Article
Physics, Particles & Fields
Laurits Tani, Diana Rand, Christian Veelken, Mario Kadastik
Summary: Analyzing vast amounts of data in modern high energy physics experiments is a challenge, often requiring the use of machine learning methods trained on simulated data. Choosing the right parameters for the machine learning algorithm is crucial for performance optimization.
EUROPEAN PHYSICAL JOURNAL C
(2021)
Article
Astronomy & Astrophysics
Isidro Gomez-Vargas, Joshua Briones Andrade, J. Alberto Vazquez
Summary: The applications of artificial neural networks in cosmology have been successful due to their ability to model large datasets and nonlinear functions. However, their use can be controversial due to the potential for inaccurate results when hyperparameters are not carefully selected. In this paper, the use of genetic algorithms to find optimal hyperparameter combinations for neural networks is proposed and tested on three different cosmological cases, showing improved performance compared to a standard grid method.
Article
Computer Science, Information Systems
Israa Al-Badarneh, Maria Habib, Ibrahim Aljarah, Hossam Faris
Summary: This paper introduces three stochastic and metaheuristic algorithms to train MLP neural network for solving the problem of imbalanced classifications. The algorithms are evaluated using accuracy, F-score, and G-mean, and the results show that F-score and G-mean are more advantageous when the datasets are imbalanced.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2022)
Article
Chemistry, Physical
Bartosz Miller, Leonard Ziemianski
Summary: This study investigates the application of surrogate models in multi-objective optimization of composite shells. By using deep neural networks to approximate the relationships between input parameters and key metrics, incorporating mode shape identification, and employing network ensembles, the efficiency and accuracy of the optimization process are enhanced.
Article
Engineering, Multidisciplinary
Zhen-Jie Guan, Rui Li, Jian-Tang Jiang, Bo Song, Yuan-Xun Gong, Liang Zhen
Summary: In this study, a method for predicting electromagnetic properties of coatings based on GA-ANN algorithm was proposed, which showed lower error and higher accuracy compared to conventional methods. The optimized mixture coating with specific particles composition achieved an effective absorption bandwidth of 9.3 GHz.
COMPOSITES PART B-ENGINEERING
(2021)
Article
Construction & Building Technology
Cheng Lin, Yunting Lin
Summary: In this study, the water cycle algorithm (WCA) is used to analyze and predict the thermal energy demand (TDA) and weighted average discomfort degree-hours (DDA) for a residential building. The connections between TDA and DDA with the geometry and architecture of the building are established using a double-target multi-layer perceptron (2TMLP) model. The results show high goodness-of-fit for both TDA and DDA, with 98.67% and 99.74% correlation, respectively. The WCA-2TMLP model outperforms other hybrid models in terms of prediction accuracy, although the shuffled complex evolution (SCE) optimizer has a better convergence rate. The final mathematical equation of the SCE-2TMLP model is derived for direct prediction of TDA and DDA.
Article
Construction & Building Technology
Rodrigo Polo-Mendoza, Gilberto Martinez-Arguelles, Rita Penabaena-Niebles, Elvis Covilla-Valera
Summary: Compared to traditional Hot Mix Asphalt (HMA), Warm Mix Asphalt (WMA) with Recycled Concrete Aggregate (RCA) contents can be produced at lower temperatures, reducing natural aggregates consumption. However, the benefits may be offset by the higher asphalt binder demand. This research develops a computational model using Artificial Neural Networks (ANNs) to optimize the WMA-RCA design.
ROAD MATERIALS AND PAVEMENT DESIGN
(2023)
Article
Agriculture, Dairy & Animal Science
Gustavo A. Quintana-Ospina, Maria C. Alfaro-Wisaquillo, Edgar O. Oviedo-Rondon, Juan R. Ruiz-Ramirez, Luis C. Bernal-Arango, Gustavo D. Martinez-Bernal
Summary: This study evaluated the impact of temperature, relative humidity, thermal humidity index, management, and farm-associated factors on the performance of broilers raised under commercial tropical conditions. The results showed that temperature significantly affected the weight, weight gain, feed conversion ratio, and mortality of the broilers. Farm-associated factors also had an impact on the performance of the broilers.
Article
Energy & Fuels
Hugo T. V. Gouveia, Murilo A. Souza, Aida A. Ferreira, Jonata C. de Albuquerque, Otoni Nobrega Neto, Milde Maria da Silva Lira, Ronaldo R. B. de Aquino
Summary: In this research, an evolutionary algorithm called RCDESIGN-AS was developed to optimize the ESN-AS for wind speed forecasting. The method is efficient and has shown significant improvement in prediction accuracy compared to traditional methods.
Article
Computer Science, Artificial Intelligence
Hamit Taner Unal, Fatih Basciftci
Summary: We introduce a new neural architecture called Neural Logic Circuits (NLC), which is an evolutionary, weightless, and learnable model inspired by the brain's neuroplasticity. This model achieves learning through the evolution of its architecture by reorganizing synaptic connections and generating artificial neurons as logic gates. Unlike Artificial Neural Networks (ANN), our model achieves generalization ability without intensive weight training and focuses computational resources on building optimal network architecture. Experimental results show the remarkable superiority of our initial model, NLCv1, on well-known binary classification datasets compared to modern and competitive machine learning algorithms.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Construction & Building Technology
Yanni Bouras, Le Li
Summary: This paper develops a creep compliance prediction model for concretes containing fly ash and slag using supervised machine learning techniques. The best-performing models were found to be random forest regression (RFR) and Gaussian process regression (GPR). The importance of input variables was analyzed and time was identified as the most significant parameter. Experimental data showed that the trained models accurately reflected creep behavior, with the exception of the decision tree regression (DTR) model.
Article
Computer Science, Information Systems
Nebojsa Bacanin, Timea Bezdan, K. Venkatachalam, Miodrag Zivkovic, Ivana Strumberger, Mohamed Abouhawwash, Abeer B. Ahmed
Summary: The study proposes an enhanced artificial bee colony optimization algorithm for optimizing connection weights and hidden units of artificial neural networks. Through testing and comparison, the results show that the algorithm outperforms other metaheuristics in terms of accuracy and convergence speed. The improved learning mechanism significantly enhances the convergence speed of the original algorithm and the exploitation capability is enhanced, resulting in significantly better accuracy.
Article
Computer Science, Hardware & Architecture
Erol Egrioglu, Crina Grosan, Eren Bas
Summary: In this study, a new genetic algorithm with a statistical-based chromosome replacement strategy is proposed and applied in the training process of a multiplicative neuron model artificial neural network. Results show that this algorithm outperforms other artificial intelligence optimization methods in time-series prediction tasks.
JOURNAL OF SUPERCOMPUTING
(2023)
Article
Energy & Fuels
Minwoong Kang, Stefan Elbel
Summary: This study proposes a new type of regenerator with high heat transfer rate for the caloric cycle using additive manufacturing. The regenerator is optimized through an artificial neural network - genetic algorithm method, resulting in improved system efficiency and cooling capacity. This contributes to the development and energy saving of magnetic refrigeration cycle and may also enhance the performance of other caloric cycles.
Article
Computer Science, Artificial Intelligence
Yaojun Liu, Ping Wang, Jingjing Liu, Chuanyang Liu
Summary: X-ray circuits are widely used in electronic devices, and their quality is crucial for the overall quality of electronic products. This paper introduces deep learning and artificial intelligence technology to automatically detect defects in X-ray circuit boards. The study developed a defect detection system and tested its performance, making a certain contribution to future AI algorithms for X-ray PCB defect diagnosis.
NEURAL COMPUTING & APPLICATIONS
(2023)
Article
Engineering, Aerospace
Tolga Baklacioglu
AEROSPACE SCIENCE AND TECHNOLOGY
(2016)
Article
Engineering, Aerospace
Tolga Baklacioglu
INTERNATIONAL JOURNAL OF TURBO & JET-ENGINES
(2017)
Article
Engineering, Aerospace
Altug Piskin, Himmet Emre Aktas, Ahmet Topal, Onder Turan, Tolga Baklacioglu
Summary: This paper introduces a novel method for turbine balancing using Ant Colony Optimization, which has been found to outperform other optimization methods. By separating turbine blade sets into two groups, the efficiency of turbine balancing for aircraft gas turbines can be improved.
INTERNATIONAL JOURNAL OF TURBO & JET-ENGINES
(2021)
Article
Engineering, Aerospace
Tolga Baklacioglu, Onder Turan, Hakan Aydin
Summary: This study demonstrates the use of deep learning method and metaheuristic design in predicting relative exergy destruction of turboprop components, achieving more accurate testing results through the combination of multiple hidden layers neural networks and genetic algorithms.
INTERNATIONAL JOURNAL OF TURBO & JET-ENGINES
(2021)
Article
Engineering, Aerospace
A. Piskin, T. Baklacioglu, O. Turan, H. Aydin
AERONAUTICAL JOURNAL
(2020)
Article
Engineering, Aerospace
Ridvan Oruc, Tolga Baklacioglu
AIRCRAFT ENGINEERING AND AEROSPACE TECHNOLOGY
(2020)
Article
Engineering, Aerospace
Ridvan Oruc, Tolga Baklacioglu
AIRCRAFT ENGINEERING AND AEROSPACE TECHNOLOGY
(2020)
Article
Engineering, Multidisciplinary
Ridvan Oruc, Tolga Baklacioglu, Onder Turan
JOURNAL OF THE FACULTY OF ENGINEERING AND ARCHITECTURE OF GAZI UNIVERSITY
(2020)
Article
Engineering, Aerospace
T. Baklacioglu
Summary: Different Artificial Neural Network models were used to predict fuel flow rate of a commercial aircraft in different flight phases, showing good fit for all models. Optimum network topologies were sought through trial-and-error method to obtain more accurate fuel intake estimations.
AERONAUTICAL JOURNAL
(2021)
Article
Engineering, Aerospace
Ridvan Oruc, Tolga Baklacioglu
Summary: This study introduces a new fuel flow rate model for the descent phase of the flight using PSO algorithm, demonstrating high precision in predicting real fuel flow rate values. The use of real FDR data in the analysis adds originality to the study.
AIRCRAFT ENGINEERING AND AEROSPACE TECHNOLOGY
(2021)
Article
Engineering, Aerospace
Altug Piskin, Tolga Baklacioglu, Onder Turan
Summary: This paper introduces a novel hybrid optimization code for aerospace propulsion system optimization. ACOPSO is capable of solving multi-objective and multimodal problems, providing an alternative approach for turbine engine performance calculations.
AIRCRAFT ENGINEERING AND AEROSPACE TECHNOLOGY
(2022)
Article
Thermodynamics
Ozer Can, Tolga Baklacioglu, Erkan Ozturk, Onder Turan
Summary: In this study, various artificial neural networks were utilized to predict the combustion characteristics of a diesel engine. The results showed that all the proposed neural networks had high accuracy in predicting the parameters, and the MLP architecture with the LM algorithm achieved the best results.
Article
Thermodynamics
Ridvan Oruc, Tolga Baklacioglu
Summary: This study presents a new method for analyzing the climb of high performance aircraft using the energy method, which calculates the total energy of the aircraft. By establishing an aircraft performance model and utilizing the energy maneuverability method, specific excess power contours are obtained. Real data is used in the study and the cuckoo search algorithm and particle swarm optimization methods are applied to achieve accurate results in the optimization processes.
Article
Thermodynamics
Ridvan Oruc, Tolga Baklacioglu
Summary: In this study, an estimation model was created to predict specific excess power (Ps) contours based on flight altitude and Mach number. The cuckoo search algorithm (CSA) method was used for modeling, and highly accurate results were obtained. These models are the first attempt in the current literature and include real Flight Data Recorder (FDR) values.