Design Of Artificial Neural Networks Research Articles

Efficient prediction of mechanical properties of Boron Nitride Nanosheets using artificial neural networks

This paper aims to evaluate the mechanical properties of Boron Nitride Nanosheets (BNNS) and their vital use in nano-electromechanical systems (NEMS). By employing molecular Dynamics (MD) simulation, modelling of the atomic structure was done. The mechanical response of BNNS under various parameters (strain rate, temperature, chirality and dimension) enabled the generation of a comprehensive data set that accurately represents their elastic properties. The dataset obtained from MD simulation was subsequently utilized to construct an artificial neural network (ANN) model, tailored to predict the Young’s modulus of BNNS accurately. This work aimed to improve the model’s efficiency by refining the design of ANN, which significantly reduces the computational time while maintaining higher accuracy predictions. The findings demonstrate precise and rapid prediction for developing components based on BNNS in NEMS. This paper establishes an analogy between in-depth atomistic simulations and real-world engineering applications presenting a new approach for precisely predicting the attributes of nanomaterials.

Application to novel smart techniques for decarbonization of commercial building heating and cooling through optimal energy management

The present article proposes a novel smart building energy system utilizing deep geothermal resources through naturally-driven borehole thermal energy storage interacting with the district heating network. It includes an intelligent control strategy for lowering operational costs, making better use of renewables, and avoiding CO2 emissions by eliminating heat pumps and cooling machines to address the heating and cooling demands of a commercial building in Uppsala, a city near Stockholm, Sweden. After comprehensively conducting techno-environmental and economic assessments, the system is fine-tuned using artificial neural networks (ANN) for optimization. The study aims to determine which ANN design and training procedure is the most efficient in terms of accuracy and computing speed. It also assesses well-known optimization algorithms using the TOPSIS decision-making technique to find the best trade-off among various indicators. According to the parametric results, deeper boreholes can collect more geothermal energy and reduce CO2 emissions. However, deep drilling becomes more expensive overall, suggesting the need for multi-objective optimization to balance costs and techno-environmental benefits. The results indicate that Levenberg-Marquardt algorithms offer the optimum trade-off between computation time and error minimization. From a TOPSIS perspective, while the dragonfly algorithm is not ideal for optimizing the suggested system, the non-dominated sorting genetic algorithm is the most efficient since it yields more ideal points rated below 100. The optimization yields a higher energy production of 120 kWh/m2, as well as a decreased levelized cost of energy of 57 $/MWh, a shorter payback period of two years, and a reduced CO2 index of 1.90 kg/MWh. The analysis reveals that despite the high investment costs of 382.50 USD/m2, the system is financially beneficial in the long run due to a short payback period of around eight years, which aligns with the goals of future smart energy systems: reduce pollution and increase cost-effectiveness.

Activation Functions for Neural Networks: Application and Performance-based Comparison

Past decade has seen explosive growth of Deep Learning (DL) algorithms based on Artificial Neural Networks (ANNs) and its applications in vast emerging domains to solve real world complex problems. The DL architecture uses Activation Functions (AFs), to perform the task of finding relationship between the input feature and the output. Essential building blocks of any ANN are AFs which bring the required non-linearity of the output in the Output layer of network. Layers of ANNs are combinations of linear and nonlinear AFs. Most extensively used AFs are Sigmoid, Hyperbolic Tangent (Tanh), Rectified Linear Unit (ReLU) etc to name a few. Choosing an AF for a particular AF depends on various factors such as Nature of Application, Design of ANN, Optimizers used in the network, Complexity of Data etc. This paper presents a survey on most widely used AFs along with the important consideration while selecting an AF on a specific problem domain. A broad guideline on selecting an AF based on the literature survey has been presented to help researchers in employing suitable AF in their problem domain.

A memetic dynamic coral reef optimisation algorithm for simultaneous training, design, and optimisation of artificial neural networks

Artificial Neural Networks (ANNs) have been used in a multitude of real-world applications given their predictive capabilities, and algorithms based on gradient descent, such as Backpropagation (BP) and variants, are usually considered for their optimisation. However, these algorithms have been shown to get stuck at local optima, and they require a cautious design of the architecture of the model. This paper proposes a novel memetic training method for simultaneously learning the ANNs structure and weights based on the Coral Reef Optimisation algorithms (CROs), a global-search metaheuristic based on corals’ biology and coral reef formation. Three versions based on the original CRO combined with a Local Search procedure are developed: (1) the basic one, called Memetic CRO; (2) a statistically guided version called Memetic SCRO (M-SCRO) that adjusts the algorithm parameters based on the population fitness; (3) and, finally, an improved Dynamic Statistically-driven version called Memetic Dynamic SCRO (M-DSCRO). M-DSCRO is designed with the idea of improving the M-SCRO version in the evolutionary process, evaluating whether the fitness distribution of the population of ANNs is normal to automatically decide the statistic to be used for assigning the algorithm parameters. Furthermore, all algorithms are adapted to the design of ANNs by means of the most suitable operators. The performance of the different algorithms is evaluated with 40 classification datasets, showing that the proposed M-DSCRO algorithm outperforms the other two versions on most of the datasets. In the final analysis, M-DSCRO is compared against four state-of-the-art methods, demonstrating its superior efficacy in terms of overall accuracy and minority class performance.

Application of Artificial Neural Networks for Identification of Lithofacies by Processing of Core Drilling Data

Identifying lithofacies types from core drilling data presents significant challenges, especially given the limited number of physical drilling characteristics available for analysis. Traditional machine learning methods often face issues with poor training and testing due to these limitations. Addressing this, we propose a new method for processing core drilling data to improve the accuracy of deep artificial neural networks (DANNs) in lithofacies recognition. Our approach transforms torque, weight on bit (WOB), and rotational speed data into three square matrices, creating a novel three-channel lithofacies image. This method allows for the application of DANNs by converting the complex lithofacies recognition task into a more standard image recognition problem. The developed method dramatically increases the input vector dimensions, enhancing the richness of the data input. The validation of results revealed that the DANN model trained for merely 3000 iterations successfully predicted lithofacies types of all eight testing samples in a mere 2.85 ms, showcasing superior accuracy. The innovative drilling data processing method proposed in this study enables DANNs to identify lithofacies with increased speed and accuracy. This offers a new direction for other DANNs utilizing drilling data.

Artificial neural network identification of exercise expiratory flow-limitation in adults

Identification of ventilatory constraint is a key objective of clinical exercise testing. Expiratory flow-limitation (EFL) is a well-known type of ventilatory constraint. However, EFL is difficult to measure, and commercial metabolic carts do not readily identify or quantify EFL. Deep machine learning might provide a new approach for identifying EFL. The objective of this study was to determine if a convolutional neural network (CNN) could accurately identify EFL during exercise in adults in whom baseline airway function varied from normal to mildly obstructed. 2931 spontaneous exercise flow-volume loops (eFVL) were placed within the baseline maximal expiratory flow-volume curves (MEFV) from 22 adults (15 M, 7 F; age, 32 yrs) in whom lung function varied from normal to mildly obstructed. Each eFVL was coded as EFL or non-EFL, where EFL was defined by eFVLs with expired airflow meeting or exceeding the MEFV curve. A CNN with seven hidden layers and a 2-neuron softmax output layer was used to analyze the eFVLs. Three separate analyses were conducted: (1) all subjects (n = 2931 eFVLs, [GRALL]), (2) subjects with normal spirometry (n = 1921 eFVLs [GRNORM]), (3) subjects with mild airway obstruction (n = 1010 eFVLs, [GRLOW]). The final output of the CNN was the probability of EFL or non-EFL in each eFVL, which is considered EFL if the probability exceeds 0.5 or 50%. Baseline forced expiratory volume in 1 s/forced vital capacity was 0.77 (94% predicted) in GRALL, 0.83 (100% predicted) in GRNORM, and 0.69 (83% predicted) in GRLOW. CNN model accuracy was 90.6, 90.5, and 88.0% in GRALL, GRNORM and GRLOW, respectively. Negative predictive value (NPV) was higher than positive predictive value (PPV) in GRNORM (93.5 vs. 78.2% for NPV vs. PPV). In GRLOW, PPV was slightly higher than NPV (89.5 vs. 84.5% for PPV vs. NPV). A CNN performed very well at identifying eFVLs with EFL during exercise. These findings suggest that deep machine learning could become a viable tool for identifying ventilatory constraint during clinical exercise testing.

ANN SUPPORTED STUDY ON THE PERFORMANCE AND SLURRY EROSION RESISTANCE OF THERMAL SPRAYED WC20CR3C27NI COATINGS

The thermal spray coatings are commonly employed in slurry pump components and hydrodynamic turbine blades, where wear progression is an intricate phenomenon. In this research work, the performance analysis of HVOF and APS sprayed WC20Cr3C27Ni coatings for slurry erosion wear is carried out by using artificial neural networks (ANN). The influence of time, particle size, impact angle, speed, and slurry concentration on wear performance of coatings and turbine steel substrate are evaluated. Under the experimental settings, slurry erosion wear rates and mass loss for both coatings and substrate were determined. When ASTM A743 steel was coated with thermal sprayed WC20Cr3C27Ni coatings, the slurry erosion wear resistance of the steel was enhanced by 2 and 3.5 times for APS and HVOF coatings, respectively. The design of ANN made it possible to examine the interactions between the seven input variables. A robust model was formed by the two outputs that followed. This model enables the prediction of slurry erosion wear rate and mass loss of WC20Cr3C27Ni coatings and substrate.

Energy captivation in vibration of laminate nanocomposite cylinders using nonlocal strain gradient theory and artificial neural network

Artificial intelligence has revolutionized the process of solving engineering problems by facilitating predictions and optimizations with ease. The use of artificial neural networks (ANNs) for predicting the mechanical behavior of small-scale structures, particularly in the instability field, is widely prevalent. The present study aims to use an ANN to predict the frequency response of microscale spinning cylindrical laminate shells. Our ANN design utilizes a novel modified Adam optimizer for dimensional consistency, and a genetic algorithm is used for optimizing the number of nodes in the hidden layers of the ANN. ANNs require a valid dataset for training and testing, thus, a numerical solution of the size-dependent elasticity theory is generated, and a parametric study is conducted to find the dependency of vibrational responses on different parameters of the cylindrical shell resting on an elastic foundation equipped with a piezoelectric layer. These numerical results are utilized to train and test the designed ANN. The unique feature of the current ANN is the utilization of several well-known optimizers to observe their efficacy in solving the vibrational problem. The results of the vibration study from both numerical and ANN are presented using transverse displacement of the structure. The study finds that the AdaDelta optimizer performs satisfactorily in comparison to other optimization algorithms in terms of both total error and convergence rate.

An efficient design methodology to speed up the FPGA implementation of artificial neural networks

In this paper, we propose and formulate a C++ based training methodology for speeding up the implementation of an Artificial Neural Network (ANN) in a Field Programmable Gate Array (FPGA). The proposed ANN implementation methodology uses a custom C++ based program referred to as Neural Network Design Parameter Extraction (NNDPE) program developed using an open source library namely Fast Artificial Neural Network (FANN). The proposed C++ based NNDPE program reduces the time required to train and extract the design parameters of the ANN, such as the number of layers, the number of neurons in each layer, and the weights. The extracted ANN design parameters, custom hardware arithmetic units, and function-approximate activation functions are used to implement the ANN hardware architecture of a linear function on the Virtex-7 FPGA platform. The Vivado 2018.3 tool is used to simulate, synthesize, and implement the ANN-based linear function. It was concluded from the simulation that the ANN hardware implementation has high precision due to the usage of floating-point arithmetic operations compared to the ANN which uses fixed-point arithmetic operations. The ANN-implemented on the Virtex-7 FPGA operates at 150.76 MHz, which is approximately 11 to 18 times faster than the software implementation running on various CPU cores with operating frequencies ranging over 2.2 GHz to 4.70 GHz.

Design of Time-Encoded Spiking Neural Networks in 7-nm CMOS Technology

In biologically inspired spiking neural networks (SNNs) neurons communicate by short pulses, called spikes. SNNs have the potential to be more power efficient than artificial neural networks (ANNs), thanks to the fewer computational steps required by the spike transmission and processing, as compared to the multiply-and-accumulate (MAC) operations with wide bit-vectors usually adopted in ANNs. We present the design of two types of SNNs with integrate-and-fire dynamics and single-spike per neuron operation, where neural communication is based on synchronous time-to-first-spike (sTTFS) and time-to-first-spike (TTFS) encoding schemes. In the considered time-encoded SNNs, the information is carried by the timing of the spikes with respect to a reference time. In 7nm CMOS technology both designs are synthesized as VHDL-based random-logic-macros (RLMs) and compared to an equivalent ANN design in terms of power consumption, latency and silicon area, using the Iris data set for inference. A cost function expressed as a product of energy consumption and silicon area is introduced to compare the three network designs. With respect to this cost function, it turns out that the SNN-TTFS implemented for the considered classification task outperforms the ANN used as baseline model.

Suicide Risk Analysis and Psycho-Emotional Risk Factors Using an Artificial Neural Network System.

Suicidal behavior among young people has become an increasingly relevant topic after the COVID-19 pandemic and constitutes a public health problem. This study aimed to examine the variables associated with suicide risk and determine their predictive capacity. The specific objectives were: (1) to analyze the relationship between suicide risk and model variables and (2) to design an artificial neural network (ANN) with predictive capacity for suicide risk. The sample comprised 337 youths aged 18-33 years. An ex post facto design was used. The results showed that emotional attention, followed by problem solving and perfectionism, were variables that contributed the most to the ANN's predictive capacity. The ANN achieved a hit rate of 85.7%, which is much higher than chance, and with only 14.3% of incorrect cases. This study extracted relevant information on suicide risk and the related risk and protective factors via artificial intelligence. These data will be useful for diagnosis as well as for psycho-educational guidance and prevention. This study was one of the first to apply this innovative methodology based on an ANN design to study these variables.

A comprehensive framework based on Bayesian optimization and skip connections artificial neural networks to predict buildings energy performance

Artificial Neural Networks (ANNs) are a powerful modeling and prediction tool. However, in the energy and buildings fields, certain articles have been highlighted for reducing ANNs' ability to aid in building energy prediction. This result stems from the author's limited ability to regulate a large number of hyper-parameters. As a result, this paper provides a comprehensive study of ANN challenges and solutions in a single framework. The study's research aims may be stated as defining the necessary data preparation stages; selecting the appropriate ANN design for each study scenario with its hyper-parameters; and presenting the approach for training and evaluating ANNs to obtain the best performance. All of these concerns are presented in this study, compiled into a comprehensive framework, and implemented on two different types of real-life data: 1) CBECS data for US commercial buildings, and 2) ASHRAE data for different buildings around the world. The framework optimization loop mainly relies on the Bayesian optimization technique, which employs Gaussian Processes (GP) as a surrogate function; a combination of two concepts (i.e., Expected Improvement (EI) and Probability Improvement (PI)) as an acquisition function; hyper-parameters as a dimension space optimization; and adjusted R2 value as a target optimization score to be improved. Various skip connection ANN architectures are applied to demonstrate the ability to modify ANNs' limitations in the energy and building fields, assuring excellent accuracy and reliability in forecasting building energy performance, with adjusted R2 of more than 0.93 for both data sets.

Ultra-low power resistive random-access memory based on VO2/TiO2 nanotubes composite film

In this study, a resistive random access memory (RRAM) based on VO2/TiO2 nanotubes was prepared. Compared with single VO2-based RRAM or TiO2 nanotube-based RRAM, the Ag/VO2/TiO2 nanotube/FTO RRAM device has advantages such as low power consumption (∼0.154 pJ), excellent stability (3000 DC cycles), and high switching rate (300 times). It still shows certain advantages compared with TiO2-based low-power RRAM (0.5 pJ). The device can also exhibit 20 consecutive long-term enhancement/suppression behaviors, which can be used to simulate brain-like activity and achieve recognition accuracy of 95% (8 × 8 pixels) and 88% (28 × 28 pixels) in artificial neural network identification field.

Predicting mechanical properties of defective h-BN nanosheets using Data-Driven models

Machine learning (ML) methods have assisted different fields of science in overcoming challenging problems. In this work, with the aid of ML and deep learning algorithms, predicting the mechanical properties of defective single-layer hexagonal boron nitride (h-BN) nanosheets is accelerated compared to conventional atomistic simulations. To provide the required data for ML models, molecular dynamic (MD) simulation is used to determine the Young modulus, ultimate tensile strength, and fracture strain of these 2D nanostructures under the influence of various defect features, namely density, type, structure, and distribution. The obtained MD results exhibit strong correlation between the mechanical properties of defective h-BN monolayers and the introduced defect characteristics. Accuracy of the results is supported by the analysis of variance (ANOVA) and correlation coefficient between independent and dependent variables. In the next step, three ML algorithms including SVR, Random Forest and XGBoost, and three artificial neural network (ANN) models with different hidden layers are trained to accomplish our goals. Among all the developed models, an ANN design with four hidden layers performs better with the highest R2 score of 0.86. Our findings provide significant assets for defect engineering by accelerating the mechanical properties prediction of defective h-BN monolayers compared to traditional MD simulations.

Energy-Efficient Hardware Implementation of Fully Connected Artificial Neural Networks Using Approximate Arithmetic Blocks.

In this paper, we explore efficient hardware implementation of feedforward artificial neural networks (ANNs) using approximate adders and multipliers. Due to a large area requirement in a parallel architecture, the ANNs are implemented under the time-multiplexed architecture where computing resources are re-used in the multiply accumulate (MAC) blocks. The efficient hardware implementation of ANNs is realized by replacing the exact adders and multipliers in the MAC blocks by the approximate ones taking into account the hardware accuracy. Additionally, an algorithm to determine the approximate level of multipliers and adders due to the expected accuracy is proposed. As an application, the MNIST and SVHN databases are considered. To examine the efficiency of the proposed method, various architectures and structures of ANNs are realized. Experimental results show that the ANNs designed using the proposed approximate multiplier have a smaller area and consume less energy than those designed using previously proposed prominent approximate multipliers. It is also observed that the use of both approximate adders and multipliers yields, respectively, up to 50% and 10% reduction in energy consumption and area of the ANN design with a small deviation or better hardware accuracy when compared to the exact adders and multipliers.

Comparison of the characteristics of the control strategies based on artificial neural network and genetic algorithm for air conditioning systems

Air conditioning systems are playing an increasingly important role in our daily life, and the control of air conditioning systems is related to the intelligence of the system, indoor thermal comfort and energy consumption. Therefore, the control of air conditioning systems has always been an important research topic. Artificial neural network (ANN) and genetic algorithm (GA) are two commonly used intelligent control methods. However, there is no guidance until now on the applicability of these two control methods and how to make a choice between them in the control of air conditioning systems. In this study, an ANN control strategy and a GA control strategy were developed and experimentally verified. The experimental results indicated that both the ANN control strategy and the GA control strategy could control the air conditioning system properly under different control commands. To compare the control performances, the convergence speed, discrete characteristic, energy consumption and interference resistance ability were calculated or experimentally validated. The ANN control strategy showed better performances in the convergence speed and energy consumption. While the GA control strategy performed better in maintaining the stable state of the air conditioning system.The innovation of this study lies in two points. First, when designing the ANN control strategy and the GA control strategy, ANN and GA were applied as the central control algorithm, rather than only as auxiliary algorithms for system identification, prediction or optimization. This methodology is relatively novel in the design of ANN and GA control strategies for the air conditioning system, and could further expand the application of ANN and GA. Second, this study employed 4 evaluating indicators including convergence speed, discrete characteristic, energy consumption and interference resistance ability, to comprehensively evaluated the control performance of ANN control strategy and GA control strategy.

Design of ANN Based Non-Linear Network Using Interconnection of Parallel Processor

Suspicious mass traffic constantly evolves, making network behaviour tracing and structure more complex. Neural networks yield promising results by considering a sufficient number of processing elements with strong interconnections between them. They offer efficient computational Hopfield neural networks models and optimization constraints used by undergoing a good amount of parallelism to yield optimal results. Artificial neural network (ANN) offers optimal solutions in classifying and clustering the various reels of data, and the results obtained purely depend on identifying a problem. In this research work, the design of optimized applications is presented in an organized manner. In addition, this research work examines theoretical approaches to achieving optimized results using ANN. It mainly focuses on designing rules. The optimizing design approach of neural networks analyzes the internal process of the neural networks. Practices in developing the network are based on the interconnections among the hidden nodes and their learning parameters. The methodology is proven best for nonlinear resource allocation problems with a suitable design and complex issues. The ANN proposed here considers more or less 46k nodes hidden inside 49 million connections employed on full-fledged parallel processors. The proposed ANN offered optimal results in real-world application problems, and the results were obtained using MATLAB.

Do international resources configure SMEs' sustainable performance in the digital era? Evidence from Pakistan

This study examines the impact of international resources on sustainable performance through the mediating role of green entrepreneurial orientation. In emerging economies, the business sector, especially Small and Medium Enterprises (SMEs), has a number of obstacles, including a lack of resources, financial support, and professional skills. Furthermore, SMEs depend on international resources to compete in an unpredictable industry in the long term. A structured questionnaire survey was undertaken to collect data from 380 Pakistani SMEs in order to preserve the productive insights. For data processing, a hybrid partial least square structural equation modelling (PLS-SEM) and Artificial Neural Network (ANN) technique was used. The findings show that financial resources have a favorable and substantial influence on SMEs' long-term performance, but technical resources have no meaningful impact on SMEs' long-term performance. Furthermore, green entrepreneurial approach somewhat mediators the association between financial resources and long-term success while totally mediating the link between technical resources and long-term performance. Finally, we run a sensitivity analysis to confirm our results. The ANN design explains 71% of the suggested model's accuracy. This report offers significant advice to managers, politicians, and other stakeholders about the importance of international resources and competencies. It is advised that the Small and Medium Enterprise Development Authority (SMEDA) and the government of Pakistan concentrate more on these chances for long-term performance in order to compete in the worldwide market.

Increasing the accuracy of estimating the viscosity of SAE40-based nanofluid containing MWCNT-TiO2 nanoparticles using a creative method in optimizing the structure of the artificial neural network

The viscosity (μnf) is one of the influencing parameters in choosing that nanofluid (NF) that affects its thermal behavior and heat transfer. In this regard, the present study is conducted to investigate the μnfof MWCNT-TiO2 (10:90) /SAE40 NF using an artificial neural network (ANN). ANN design in temperature conditions between 25 and 50°C and in different solid volume fraction (SVF) of nanoparticles (φ =0.0625%−1%). 174 laboratory data were used. Three inputs (temperature(T), φ and shear rate(SR)) and one output (μnf) are determined for ANN. In this study, ANN modeling was done with MLP method and Levenberg-Marquardt (LM) training algorithm. The selected optimal structure, among different ANN structures for MWCNT-TiO2(10:90)/SAE40 NF, has two hidden layers with an optimal structure of 10 and 4 neurons in the first and second layers. The results of R and MSE coefficients prove the accuracy of the proposed model., which in the final stage was set equal to 0.9999507 and 0.004599708, respectively. The margin of deviation (MOD) in the grid data set is in the range − 3% <MOD < + 3%. Based on studies that considered ANN correlation with computational data compared to laboratory data, the results of this comparison show that the ANN is more accurate and reliable for estimating the μnf. The turning point of this article is designing and predicting the viscosity of MWCNT-TiO2/SAE40 NF in the least time and financial costs compared to repetitive and time-consuming experiments through ANN.

Artificial intelligence-aided nanoplasmonic biosensor modeling

The use of monolithic biosensors is increasingly popular in fundamental biological studies. Rapid advances in nanophotonic biosensors are leading to lab-on-a-chip platforms. In this paper we propose a method to use artificial neural networks (ANNs) to predict the output electrical signal of biosensor. Multilayer perceptron model is developed by assuming electrical current as outputs, and the refractive index of biosamples, central wavelength (λ) and full width half maximum (FWHM) of input light source as inputs. A comparative approach was applied between finite-difference time-domain (FDTD) method and ANN results to evaluate the biosensor’s ANN model. Results showed that the ANN design with topology of (3 5 4 4 6 21) can predict the output accurately based on the value of mean square error (MSE) about 2.9 × 10−8 as evaluation parameter. It was shown that the developed ANN model can approximate the outcome to high precision with only a small sampling of the data. Using the developed model, pre-optimization was run to find the optimum condition for electrical sensitivity and responsivity of the device. We found that the light source with central wavelength of 735 nm and FWHM of 70 nm can simultaneously satisfy the optimum conditions for sensitivity and responsivity.