardening, and development of strain-induced martensite of nickel-titanium superalloys made the micro-wire electro discharge machining (mu wire-EDM) process one of the main practical options to cut such alloys in micro-scale. This paper presents the results of a comprehensive study to address the response variables of Ni55.8Ti superalloy in mu wire-EDM process, including the kerf width (KW), material removal rate (MRR), arithmetic mean surface roughness (R-a) and white layer thickness (WLT). To this aim, the effects of pulse on-time (T-on), pulse off-time (T-off), discharge current (I-d) and servo voltage (SV) as input parameters were investigated using the experiments conducted based on Taguchi L-27 orthogonal array. The results were employed in the analysis of variance (ANOVA) to examine the significance of input parameters and their interactions with the output variables. An optimization approach was adopted based on a hybrid neural network/particle swarm optimization (ANN/PSO) technique. The ANN was employed to achieve the models representing the correlation between the input parameters and output variables of the mu wire-EDM process. The weight and bias factor matrices were obtained by ANN in MATLAB and together with the feed forward/backpropagation model and developed functions based on PSO methodology were used to optimize the input parameters to achieve the minimum quantities of KW, R-a and WLT and the maximum value of MRR, individually and in an accumulative approach. The results represented a maximum accumulative error of nearly 8% that indicated the precision of the developed model and the reliability of the optimization approach. At the optimized level of input parameters obtained through the accumulative optimization approach, the KW, R-a, and WLT remained nearly intact as compared with the levels of responses obtained in the individual optimization approach, while there was a sacrifice in the machining efficiency and reduction in the MRR in the mu wire-EDM process of Nitinol superalloy.