Article
Engineering, Chemical
Robert Makomere, Hilary Rutto, Lawrence Koech, Musamba Banza
Summary: This research project explores the effectiveness of artificial neural networks (ANN) in dry flue gas desulphurization (DFGD). BR and LM training algorithms are used for DFGD modeling. Input data includes diatomite to Ca(OH)(2) ratio, hydration time, hydration temperature, sulphation temperature, and inlet gas concentration, while output data includes sorbent conversion and sulphation responses. By comparing activation functions and hidden layer cells, it is found that BR performs better than LM with lower RMSE and MSE values. The ANN using BR and LM can accurately predict DFGD outcomes, and the shrinking core model indicates that the chemical reaction is the controlling step.
CANADIAN JOURNAL OF CHEMICAL ENGINEERING
(2023)
Article
Energy & Fuels
Ahsan Ali, Muhammad Adnan Khan, Naseem Abbas, Hoimyung Choi
Summary: This study proposes a hydrogen storage prediction system empowered with machine learning, and compares different artificial neural network approaches to determine the optimal method for predicting hydrogen storage capacities.
JOURNAL OF ENERGY STORAGE
(2022)
Article
Energy & Fuels
Hung Vo-Thanh, Menad Nait Amar, Kang-Kun Lee
Summary: This study used three machine learning models to predict the residual and solubility trapping indices of CO2 in saline aquifers. The results showed that the general regression neural network (GRNN) model had better accuracy compared to the other models. The findings are important for evaluating the feasibility of future GCS projects in saline aquifers.
Article
Engineering, Aerospace
Kemal Guven, Andac Tore Samiloglu
Summary: Neural networks, specifically the NARX network, were used to identify an unknown aerial delivery system in this study. The performance of the NARX network was evaluated based on various parameters, and it was found that a model with one hidden layer and five neurons trained using the Bayesian regularization algorithm was sufficient for this problem.
Article
Environmental Studies
Yoochan Kim, Apurna Ghosh, Erkan Topal, Ping Chang
Summary: Future prediction of commodity price is crucial for mining investors and operators. This research evaluated five different estimation techniques and found that the purelin model using Levenberg-Marquardt technique exhibited the best forecast results for iron ore prices. The accuracy of the forecasts was particularly high for up to 2 months ahead.
Article
Mathematics
Jesus de-Prado-Gil, Osama Zaid, Covadonga Palencia, Rebeca Martinez-Garcia
Summary: This paper aims to predict the 28-day splitting tensile strength of self-compacting concrete (SCC) with recycled aggregates (RA) using different artificial neural network (ANN) algorithms. The results show that the Bayesian regularization (BR) algorithm performs the best. Sensitivity analysis reveals that cement is the most influential variable in the prediction.
Article
Computer Science, Artificial Intelligence
Xin Ma, Mei Xie, Johan A. K. Suykens
Summary: This study adopts the idea of Grey-box modelling and develops a neural grey system model. Through six real world case studies, it is demonstrated that the proposed model outperforms others and shows good generality.
Article
Chemistry, Physical
Jesus de-Prado-Gil, Covadonga Palencia, P. Jagadesh, Rebeca Martinez-Garcia
Summary: A considerable amount of discarded building materials are produced each year, leading to ecosystem degradation. Replacing self-compacting concrete with recycled aggregate helps reduce the cost. This study compares three algorithm techniques to estimate the 28-day compressive strength of self-compacting concrete with recycled aggregate. The findings show that all three models are accurate, with Bayesian regularization performing the best.
Article
Computer Science, Interdisciplinary Applications
L. Borkowski, C. Sorini, A. Chattopadhyay
Summary: The RNN-based model is developed to predict nonlinear plastic response under multiaxial loading, utilizing a novel approach to enforce physical conditions and maintain accuracy, efficiency, and widespread applicability in various fields such as metal forming and part life prediction.
COMPUTERS & STRUCTURES
(2022)
Article
Chemistry, Analytical
Sidra Naz, Muhammad Asif Zahoor Raja, Ammara Mehmood, Aneela Zameer Jaafery
Summary: This paper investigates the hysteresis effects of piezoelectric actuators and utilizes artificial intelligence-based neural network methods for numerical analysis and optimization. The effectiveness of these methods is validated through experimental verification.
Article
Engineering, Multidisciplinary
Roshana Mukhtar, Chuan-Yu Chang, Muhammad Asif Zahoor Raja, Naveed Ishtiaq Chaudhary
Summary: The objective of this paper is to propose a novel design of intelligent neuro-supervised networks (INSNs) for studying the dynamics of Parkinson's disease illness (PDI). The designed INSNs utilize multilayer structure neural networks with the Levenberg-Marquardt (LM) and Bayesian regularization (BR) optimization approaches. The outcomes suggest the efficacy of both INSNs solvers for different scenarios in PDI models, with the BR-based method being relatively more accurate.
Article
Education, Scientific Disciplines
Guoqing Zhu, Lin Huang, Jiapeng Yin, Wen Gai, Lijiang Wei
Summary: Fault diagnosis technologies play a crucial role in ensuring the safety and reliability of ocean-going marine diesel engines. This study proposes a diagnosis method based on thermal parametric analysis and neural network algorithms for multiple faults in marine diesel engines. The results indicate that the Levenberg Marquardt back propagation neural network achieves the highest diagnostic accuracy rate and relatively short diagnostic time, while the probabilistic neural network has the fastest diagnosis speed but lower diagnostic accuracy rates.
Article
Engineering, Electrical & Electronic
Guocai Nan, Zhengkuan Wang, Chenghua Wang, Bi Wu, Zhican Wang, Weiqiang Liu, Fabrizio Lombardi
Summary: This work introduces a hybrid-iterative compression algorithm for LSTM/GRU and proposes an energy-efficient accelerator for bidirectional RNNs. By grouping gating units and using different compression algorithms, significant reduction in storage and computation requirements can be achieved without compromising accuracy. Improvements in the data flow of matrix operation unit and BRAM utilization, along with a timing matching strategy, address the load-imbalance issue and result in enhanced energy efficiency compared to state-of-the-art designs.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I-REGULAR PAPERS
(2021)
Article
Energy & Fuels
Muhammad Sabbar Hassan, Khurram Kamal, Tahir Abdul Hussain Ratlamwala
Summary: This research focuses on detecting and classifying power plant faults using intelligent artificial neural network approach. The findings show that both Levenberg-Marquardt and Bayesian regularization methods have high accuracy in detecting and classifying faults, with LM algorithm having faster training speed. The neural network-based methods are effective in detecting and classifying faults with good performance.
ENERGY SOURCES PART A-RECOVERY UTILIZATION AND ENVIRONMENTAL EFFECTS
(2022)
Article
Nuclear Science & Technology
Khalil Moshkbar-Bakhshayesh
Summary: This study introduces a new technique for accurate estimation of nuclear power plant parameters using genetic algorithm, Bayesian regularization, and LM learning algorithm. The results show the superiority of this technique in estimating target parameters during NPP transients.
ANNALS OF NUCLEAR ENERGY
(2021)