Faculty of Engineering

Permanent URI for this communityhttps://hdl.handle.net/10679/10

Browse

Now showing 1 - 7 of 7

Open Access
Accelerating discovery of COFs for CO2 capture and H2 purification using structurally guided computational screening
(Elsevier, 2022-01-01) Aksu, G. O.; Fındıkçı, İlknur Eruçar; Haslak, Z. P.; Keskin, S.; Mechanical Engineering; FINDIKÇI, Ilknur Eruçar
Screening of hypothetical covalent organic framework (hypoCOF) database enables to go beyond the current synthesized structures to design high-performance materials for CO2 separation. In this work, we followed a structurally guided computational screening approach to find the most promising candidates of hypoCOF adsorbents and membranes for CO2 capture and H2 purification. Grand canonical Monte Carlo (GCMC) simulations were used to evaluate CO2/H2 separation performance of 3184 hypoCOFs for pressure-swing adsorption (PSA) and vacuum-swing adsorption (VSA) processes. CO2/H2 adsorption selectivities and CO2 working capacities of hypoCOFs were calculated in the range of 6.13–742 (6.39–954) and 0.07–8.68 mol/kg (0.01–3.92 mol/kg), achieving higher values than those of experimentally synthesized COFs at PSA (VSA) conditions. Density functional theory (DFT) calculations revealed that the strength of hydrogen bonding between CO2 and the functional group of linkers is an important factor for determining the CO2 selectivity of hypoCOFs. The most predominant topologies and linker types were identified as bor and pts, linker91 (a triazine linker) and linker92 (a benzene linker) for the top-performing hypoCOF adsorbents, respectively. Molecular dynamics (MD) simulations of 794 hypoCOFs showed that they exceed the Robeson's upper bound by outperforming COF, zeolite, metal organic framework (MOF), and polymer membranes due to their high H2/CO2 selectivities, 2.66–6.14, and high H2 permeabilities, 9×105–4.5×106 Barrer. Results of this work will be useful to guide the synthesis of novel materials by providing molecular-level insights into the structural features of hypothetical COFs to achieve superior CO2 separation performance.
Open Access
Deep reinforcement learning approach for trading automation in the stock market
(IEEE, 2022) Kabbani, Taylan; Duman, Ekrem; Industrial Engineering; DUMAN, Ekrem; Kabbani, Taylan
Deep Reinforcement Learning (DRL) algorithms can scale to previously intractable problems. The automation of profit generation in the stock market is possible using DRL, by combining the financial assets price 'prediction' step and the 'allocation' step of the portfolio in one unified process to produce fully autonomous systems capable of interacting with their environment to make optimal decisions through trial and error. This work represents a DRL model to generate profitable trades in the stock market, effectively overcoming the limitations of supervised learning approaches. We formulate the trading problem as a Partially Observed Markov Decision Process (POMDP) model, considering the constraints imposed by the stock market, such as liquidity and transaction costs. We then solve the formulated POMDP problem using the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm reporting a 2.68 Sharpe Ratio on unseen data set (test data). From the point of view of stock market forecasting and the intelligent decision-making mechanism, this paper demonstrates the superiority of DRL in financial markets over other types of machine learning and proves its credibility and advantages in strategic decision-making.
Open Access
Experimental and numerical modal characterization for additively manufactured triply periodic minimal surface lattice structures: Comparison between free-size and homogenization-based optimization methods
(Wiley, 2023-06) Özdemir, Mirhan; Simsek, U.; Kuşer, Engin; Gayir, C. E.; Celik, A.; Şendur, Polat; Mechanical Engineering; ŞENDUR, Polat; Özdemir, Mirhan; Kuşer, Engin
Homogenization-based topology optimization (HMTO) is one of the most extensively used grading methods to generate functionally graded lattice structures (FGLs). However, it requires a precharacterization of the lattices, which is time-consuming. As a remedy, free-size optimization-based graded lattice generation (FOGLG) is explored as an alternative method to generate the FGLs. This article builds on the authors’ previous work in which the HMTO and FOGLG approaches are studied to improve the dynamic characteristic of a design by using a single lattice type, namely, double gyroid (DG) structure. To show applicability of the proposed methods, different lattice types including diamond (D), gyroid (G), and I-WP are employed to create FGLs herein. The frequency response analysis is performed, and the results from HMTO and FOGLG are compared in terms of their accuracy and efficiency. The optimized designs are then reconstructed by relative density mapping (RDM) and enhanced relative density mapping (ERDM) methods. The fabricated test samples made of cobalt–chromium using the direct metal laser melting (DMLM) technique are then experimentally validated using a laser vibrometer. The results reveal that HMTO and FOGLG can be used on the lattice types with a variety of configurations and relative densities.
Open Access
Exploring the performance limits of MOF/polymer MMMs for O2/N2 separation using computational screening
(Elsevier, 2021-01-15) Dağlar, H.; Fındıkçı, İlknur Eruçar; Keskin, S.; Mechanical Engineering; FINDIKÇI, Ilknur Eruçar
Air separation is one of the most challenging separations because of the very similar molecular dimensions of gas molecules. We used a high-throughput computational screening approach to identify the upper performance limits of metal organic framework (MOF) membranes and MOF/polymer mixed matrix membranes (MMMs) for O2/N2 separation. Gas permeabilities and selectivities were calculated for 5629 MOF membranes and 78,806 different types of MOF/polymer MMMs, which represent the largest number of MOF-based membranes studied to date for air separation. Our results showed that many MOF membranes exceed the upper bound established for polymer membranes due to their high permeabilities and/or selectivities. The maximum achievable O2 permeability and O2/N2 selectivity of MOF/polymer MMMs were computed as 2710.8 Barrer and 19.8, respectively. Results revealed that MOF/polymer MMMs can outperform MMMs composed of traditional fillers, such as zeolites, in terms of O2 permeability and O2/N2 selectivity. The impacts of purity of air mixture and the structural flexibility of MOFs on the gas separation performances of MMMs were also discussed. These results provide molecular-level insights into adsorption and diffusion behaviors of O2 and N2 in MOF membranes in addition to presenting structure-performance relations of MOFs that can lead to high-performance membranes and fillers for MMMs.
Open Access
Simulation of vehicles’ gap acceptance decision at unsignalized intersections using SUMO
(Elsevier, 2022) Bagheri, Mohammad; Bartın, Bekir Oğuz; Ozbay, K.; Civil Engineering; BARTIN, Bekir Oğuz; Bagheri, Mohammad
In this paper, an artificial neural network (ANN)-based gap acceptance behavior model was proposed. The feasibility of implementing this model in a microscopic simulation tool was tested using the application programming interface of Simulation of Urban Mobility (SUMO) simulation package. A stop-controlled intersection in New Jersey was selected as a case study. The simulation model of this intersection was calibrated using ground truth data extracted during the afternoon peak hours. The ANN-based SUMO model was compared to SUMO model with default gap acceptance parameters and the SUMO model with calibrated gap acceptance parameters. The comparison was based on wait time and accepted gap values at the minor approach of the intersection. The results showed that the ANN-based model produced superior results based on the selected outputs. The analysis results also indicated that the ANN-based model leads to significantly more realistic driving behavior of vehicles on the major approach of the intersection.
Open Access
Stochastic production planning with flexible manufacturing systems and uncertain demand: A column generation-based approach
(Elsevier, 2022) Elyasi, Milad; Özener, Başak Altan; Ekici, Ali; Özener, Okan Örsan; Yanıkoğlu, İhsan; Economics; Industrial Engineering; ÖZENER, Başak Altan; EKİCİ, Ali; ÖZENER, Okan Örsan
The ongoing pandemic, namely COVID-19, has rendered widespread economic disorder. The deficiencies have delayed production at manufacturers in several industries on the supply side. The effects of disruption were more notable for industries with longer supply chains, especially reaching East Asia. Regarding the demand, sectors can be divided into three categories: i) the ones, like e-commerce companies, that experienced augmented demand; ii) the ones with a plunged demand, like what hotels and restaurants experience; iii) the companies experiencing a roller-coaster-ride business. After mitigation efforts, the economy started recovering, resulting in increased demand. However, regardless of their struggles, the companies have not fully returned to their pre-pandemic levels. One of the strategies to gain resilience in its supply chain and manage the disruptions is to employ flexible/hybrid manufacturing systems. This paper considers a flexible/hybrid manufacturing production setting with typically dedicated machinery to satisfy regular demand and a flexible manufacturing system (FMS) to handle surge demand. We model the uncertainty in demand using a scenario-based approach and allow the business to make here-and-now and wait-and-see decisions exploiting the cost-effectiveness of the standard production and responsiveness of the FMS. We propose a column generation-based algorithm as the solution approach. Our computational analysis shows that this hybrid production setting provides highly robust response to the uncertainty in demand, even with high fluctuations.
Open Access
Symbolic knowledge extraction for explainable nutritional recommenders
(Elsevier, 2023-06) Magnini, M.; Ciatto, G.; Cantürk, Furkan; Aydoğan, Reyhan; Omicini, A.; Computer Science; AYDOĞAN, Reyhan; Cantürk, Furkan
Background and objective: This paper focuses on nutritional recommendation systems (RS), i.e. AI-powered automatic systems providing users with suggestions about what to eat to pursue their weight/body shape goals. A trade-off among (potentially) conflictual requirements must be taken into account when designing these kinds of systems, there including: (i) adherence to experts’ prescriptions, (ii) adherence to users’ tastes and preferences, (iii) explainability of the whole recommendation process. Accordingly, in this paper we propose a novel approach to the engineering of nutritional RS, combining machine learning and symbolic knowledge extraction to profile users—hence harmonising the aforementioned requirements. MethodsOur contribution focuses on the data processing workflow. Stemming from neural networks (NN) trained to predict user preferences, we use CART Breiman et al.(1984) to extract symbolic rules in Prolog Körner et al.(2022) form, and we combine them with expert prescriptions brought in similar form. We can then query the resulting symbolic knowledge base via logic solvers, to draw explainable recommendations. ResultsExperiments are performed involving a publicly available dataset of 45,723 recipes, plus 12 synthetic datasets about as many imaginary users, and 6 experts’ prescriptions. Fully-connected 4-layered NN are trained on those datasets, reaching ∼86% test-set accuracy, on average. Extracted rules, in turn, have ∼80% fidelity w.r.t. those NN. The resulting recommendation system has a test-set precision of ∼74%. The symbolic approach makes it possible to devise how the system draws recommendations. Conclusions Thanks to our approach, intelligent agents may learn users’ preferences from data, convert them into symbolic form, and extend them with experts’ goal-directed prescriptions. The resulting recommendations are then simultaneously acceptable for the end user and adequate under a nutritional perspective, while the whole process of recommendation generation is made explainable.

Browse

Browsing by Rights "Attribution-NonCommercial-NoDerivatives 4.0 International"