Makale Koleksiyonu

Bu koleksiyon için kalıcı URI

https://hdl.handle.net/20.500.12451/267

Listeleniyor 1 - 20 / 100

Optimum cable bonding with pareto optimal and hybrid neural methods to prevent high-voltage cable insulation faults in distributed generation systems
(Multidisciplinary Digital Publishing Institute (MDPI), 2024) Akbal, Bahadır
The high voltage, current and harmonic distortion in high-voltage cable metal sheaths cause cable insulation faults. The SSBLR (Sectional Solid Bonding with Inductance (L) and Resistance) method was designed as a new cable grounding method to prevent insulation faults. SSBLR was optimized using multi-objective optimization (MOP) with the prediction method (PM) to minimize these factors. The Pareto optimal method was used for MOP. The artificial neural network, hybrid artificial neural network and regression methods were used as the PM. When the artificial neural network–genetic algorithm hybrid method was used as the PM, and the genetic algorithm was used as the optimization method, the voltage and current were significantly reduced in the metal sheath of the cable.
Zırhlı yüksek gerilim kablolarında yalıtım arızalarını önlemek için çok amaçlı optimizasyon ve hibrit yapay zekâ tabanlı kablo topraklama yöntemi
(Oğuzhan YILMAZ, 2024) Akbal, Bahadır
Bir yüksek gerilim kablosunda, yalıtkan tabaka üzerinde kullanılan metal kılıf ve zırh elektrik alanını sınırladıkları gibi mekanik etkilere karşı yalıtkanı korurlar. Ancak kablodan yük akımı geçince, zırh ve metal kılıf üzerinde kablo sıcaklığını artıran ve tehlikeli gerilimlere neden olan akım ve gerilimler oluşur. Aşırı kablo sıcaklığı ve tehlikeli gerilimler yalıtım arızalarına neden olmaktadır. Literatürde bu yalıtım arızalarını önlemek için farklı topraklama yöntemleri önerilmektedir. Ancak, son yıllarda harmonik akımlarının da etkisinin artmasından dolayı bu topraklama yöntemleri yalıtım hatalarını önlemek için yetersiz kalmaktadır. Bu çalışmada yüksek harmonikli akım ve gerilimden dolayı oluşan yalıtım arızalarını önlemek için optimizasyon ve yapay zekâ tabanlı yeni bir topraklama yöntemi önerilmektedir. Bu yöntemde, optimum bir topraklama yapabilmek için metal kılıf ve zırh üzerindeki geriliminin, akımının ve akım harmonik bozunumunun bilinmesi gerekir. Dolayısıyla, bu parametre değerlerinin tespiti için hibrit sinir ağları ve regresyon yöntemlerinden oluşan tahmin yöntemleri kullanılmıştır. Hibrit yapay arı kolonisi-yapay sinir ağı (H-YAK) ve gauss proses regresyon (GPR) yöntemleri minimum eğitim hatalarına göre bu gruplar içinden seçilmiş ve optimizasyon algoritmalarında amaç fonksiyonu olarak kullanılmışlardır. Önerilen topraklama yönteminin optimizasyonunda birçok amaç olduğu için çok amaçlı optimizasyon yöntemi kullanılmıştır. Çok amaçlı optimizasyonda, tahmin yöntemi olarak H-YAK kullanıldığında, metal parçalar üzerinde gerilim, akım ve akım harmoniklerinde optimum değerler sağlanmıştır.
Compact laser spectroscopy-based sensor using a transformer-based model for analysis of multiple molecules
(Optica Publishing Group (formerly OSA), 2024) Bayraklı, İsmail; Eken, Enes
Interest in the development of compact sensors that consume low energy is increasing day by day. This study reports, to our knowledge, such a novel sensor system that can analyze multiple molecules simultaneously with high sensitivity under ambient conditions (900 mbar and 300 K). To quantify molecules, a distributed feedback quantum cascade laser (DFB QCL) was combined with a compact multi-pass absorption (mpass) cell without the need for vacuum components, lock-in amplifier, or any electric filters. By using a transformer-encoder-based model, the noise level was reduced and the pressure-broadened absorption lines of the molecules were separated, narrowed (resolved), and displayed one by one. In this way, molecules can be quantified using pressure-broadened overlapping absorption lines under ambient conditions. To test our sensor system, CO2 and N2O molecules were used. Depending on the concentration values, SNR can be improved by up to 50 times. Better results are obtained at higher concentration values. Detection limits for N2O and CO2 molecules were determined to be 30 ppb and 180 ppm, respectively. The analysis time of molecules is around 80 ms.
Optimal fuzzy P + Dµ controller for cancer chemotherapy
(Elsevier Ltd, 2024) Ay, Sena; Soylu, Selim
This paper aims to develop a robust, effective, and optimal controller for an automatic intravenous drug delivery system for the chemotherapy treatment of cancer patients. For this purpose, a novel hybrid control method, namely the Fuzzy P + Dµ controller, is designed by combining fractional calculus and fuzzy logic controller (FLC). Particle swarm optimization (PSO) algorithm is employed to tune the controller parameters optimally. To evaluate the performance of the proposed controller, a cancer patient model that considers cumulative drug toxicity, one of the most common side effects of chemotherapy, is utilized. Limits on toxicity and drug concentration for patient safety are incorporated into the controller design and optimization process. The graphical and numerical simulation results show that almost no cancerous cells are left at the end of the treatment period. The effectiveness of the controller is proven by the performance index (PI) obtained. Our optimal hybrid Fuzzy P + Dµ outclasses other control techniques in previous related studies with the best PI of 44.426 and the minimum final number of cancerous cells (FNCCs) of 5.084E-08. Moreover, in robustness tests perturbing intra-patient parameters and using different treatment protocols, the cumulative drug toxicity level is kept within safe limits, and almost no cancerous cells remain
Design of Optimal FOPI Controller for Two-Area Time-Delayed Load Frequency Control System with Demand Response
(Korean Institute of Electrical Engineers, 2024) Katipoğlu, Deniz; Soylu, Selim
This study comprehensively investigates modeling and controller design for two-area load frequency control (LFC) system with demand response (DR) and time delays. An optimized fractional order proportional-integral (FOPI) controller is designed both to improve the stability of grid frequencies and to deal with the time delays. Using the stability boundary locus (SBL) method, the stability region for the FOPI controller parameter space is obtained. The boundaries of these stability regions are used in the optimization process of the FOPI controller parameters. A new objective function is used to minimize the integral time-weighted absolute error (ITAE) and time domain specifications such as overshoot (OS) and undershoot (US) of the frequency and tie-line power deviations. The parameters of the FOPI controller are optimized by Artificial Hummingbird Algorithm (AHA). The simulation results are compared with gray wolf optimizer (GWO) and particle swarm optimization (PSO) tuned FOPI controllers. The AHA tuned FOPI (AHA-FOPI) controller gives very promising results, especially in terms of improving the transient response of the system by reducing the settling time (ST) and steady-state error (ess). Additionally, the effectiveness of the AHA-FOPI is demonstrated under increased time delay. It is evident from the simulation results that the AHA-FOPI controller has successfully improved the time domain characteristics of the LFC-DR system.
A novel approach for improving the performance of deep learning-based state of charge estimation of lithium-ion batteries: Choosy SoC Estimator (ChoSoCE)
(Elsevier Ltd, 2024) Korkmaz, Mehmet
Deep learning-based (DL) methods have recently come to the forefront among the data-driven models due to their success in capturing the complexities of the battery. Many previous DL-based studies for SoC estimation have almost exclusively focused on improving DL structure by proposing various architectures. Questions regarding the outlier or atypical predictions have yet to be adequately addressed. Furthermore, few works benefit from optimization algorithms to determine the hyperparameters of DL. In this study, we have addressed the problem of how to obtain the hyperparameter of DL and fix the improper DL predictions. To this aim, we used two different optimization algorithms to determine the hyperparameters of DL and proposed a novel algorithm that considers the previous SoC estimations. The algorithm either approves or rejects the DL predictions for the relevant step and offers new values for the rejected ones. The proposed scheme is evaluated using a battery dataset which includes different driving cycles. According to the results, it is observed that the optimized DL outperforms the empirical one by at least 35% in terms of performance indices. Moreover, the proposed novel algorithm successfully integrates into all variations and significantly improves the performance index scores.
Joint Pulse Index and Spatial Modulation
(Institute of Electrical and Electronics Engineers Inc., 2024) Aldırmaz Çolak, Sultan; Aydın, Erdoğan; Gündem, Sümeyra; Çelik, Yasin; Başar, Ertuğrul
According to the planned key performance indicator (KPI) standards, 6G technology should achieve higher throughput than 5G. More efficiency in transceiver schemes is required to meet this demand. In this study, we take advantage of spatial modulation (SM) and pulse index modulation (PIM) techniques to increase spectral efficiency. The proposed PIM-SM scheme utilizes well-localized and orthogonal Hermite-Gaussian pulses along with spatial indexing. Thanks to the orthogonality between pulses in the set, multiple pulses are transmitted together. The design, simulation, and analytical bit error rate performance derivations of PIM-SM are discussed in this letter to verify the viability and compatibility of pulse-based data transfer utilizing the spatial domain. The performance is compared with generalized code index modulation-spatial modulation (GCIM-SM), code index modulation-quadrature spatial modulation (CIM-QSM), and classical spatial modulation (SM) schemes.
Dinamik talep cevabı içeren zaman gecikmeli iki bölgeli yük frekans kontrol sistemlerinin kararlılık bölgelerinin hesaplanması
(Gazi Üniversitesi, 2024) Katipoğlu, Deniz; Sönmez, Şahin; Ayasun, Saffet
Bu çalışmada, dinamik talep cevabı (DTC) ve haberleşme zaman gecikmesi içeren iki bölgeli yük frekans kontrol (YFK-DTC) sisteminin kararlılık sınır eğrisi yöntemi kullanılarak denetleyici parametre düzleminde kararlılık bölgeleri hesaplanmıştır. DTC kontrol, kontrol edilebilir yük gruplarını frekans kontrol servisine dahil ederek, üretim ve puant yük talebi arasında dengenin daha kısa sürede sağlanması ve yenilenebilir enerji kaynaklarında güç dengesizlikleri problemlerine karşı önemli bir çözüm sunmaktadır. DTC kontrol mekanizmasının yük frekans kontrol sistemlerinde kullanımı, sistemin güvenliği ve güvenilirliğini sağlamasına rağmen, haberleşme ağlarından kaynaklanan zaman gecikmeleri, denetleyici performansını ve sistemin kararlılığını olumsuz etkileyebilmektedir. Dolayısıyla, bu çalışma zaman gecikmesi içeren iki bölgeli YFK-DTC sisteminin kararlılığını garanti edecek tüm oransal-integral (PI) denetleyici kazanç değerlerini elde etmektedir. Bu amaçla, zaman gecikmeli YFK-DTC sisteminin denetleyici parametre düzleminde kararlılık bölgelerini oluşturan kompleks kök sınır (Complex Root Boundary, CRB) eğrisini ve reel kök sınır (Real Root Boundary, RRB) eğrisini bulmak için kararlılık sınır eğrsi yöntemi kullanılmıştır. Elde edilen teorik sonuçların doğruluğu, quasi-polynomial mapping root (QPmR) algoritması ve zaman düzleminde yapılan benzetim çalışmaları ile gösterilmiştir. Sonuçlar, DTC kontrol çevriminin katkısı ile zaman gecikmeli YFK sisteminin kararlılık bölgelerinin ve kararlılık payının arttığını göstermektedir.
Nickel-coated carbon fillers for polyurethane foam with improved microwave absorption performance: A comparative analysis
(Wiley, 2024) Kurt, Gökçe; Korkmaz, Mehmet
In this study, polyurethane foam (PUF) composites were prepared by incorporating different functional fillers, and morphological, rheological, dielectric, and microwave absorption properties of these composite samples were investigated. Graphite (G), nickel (Ni), nickel-coated hybrid carbon (NiC), and iron II-III oxide (Fe3O4) powders were used as functional fillers, while rigid polyurethane foam (PUF) with a 40 kg/m3 density and a closed-cell structure was used as the carrier foam phase. Morphological analyses that were carried out by scanning electron microscopy showed that all fillers were dispersed between the cell walls. In dielectric characterization, the highest electrical permittivity values were obtained by incorporation of G, while the highest magnetic permeability values were obtained by incorporation of Fe3O4 at the same filler concentration. Furthermore, the PUF composites could be prepared using NiC at a higher filler concentration than the others due to the lower viscosity increasing effect of NiC. The highest microwave absorption performance was obtained in the NiC-filled PUF composite at a filler concentration of 120 phr. The minimum reflection loss (RL) value of -45.2 dB at 10.62 GHz and an RL value lower than -10 dB in the entire X-band region were obtained in the NiC-filled PUF composite.HighlightsNickel-coated carbon (NiC) filled polyurethane foam (PUF) was prepared.The absorption performance of NiC was compared to conventional filler.PUF composites containing NiC showed lower viscosity and processability.The highest absorption could be obtained by NiC compared to conventional fillers.
Computation of stability regions for time-delayed two-area load frequency control system including dynamic demand response
(Gazi Üniversitesi, 2024) Katipoğlu, Deniz; Sönmez, Şahin; Ayasun, Saffet
This study focuses on the computation of stability regions of the time-delayed two-area load frequency control including dynamic demand response (LFC-DDR) using stability boundary locus method. With the participation of controllable responsive loads into the frequency regulation service, dynamic demand response (DDR) has become an important solution for proper balancing between generation and peak load and to overcome the intermittent nature of renewable power generations. Although the utilization of DDR control technique increases the reliability and security of the load frequency control (LFC) systems, communication time delays because of the communication networks adversely affect the controller performance and LFC system stability. Therefore, this study obtains the all stabilizing proportional-integral (PI) controller gains that guarantee the stability of the LFC-DDR system. For that purpose, stability boundary locus method is used to obtain stability regions in the controller parameters space that constitute of complex root boundary (CRB) and real root boundary (RRB) loci of the time-delayed LFC-DDR systems. The accuracy of theoretical results is verfied by an independent algoirithm, quasi-polynomial mapping root (QPmR) finder algorithm, and time-domain simulations. Results indicate that the participation of DDR control loop increases the stability regions and stability margin of the LFC system even in the presence of communication time delays.
Multi-objective optimization and hybrid AI-based cable grounding method to prevent insulation failures in armored high voltage cables
(Gazi Universitesi, 2024) Akbal, Bahadır
High voltage cable insulation fault is a major problem in high voltage lines. Armour current, armour voltage and current harmonics are major causes of the insulation faults in high voltage underground cables. These factors are major factors for high voltage insulation faults. Thus, cable bonding methods are used to prevent the cable insulation faults. The used methods in the literature do not prevent insulation faults that are based on armour harmonic current. In this study, a new cable bonding method is developed with multi-objective optimization and hybrid intelligence algorithm to prevent cable insulation faults. The proposed bonding method is shown in Figure A, and the parameters of these method is optimized with multi-objective optimization and hybrid intelligence algorithms. (Figure Presented) Purpose: The aim of this study is to prevent the cable insulation fault based on armour current harmonic and voltage by using intelligence algorithms. Theory and Methods: Metallic sheath voltage should be known for optimization of the proposed method. Thus, the forecasting methods are used to determine the sheath voltage. Hybrid artificial neural networks and regression methods are used to forecast sheath voltage. Also, multi objective optimization methods are used because there are many objective functions for optimization of the proposed bonding method. Artificial bee colony, genetic algorithm, gravitational search algorithm and particle swarm optimization are used as optimization methods, and the forecasting methods are used as objective function in optimization algorithms. Results: When H-ABC method is used as objective function for optimization of the proposed method, armour current, armour voltage and current harmonics values decrease according to solid bonding method that is used in the literature. Also, armour voltage is lower than touch voltage limit, and Armour current is lower than armour ampacity of the high voltage cable. Conclusion: Since armour current does not exceed armour ampacity, temperature of the insulation layer does not increase. Hence, insulation faults do not occur in the high voltage cable terminations. Also, since armour voltage is lower than touch voltage limit, electroshock for person is prevented by the suggested method.
Retraction: Metabolism Determination By Soft Computing Methods From Breath Molecules
(Sakarya Üniversitesi, 2023) Metlek, Sedat; Akman, Hatice; Bayraklı, İsmail
The article whose title is given above was retracted by the decision of the Editorial Board upon the request of the authors.
Sensor using a photo-acoustic absorption cell with two perpendicular acoustic resonators to analyze multiple molecules
(Optica Publishing Group (formerly OSA), 2023) Bayraklı, İsmail; Akman, Hatice; Sarı, Filiz
An ultra-high sensitivity multi-molecule sensor based on a photo-acoustic cell with two perpendicular acoustic resonators and a common microphone has been reported. In this work, a 4.5 ?m distributed-feedback quantum cascade laser and a 1.5 ?mexternal cavity diode laser (EC-DL)were used as optical excitation sources. Considering the spectral ranges of the lasers used, it is possible to analyze eight molecules (QCL V N2Oand CO2, EC-DL:H2O, H2S, NH3, CO, CH4, and C2H2). The N2O molecule was used to evaluate the performance of the photo-acoustic spectroscopy (PAS)-based sensor. A sensitivity of 0.073 V/ppm and a linearity of 0.99 were found by analyzing the PAS signal as a function of N2O concentration at 2237.656 cm-1. The long-term performance of the sensor was determined by performing an Allan deviation analysis. A minimum detection limit of 9.8 ppb for 90 s integration time was achieved. The simultaneous multi-trace gas detection capability was verified by measurement of N2O, CO2, and H2O. Depending on the coarse/fine-tuning ranges of the lasers used, the number of molecules analyzed can be further increased. Such a sensor could provide simultaneous diagnosis of many diseases through an analysis of breath air and simultaneous monitoring of the most important greenhouse gases.
Stability Analysis Using Fractional-Order PI Controller in a Time-Delayed Single-Area Load Frequency Control System with Demand Response
(Universitatea "Stefan cel Mare" din Suceava, 2023) Katipoğlu, Deniz
The current study investigates the stability analysis based on gain and phase margin (GPM) using fractional-order proportional-integral (FOPI) controller in a time-delayed single-area load frequency control (LFC) system with demand response (DR). The DR control loop is introduced into the classical LFC system to improve the frequency deviation. Although the DR enhances the system’s reliability, the excessive use of open communication networks in the control of the LFC results in time delays that make the system unstable. A frequency-domain approach is proposed to compute the time delay that destabilizes the system using GPM values and different parameter values of the FOPI controller. This method converts the equation into an ordinary polynomial with no exponential terms by eliminating the exponential terms from the system’s characteristic equation. The maximum timedelay values at which the system is marginally stable are calculated analytically using the new polynomial. Finally, the verification of the time delays calculated is demonstrated by simulation studies in the Matlab/Simulink environment and the root finder (quasi-polynomial mapping-based root finder, QPmR) algorithm to define the roots of polynomials with exponential terms providing information about their locations.
SoC estimation of lithium-ion batteries based on machine learning techniques: A filtered approach
(Elsevier Ltd, 2023) Korkmaz, Mehmet
Accurate state-of-charge (SoC) estimation is an essential requirement for many situations where Li-Ion batteries (LiBs) are used. This ensures an efficient battery management system (BMS), so the battery can be protected from excessive discharge, and its life span can be maximized. But when it comes to electrified vehicles (xEVs), the SoC estimation accuracy becomes a more critical and indispensable prerequisite. Because the safety of xEVs during driving and the remaining range, which is an indicator of how far the vehicle can go, are directly related to the accurate SoC. However, the complex electrochemical reactions in the battery and the dependence on environmental variables make SoC estimation a challenging task. Traditionally, this is tackled by establishing either electrochemical or electrical battery equivalent models. Both methods suffer from some limitations, such as parameter identification, complex calculations, and model mismatching due to the aging factor. On the other hand, data-driven methods have recently become a popular choice for SoC estimation since they enable building data-based models rather than chemical reactions or equivalent circuit calculations. The model is built based on battery parameters such as current, voltage, battery type, and then used for SoC estimation. However, many studies in the literature examine only a few methods for SoC estimation. Also, these data-driven black box models can lead to outlier data as they are not observers. Thus, the aims of this study are twofold: First, to make a comprehensive comparison based on most of the ML methods. Second, to utilize several filters for outlier removal and measure their effectiveness. For these purposes,18 ML algorithms were handled in three main groups, and SoC estimation results were analyzed. Additionally, five different filters were used to improve the SoC estimation of these methods, and their comparisons were realized. From the results, it is clear that Bagging and ExtraTree algorithms are substantially better than other ML methods for SoC estimation since their Interquartile Range (IQR) is smaller than 3%, performance indices are the lowest ones, and curve matches are the best. Also, Rloess is the best filter among the others, although they all achieved high performance in outlier removal.
Modeling, analysis and simulation of a high-efficiency battery control system
(Tech Science Press, 2023) Alkhafaji, Mohammed Ayad; Uzun, Yunus
This paper explains step-by-step modeling and simulation of the full circuits of a battery control system and connected together starting from the AC input source to the battery control and storage system. The three-phase half-controlled rectifier has been designed to control and convert the AC power into DC power. In addition, two types of direct current converters have been used in this paper which are a buck and bidirectional DC/DC converters. These systems adjust the output voltage to be lower or higher than the input voltage. In the buck converters, the main switch operates in conduction or cut-off mode and is triggered by a Pulse-Width Modulated (PWM) signal. The output and input voltage levels ratio are used to calculate the PWM signal’s duty cycle. Therefore, the duty cycle indicates the operation mode of the converter in steady-state operation. In this study, we analyze and control of a buck converter with the PWM signal. Besides, the bidirectional DC/DC converter has been achieved and optimized by PI control methods to control the battery charging and discharging modes. The simulation has been applied via the Matlab/Simulink environment. The results show the activity of each part of the designed circuits starting from the converters and the battery control system in charge and discharge modes.
Design and comprehensive analyzes of a highly efficient TLA-Type synchronous reluctance machine including the effects of conductor per slot and wire size
(MDPI, 2023) Özdil, Ali; Uzun, Yunus
Consuming energy sources and greenhouse gas emission are one of the most prominent problems of the latest century. Most of the energy consumed globally and carbon dioxide emissions originate from electric motors used in the industry. Therefore, researchers have recently focused on the production of highly efficient, eco-friendly, and low-priced machines: Synchronous Reluctance Machines. In this study, the design and comprehensive Finite Element Analysis of a TLA-SynRM including the effects of the number of conductors per slot and wire diameter directly affecting stator slot fill factor and crucial for obtaining more realistic results from experiments has been initially carried out. Moreover, power factor, saliency ratio, and efficiency of the novel SynRM are enhanced by utilizing a fine-tuning process based on d- and q-axes flux paths. Additionally, the layer structure of the initial design is changed to a double-layer structure to improve the performance of the machine in the fine-tuning process. In the final step of this study, the machine has been manufactured, and experiments have been accomplished. This study has concluded that the novel SynRM have low torque ripple, high power factor, saliency ratio, and efficiency, whose value is within the range of IE5 efficiency class.
A novel breath molecule sensing system based on deep neural network employing multiple-line direct absorption spectroscopy
(Elsevier Ltd, 2023) Bayraklı, İsmail; Eken, Enes
A novel ppb-level biomedical sensor is developed to analyze breath samples for continuous monitoring of diseases. The setup is very compact, consisting of a distributed feedback quantum cascade laser (DFB-QCL) and a single-pass absorption cell. To make the sensor more compact and functional, a deep neural network (DNN) model is utilized for predicting gas concentrations. In order to evaluate the performance of the sensor, N2O is used as the target molecule. A minimum detection limit of 500 ppb is achieved in a single-pass absorption cell configuration. The model is trained on multiple N2O/CO2 absorption lines (instead of an isolated line) with concentrations between 0 to 500 ppm generated using the HITRAN database. The trained model is tested on measured spectra and compared to a non-linear least squares fitting algorithm. The coefficients of determination (R2) were found to be 0.997 and 0.981 for the predictions of N2O concentrations in the N2O/N2 gas mixture and the breath air, respectively. The accuracies of 2.5% and 2.9% were achieved by the sensor for both cases.
High accuracy gender determination using the egg shape index
(Nature Research, 2023) Kayadan, Muhammed; Uzun, Yunus
Since only female chicks are used in layer hens, usually hatched male chicks are killed. It is estimated that around 7 billion chicks per year are killed immediately after hatching. In addition to being unethical, this situation also causes great financial losses. Sex determination in chicks can be done before or after hatching. Of course, determinations made before hatching are more advantageous, but the prediction rate is relatively low. The morphology of an egg is expressed in terms of the Shape Index (SI), which is the ratio of the short diameter to the long diameter. In this study, male and female chicks were predicted by using the shape index of the eggs using the RUSBoost Classifier using Shape Index. Although SI varied according to the egg type, a significant correlation (r = 0.78) was observed between chick sex and SI. Therefore, it was possible to estimate gender by utilizing SI in chickens, even if the accuracy of classification was not as high as in ducks. Besides the SI, mass, short axis, long axis, ovality, volume, eccentricity parameters were obtained and used for the results. With this features, females classified with 80% and males classified 81% correctly. The model predictions were applied to the probability of female chick hatching equation from the previous studies, 71% of the estimations were correctly classified according to this equation.With this work, around 80% of accurate predictions were made. In this case, killing 5.65 billion chicks can be prevented. Likewise, many eggs are not wasted. 1.13 billion USD loss can be prevented.
Content loss and conditional space relationship in conditional generative adversarial networks
(TÜBİTAK (Scientific and Technological Research Council of Turkey), 2022) Eken, Enes
In the machine learning community, generative models, especially generative adversarial networks (GANs) continue to be an attractive yet challenging research topic. Right after the invention of GAN, many GAN models have been proposed by the researchers with the same goal: creating better images. The first and foremost feature that a GAN model should have is that creating realistic images that cannot be distinguished from genuine ones. A large portion of the GAN models proposed to this end have a common approach which can be defined as factoring the image generation process into multiple states for decomposing the difficult task into several more manageable sub tasks. This can be realized by using sequential conditional/unconditional generators. Although images generated by sequential generators experimentally prove the effectiveness of this approach, visually inspecting the generated images are far away of being objective and it is not yet quantitatively showed in an objective manner. In this paper, we quantitatively show the effectiveness of shrinking the conditional space by using the sequential generators instead of utilizing single but large generator. At the light of the content loss we demonstrate that in sequential designs, each generator helps to shrink the conditional space, and therefore reduces the loss and the uncertainties at the generated images. In order to quantitatively validate this approach, we tried different combinations of connecting generators sequentially and/or increasing the capacity of generators and using single or multiple discriminators under four different scenarios applied to image-to-image translation tasks. Scenario-1 uses the conventional pix2pix GAN model which serves as the based line model for the rest of the scenarios. In Scenario-2, we utilized two generators connected sequentially. Each generator is identical to the one used in Scenario-1. Another possibility is just doubling the size of a single generator which is evaluated in the Scenario-3. In the last scenario, we used two different discriminators in order to train two sequentially connected generators. Our quantitative results support that simply increasing the capacity of one generator, instead of using sequential generators, does not help a lot to reduce the content loss which is used in addition to adversarial loss and hence does not create better images.

Güncel Gönderiler