Research Outputs
Permanent URI for this communityhttps://hdl.handle.net/20.500.14288/2
Browse
98 results
Filters
Advanced Search
Filter by
Settings
Search Results
Publication Metadata only A bi-criteria optimization model to analyze the impacts of electric vehicles on costs and emissions(Elsevier, 2017) N/A; N/A; Department of Industrial Engineering; Kabatepe, Bora; Türkay, Metin; Master Student; Faculty Member; Department of Industrial Engineering; Graduate School of Sciences and Engineering; College of Engineering; N/A; 24956Electric vehicles (EV) are emerging as a mobility solution to reduce emissions in the transportation sector. The studies environmental impact analysis of EVs in the literature are based on the average energy mix or pre-defined generation scenarios and construct policy recommendations with a cost minimization objective. However, the environmental performance of EVs depends on the source of the marginal electricity provided to the grid and single objective models do not provide a thorough analysis on the economic and environmental impacts of EVs. In this paper, these gaps are addressed by a four step methodology that analyzes the effects of EVs under different charging and market penetration scenarios. The methodology includes a bi-criteria optimization model representing the electricity market operations. The results from a real-life case analysis show that EVs decrease costs and emissions significantly compared to conventional vehicles.Publication Metadata only A computational-graph partitioning method for training memory-constrained DNNs(Elsevier, 2021) Wahib, Mohamed; Dikbayir, Doga; Belviranli, Mehmet Esat; N/A; Department of Computer Engineering; Qararyah, Fareed Mohammad; Erten, Didem Unat; PhD Student; Faculty Member; Department of Computer Engineering; Graduate School of Sciences and Engineering; College of Engineering; N/A; 219274Many state-of-the-art Deep Neural Networks (DNNs) have substantial memory requirements. Limited device memory becomes a bottleneck when training those models. We propose ParDNN, an automatic, generic, and non-intrusive partitioning strategy for DNNs that are represented as computational graphs. ParDNN decides a placement of DNN's underlying computational graph operations across multiple devices so that the devices' memory constraints are met and the training time is minimized. ParDNN is completely independent of the deep learning aspects of a DNN. It requires no modification neither at the model nor at the systems level implementation of its operation kernels. ParDNN partitions DNNs having billions of parameters and hundreds of thousands of operations in seconds to few minutes. Our experiments with TensorFlow on 16 GPUs demonstrate efficient training of 5 very large models while achieving superlinear scaling for both the batch size and training throughput. ParDNN either outperforms or qualitatively improves upon the related work.Publication Metadata only A front tracking method for particle-resolved simulation of evaporation and combustion of a fuel droplet(Pergamon-Elsevier Science Ltd, 2018) N/A; N/A; Department of Mechanical Engineering; Irfan, Muhammad; Muradoğlu, Metin; PhD Student; Faculty Member; Department of Mechanical Engineering; Graduate School of Sciences and Engineering; College of Engineering; N/A; 46561A front-tracking method is developed for the particle-resolved simulations of droplet evaporation and combustion in a liquid-gas multiphase system. One field formulation of the governing equations is solved in the whole computational domain by incorporating suitable jump conditions at the interface. Both phases are assumed to be incompressible but the divergence-free velocity condition is modified to account for the phase change at the interface. A temperature gradient based evaporation model is used. An operator-splitting approach is employed to advance temperature and species mass fractions in time. The CHEMKIN package is incorporated into the solver to handle the chemical kinetics. The multiphase flow solver and the evaporation model are first validated using the benchmark problems. The method is then applied to study combustion of a n-heptane droplet using a single-step chemistry model and a reduced chemical kinetics mechanism involving 25-species and 26-reactions. The results are found to be in good agreement with the experimental data and the previous numerical simulations for the time history of the normalized droplet size, the gasification rate, the peak temperature and the ignition delay times. The initial flame diameter and the profile of the flame standoff ratio are also found to be compatible with the results in the literature. The method is finally applied to simulate a burning droplet moving due to gravity at various ambient temperatures and interesting results are observed about the flame blow-off.Publication Metadata only A front-tracking method for computational modeling of impact and spreading of viscous droplets on solid walls(Pergamon-Elsevier Science Ltd, 2010) N/A; Department of Mechanical Engineering; Department of Mechanical Engineering; Muradoğlu, Metin; Taşoğlu, Savaş; Faculty Member; Faculty Member; Department of Mechanical Engineering; College of Engineering; College of Engineering; 46561; 291971A finite-difference/front-tracking method is developed for computational modeling of impact and spreading of a viscous droplet on dry solid walls. The contact angle is specified dynamically using the empirical correlation given by Kistler (1993). The numerical method is general and can treat non-wetting, partially wetting and fully wetting cases but the focus here is placed on the partially wetting substrates. Here the method is implemented for axisymmetric problems but it is straightforward to extend it to three dimensional cases. Grid convergence of the method is demonstrated and the validity of the dynamic contact angle method is examined. The method is first tested for the spreading and relaxation of a droplet from the initial spherical shape to its final equilibrium conditions for various values of Eotvos number. Then it is applied to impact and spreading of glycerin droplets on wax and glass substrates and, the results are compared with experimental data of Sikalo et al. (2005). The numerical results are found in a good agreement with the experimental data. Finally the effects of governing non-dimensional numbers on the spreading rate, apparent contact angle and deformation of the droplet are investigated.Publication Open Access A gated fusion network for dynamic saliency prediction(Institute of Electrical and Electronics Engineers (IEEE), 2022) Kocak, Aysun; Erdem, Erkut; Department of Computer Engineering; Erdem, Aykut; Faculty Member; Department of Computer Engineering; College of Engineering; 20331Predicting saliency in videos is a challenging problem due to complex modeling of interactions between spatial and temporal information, especially when ever-changing, dynamic nature of videos is considered. Recently, researchers have proposed large-scale data sets and models that take advantage of deep learning as a way to understand what is important for video saliency. These approaches, however, learn to combine spatial and temporal features in a static manner and do not adapt themselves much to the changes in the video content. In this article, we introduce the gated fusion network for dynamic saliency (GFSalNet), the first deep saliency model capable of making predictions in a dynamic way via the gated fusion mechanism. Moreover, our model also exploits spatial and channelwise attention within a multiscale architecture that further allows for highly accurate predictions. We evaluate the proposed approach on a number of data sets, and our experimental analysis demonstrates that it outperforms or is highly competitive with the state of the art. Importantly, we show that it has a good generalization ability, and moreover, exploits temporal information more effectively via its adaptive fusion scheme.Publication Open Access A hierarchical solution approach for a multicommodity distribution problem under a special cost structure(Elsevier, 2012) Koca, Esra; Department of Industrial Engineering; Yıldırım, Emre Alper; Faculty Member; Department of Industrial Engineering; College of EngineeringMotivated by the spare parts distribution system of a major automotive manufacturer in Turkey, we consider a multicommodity distribution problem from a central depot to a number of geographically dispersed demand points. The distribution of the items is carried out by a set of identical vehicles. The demand of each demand point can be satisfied by several vehicles and a single vehicle is allowed to serve multiple demand points. For a given vehicle, the cost structure is dictated by the farthest demand point from the depot among all demand points served by that vehicle. The objective is to satisfy the demand of each demand point with the minimum total distribution cost. We present a novel integer linear programming formulation of the problem as a variant of the network design problem. The resulting optimization problem becomes computationally infeasible for real-life problems due to the large number of integer variables. In an attempt to circumvent this disadvantage of using the direct formulation especially for larger problems, we propose a Hierarchical Approach that is aimed at solving the problem in two stages using partial demand aggregation followed by a disaggregation scheme. We study the properties of the solution returned by the Hierarchical Approach. We perform computational studies on a data set adapted from a major automotive manufacturer in Turkey. Our results reveal that the Hierarchical Approach significantly outperforms the direct formulation approach in terms of both the running time and the quality of the resulting solution especially on large instances.Publication Metadata only A limited memory BFGS based unimodular sequence design algorithm for spectrum-aware sensing systems(IEEE-inst Electrical Electronics Engineers inc, 2022) N/A; Savcı, Kubilay; PhD Student; Graduate School of Sciences and Engineering; N/AUnimodular sequences with good correlation and spectral properties are desirable in numerous applications such as active remote sensing and communication systems. therefore, designing sequences with stopband and correlation sidelobe constraints has gained a lot of attention in the last few decades. in this paper, we propose a fast and efficient iterative algorithm to design unimodular and sparse frequency waveforms with low aperiodic/periodic autocorrelation sidelobes and desired stopband properties. in our approach, the bi-objective optimization problem which minimizes both the integrated sidelobe level (ISL) of the autocorrelation function and the power density in the spectral stopbands is first turned into an unconstrained single objective optimization problem and then is treated as a nonlinear large-scale problem. for the solution of the problem, we develop an algorithm based on Limited Memory Broyden-Fletcher-Goldfarb-Shanno (L-BFGS) Quasi-Newton optimization method. Unlike most gradient based algorithms which employ line searches to deduce the step length, owing to L-BFGS method, unit step length is taken as a general rule to avoid the cost of computation at every iteration with very few exceptions. the calculation of gradient is based on Fast Fourier Transform and Hadamard product operations and thus the algorithm is fast and computationally efficient. Moreover, the algorithm is space efficient and its low-memory feature makes it possible to generate long sequences. Several numerical examples are presented to validate the efficacy of the proposed method and to show its superiority over other state-of-art algorithms.Publication Metadata only A new robust consistent hybrid finite-volume/particle method for solving the PDF model equations of turbulent reactive flows(Pergamon-Elsevier Science Ltd, 2014) Department of Mechanical Engineering; Sheikhsarmast, Reza Mokhtarpoor; Türkeri, Hasret; Muradoğlu, Metin; PhD Student; PhD Student; Faculty Member; Department of Mechanical Engineering; Graduate School of Sciences and Engineering; Graduate School of Sciences and Engineering; College of Engineering; N/A; N/A; 46561A new robust hybrid finite-volume (FV)/particle method is developed for solving joint probability density function (JPDF) model equations of statistically stationary turbulent reacting flows. The method is designed to remedy the deficiencies of the hybrid algorithm developed by Muradoglu et al. (1999, 2001). The density-based FV solver in the original hybrid algorithm has been found to be excessively dissipative and yet not very robust. To remedy these deficiencies, a pressure-based PISO algorithm in the open source FV package, OpenFOAM, is used to solve the Favre-averaged mean mass and momentum equations while a particle-based Monte Carlo algorithm is employed to solve the fluctuating velocity-turbulence frequency-compositions JPDF transport equation. The mean density is computed as a particle field and passed to the FV method. Thus the redundancy of the density fields in the original hybrid method is removed making the new hybrid algorithm more consistent at the numerical solution level. The new hybrid algorithm is first applied to simulate non-swirling cold and reacting bluff-body flows. The convergence of the method is demonstrated. In contrast with the original hybrid method, the new hybrid algorithm is very robust with respect to grid refinement and achieves grid convergence without any unphysical vortex shedding in the cold bluff-body flow case. In addition, the results are found to be in good agreement with the earlier PDF calculations and also with the available experimental data. Finally the new hybrid algorithm is successfully applied to simulate the more complicated Sydney swirling bluff-body flame 'SM1'. The method is also very robust for this difficult test case and the results are in good agreement with the available experimental data. In all the cases, the PISO-FV solver is found to be highly resilient to the noise in the mean density field extracted from the particles.Publication Metadata only A novel economic-based scheduling heuristic for computational grids(Sage Publications Ltd, 2007) N/A; Department of Computer Engineering; Sönmez, Ömer Ozan; Gürsoy, Attila; Master Student; Faculty Member; Department of Computer Engineering; Graduate School of Sciences and Engineering; College of Engineering; N/A; 8745In the economic-based computational grids we need effective schedulers not only to minimize the makespan but also to minimize the costs that are spent for the execution of the jobs. in this work, A novel economy driven job scheduling heuristic is proposed and a simulation application is developed by using GridSim toolkit to investigate the performance of the heuristic. the simulation-based experiments demonstrate the effectiveness of the proposed heuristic both in terms of parameter sweep and sequential workflow type of applications.Publication Open Access A preference-based, multi-unit auction for pricing and capacity allocation(Elsevier, 2018) Lessan, Javad; Department of Business Administration; Karabatı, Selçuk; Faculty Member; Department of Business Administration; College of Administrative Sciences and Economics; 38819We study a pricing and allocation problem of a seller of multiple units of a homogeneous item, and present a semi-market mechanism in the form of an iterative ascending-bid auction. The auction elicits buyers' preferences over a set of options offered by the seller, and processes them with a random-priority assignment scheme to address buyers' "fairness" expectations. The auction's termination criterion is derived from a mixed-integer programming formulation of the preference-based capacity allocation problem. We show that the random priority- and preference-based assignment policy is a universally truthful mechanism which can also achieve a Pareto-efficient Nash equilibrium. Computational results demonstrate that the auction mechanism can extract a substantial portion of the centralized system's profit, indicating its effectiveness for a seller who needs to operate under the "fairness" constraint.