Layer Artificial Neural Network Research Articles

Deep Learning for Walking Behaviour Detection in Elderly People Using Smart Footwear.

The increase in the proportion of elderly in Europe brings with it certain challenges that society needs to address, such as custodial care. We propose a scalable, easily modulated and live assistive technology system, based on a comfortable smart footwear capable of detecting walking behaviour, in order to prevent possible health problems in the elderly, facilitating their urban life as independently and safety as possible. This brings with it the challenge of handling the large amounts of data generated, transmitting and pre-processing that information and analysing it with the aim of obtaining useful information in real/near-real time. This is the basis of information theory. This work presents a complete system aiming at elderly people that can detect different user behaviours/events (sitting, standing without imbalance, standing with imbalance, walking, running, tripping) through information acquired from 20 types of sensor measurements (16 piezoelectric pressure sensors, one accelerometer returning reading for the 3 axis and one temperature sensor) and warn the relatives about possible risks in near-real time. For the detection of these events, a hierarchical structure of cascading binary models is designed and applied using artificial neural network (ANN) algorithms and deep learning techniques. The best models are achieved with convolutional layered ANN and multilayer perceptrons. The overall event detection performance achieves an average accuracy and area under the ROC curve of 0.84 and 0.96, respectively.

Artificial neural network integrated with bio-inspired approach for optimal VAr management and voltage profile enhancement in grid system

ABSTRACT This article presents the design of an improved smart grid model with an improved optimal VAr strategy using a bio-inspired approach. The integration of an artificial neural network technique with bio-inspired algorithms is the novelty of this research work. The objective of the work is attained by integrating the artificial neural network with three different evolving techniques like the Big Bang-Big Crunch algorithm, the gravitational search algorithm, and bacterial foraging algorithm. The three different layers of artificial neural network reconcile with the power grid for enhancing the voltage profile generated by bio-inspired techniques. The effectiveness and reliability of the algorithms are examined on the standard IEEE 30 bus system. The sample test bus system comprises six Generators, four Transformer tap changers, and 2 Static Shunt VAr Compensator. Artificial intelligence technology is applied for enhancing the voltage profile by managing reactive power across the power system network. The results produced by artificial intelligence technique are more competent and outstanding than any other conservative optimization techniques. It is found that the real power loss across the network has been reduced from 5.744 MW to 4.234 MW, while the bus voltage deviation has been minimized from 1.475 p.u. to 0.0918 p.u. and the voltage profile has been improved from 0.968 to 1.0985 p.u. after integration of the artificial neural network. The optimal values of objective functions are attained by evaluating the training and testing performance of artificial neural network criteria such as coefficient of determination, mean absolute error, root mean square error, and mean absolute percentage error. The obtained results show the remarkable improvement in voltage profile after integrating artificial intelligence technology with a power grid system.

K-fold Validation Neural Network Approach for Predicting the One-Day Compressive Strength of UHPC

Abstract This paper analyzes the application of artificial neural networks (ANN) to predict the 1-day compressive strength of ultra-high-performance concrete (UHPC) made with any combination of powders and supplementary cementitious materials (SCM) such as silica fume (SF), fly ash (FA), ground granulated blast slag furnace (GGBSF), recycled glass powder (GP), rice husk ash (RHA), fluid catalytic cracking catalyst residue (FC3R), metakaolin (MK), limestone powder (LP), and quartz powder (QP). A total of 604 data from scientific literature were used to train the one hidden layer ANN model by using the k-fold validation procedure. Furthermore, 90 UHPC mixtures were experimentally performed to validate the proposed ANN model. The performance of the model was assessed using several statistical performance indexes: ratio of the root mean square error to the standard deviation of measured data (RSR), root mean square error (RSME), normalized mean bias error (NMBE), Nash–Sutcliff efficiency, and coefficient of multiple determination (R2). Connection weight approach (CWA) algorithm was utilized to analyze the relationships between the UHPC components and the 1-day compressive strength. The results indicated that the ANN is an efficient model for predicting the early strength (1-day compressive strength) of UHPC achieving R2 values of 0.88 and 0.86 on the test data and experimental data, respectively, even when the experimental dosages included combinations of components that were not found in the training data. The results of the CWA analysis indicated that SCM such as MK, FC3R, SF, and LP, as well as other factors such as virtual packing density, improved the early strength of UHPC, whereas FA, GP, and RHA were pointed out as harmful for the one-day compressive strength. As conclusion, the ANN model could be helpful in the developing of UHPC with early strength needs by preselecting the combinations of available SCM and powders that have better results in the model at lower cost.

A Flow Information-Based Prediction Model Applied to the Nonaxisymmetric Hub Optimization of a Centrifugal Impeller

Abstract A full-range prediction model for turbomachinery based on the flow-field information code is established in this article to solve the problems that traditional models do not have enough prediction accuracy and cannot reflect the complete performance characteristics of the impeller. The model, which can predict the complete performance curve of the impeller with higher accuracy, consists of two multilayer artificial neural network (ANN) submodels. Different from the traditional model, the ANN submodel uses the flow-field information code for pretraining layer by layer. The flow-field information code is the characteristic information extracted from the impeller flow field through the proper orthogonal decomposition (POD) method. By implicitly learning the flow-field information, the prediction error of the model is reduced by 29.7% compared with the single hidden layer ANN. Based on this model, the nonaxisymmetric, but periodic, hub optimization of a centrifugal impeller with 30 variables is carried out, with the goals of the higher efficiency and the wider flow range at the specified pressure ratio and the massflow rate at the design point. The result shows that, after the optimization, the isentropic efficiency at the design point increases by 1% and the flow range increases by 2% compared to the baseline.

Deep learning and machine learning techniques to improve hand movement classification in myoelectric control system

A robust myoelectric control system (MCS) is essential for the design of electromyography (EMG) based human–machine interface (HMI) designs such as prosthetics, exoskeleton, wheelchair and humanoid robots. The functionality of the current pattern recognition (PR) technique in MCS is limited by factors such as variation in the user’s limb position. To overcome the effect of this dynamic variation, an invariant higher order statistics – frequency domain feature set (HOS-FD) is proposed in this paper. EMG data from eight hand movements in five limb positions are considered. When trained with three limb position data, the HOS-FD with three hidden layers deep neural network (DNN) achieved a significantly high classification accuracy of 97.84%±0.22 compared to other classifiers viz., single layer artificial neural network (ANN), linear discriminant analysis (LDA), support vector machine (SVM), k nearest neighbor (kNN), decision tree (DT) and Naïve Bayes (NB) classifiers with accuracies of 94.11%±1.63, 95.02%±1.89, 94.63%±2.33, 90.05 ± 4.11, 86.66 ± 4.72 and 78.78%±5.02 respectively. Further, when trained with data from all five limb positions, the proposed feature set with DNN had an accuracy of 99.16%±0.14. The statistical significance of the high classification accuracy obtained using the proposed feature set is also proven using multiple analysis of variance tests (p < 0.001). These results indicate that the proposed method is a promising technique for HMI.

Determination of a Renewable Energy System Capacity Factor Using Machining Learning Algorithms

Across the world, machining learning (ML) algorithms, such as support vector machines (SVR) and artificial neural networks (ANN), are among scientific tools for the fourth industrial revolution, or Industry 4.0, campaign. These algorithms have wide engineering applications, but their potentials in energy management are still evolving. Hence, this study investigates the performance of SVR and ANN algorithms as predictive models for wind turbines capacity factor (CF) estimation. Five independent parameters-wind speed, power density, turbulence intensity, installed capacity, and wind shear - were used as input parameters for this estimation problem. Polynomial, radial basis function (RBF), and linear kernels were used to train an SVR model that estimates CF, while Adams was used to optimize the performance of a backpropagation ANN model. These models’ applicability was evaluated using data sets from eight locations. This study used correlation coefficient used to compared the model performance. The RBF trained SVR model performed better than the other kernels during model training, but the linear kernel trained SVR model performed better than the other kernels during testing model. When these kernels performances were compared with a single hidden layer ANN model, the ANN model results were better than these kernels results.

Multilayer Backpropagation Neural Networks for Implementation of Logic Gates

ANN is a computational model that is composed of several processing elements (neurons) that tries to solve a specific problem. Like the human brain, it provides the ability to learn from experiences without being explicitly programmed. This article is based on the implementation of artificial neural networks for logic gates. At first, the 3 layers Artificial Neural Network is designed with 2 input neurons, 2 hidden neurons &amp; 1 output neuron. after that model is trained by using a backpropagation algorithm until the model satisfies the predefined error criteria (e) which set 0.01 in this experiment. The learning rate (α) used for this experiment was 0.01. The NN model produces correct output at iteration (p)= 20000 for AND, NAND &amp; NOR gate. For OR &amp; XOR the correct output is predicted at iteration (p)=15000 &amp; 80000 respectively.

Joint, accurate and robust optical signal-to- noise ratio and modulation format monitoring scheme using a single Stokes-parameter-based artificial neural network.

A joint and robust optical signal-to-noise ratio (OSNR) and modulation format monitoring scheme using an artificial neural network (ANN) is proposed and demonstrated via both numerical simulations and experiments. Before ANN, the power iteration method in Stoke space is employed to estimate the phase difference between two orthogonal polarizations caused by fiber birefringence. Then, a three layers ANN is employed to approximate the relationship between the cumulative distribution function of a single Stokes parameter (S2) and the targeted OSNR and format information. The simulation results show that the probability of OSNR estimation error within 1dB in the proposed scheme is 100%, 99.78%, 100%, 99.78% and 98.89% for 28GS/s QPSK, 8PSK, 8QAM, 16QAM and 64QAM, respectively. Meanwhile, the proposed scheme also shows high modulation format identification accuracy in the presence of nonlinear Kerr effect and residual chromatic dispersion. With 1 dB OSNR estimation error, the proposed scheme can tolerate the residual chromatic dispersion and phase-related polarization rotation rate up to 100ps/nm and 50kHz, respectively. The experimental results also further confirm that the proposed scheme shows high modulation identification accuracy for 28GS/s QPSK, 8PSK and 16QAM under the scenarios of both back-to-back and fiber transmission. Meanwhile, with the launched power of 0dBm, the mean OSNR estimation error in our scheme is smaller than 1 dB within ±160ps/nm residual chromatic dispersion after fiber transmission.

DR-Net: An Improved Network for Building Extraction from High Resolution Remote Sensing Image

At present, convolutional neural networks (CNN) have been widely used in building extraction from remote sensing imagery (RSI), but there are still some bottlenecks. On the one hand, there are so many parameters in the previous network with complex structure, which will occupy lots of memories and consume much time during training process. On the other hand, low-level features extracted by shallow layers and abstract features extracted by deep layers of artificial neural network cannot be fully fused, which leads to an inaccurate building extraction from RSI. To alleviate these disadvantages, a dense residual neural network (DR-Net) was proposed in this paper. DR-Net uses a deeplabv3+Net encoder/decoder backbone, in combination with densely connected convolution neural network (DCNN) and residual network (ResNet) structure. Compared with deeplabv3+net (containing about 41 million parameters) and BRRNet (containing about 17 million parameters), DR-Net contains about 9 million parameters; So, the number of parameters reduced a lot. The experimental results for both the WHU Building Dataset and Massachusetts Building Dataset, DR-Net show better performance in building extraction than other two state-of-the-art methods. Experiments on WHU building data set showed that Intersection over Union (IoU) increased by 2.4% and F1 score increased by 1.4%; in terms of Massachusetts Building Dataset, IoU increased by 3.8% and F1 score increased by 2.9%.

Efficient synergetic filtering in big data set using neural network technique

Presently, great accomplishment on speech-recognition, computer-vision and natural-language processing has been achieved by deep-neural networks. To tackle the major trouble in synergetic or collaborative-filtering we concentrated intensively on the techniques based on neural networks. Although a few recent researches have employed deep learning, they mostly used it to carve auxiliary facts, along with textual metaphors of objects and acoustic capabilities of music. We present a popular framework named Artificial Neural Synergetic Filtering (ANSF) to substitute the core makeup with a neural design which could be very efficient to analyse data with a random feature. ANSF is a prevalent matrix-factorisation framework. To improvise it with non-linearity we propose to leverage a multilayer perceptron to investigate customer-object communication function. In-depth experiments on actual global databases display big improvement over the latest techniques. Investigational results manifest that the application of core layers of artificial neural networks gives improved overall performance.

Machine learning and deep learning on health diagnosis of rotating fan

ABSTRACT Machine learning and deep learning on the health diagnosis of a rotating machine are studied for smart monitoring. The signals of vibration and sound pressure of a rotating fan driven by DC motor detected by an accelerometer and microphone are processed by machine/deep learning for health diagnosis of blade. For the machine learning, two methods, support vector machine (SVM) and random forest (RF), are used for classification of normal and abnormal status based on three features extracted from the signals in time domain and frequency domain. For the deep learning, convolution neural network (CNN) method is used to process the two signals in time domain for modelling; certain layers of convolution and pooling for feature extraction are followed by two layers of artificial neural network. After the learning, a confusion matrix of testing is given to evaluate the performance. In particular, the importance scores of input features are analyzed by RF, which is useful for us to screen out the non-significant features for improving the learning to avoid overfitting. We demonstrate three different methods (SVM, RF, CNN) on the diagnosis of a rotating fan with a damaged blade to illustrate their characteristics.

A single layer artificial neural network type architecture with molecular engineered bacteria for reversible and irreversible computing.

Here, we adapted the basic concept of artificial neural networks (ANNs) and experimentally demonstrate a broadly applicable single layer ANN type architecture with molecular engineered bacteria to perform complex irreversible computing like multiplexing, de-multiplexing, encoding, decoding, majority functions, and reversible computing like Feynman and Fredkin gates. The encoder and majority functions and reversible computing were experimentally implemented within living cells for the first time. We created cellular devices, which worked as artificial neuro-synapses in bacteria, where input chemical signals were linearly combined and processed through a non-linear activation function to produce fluorescent protein outputs. To create such cellular devices, we established a set of rules by correlating truth tables, mathematical equations of ANNs, and cellular device design, which unlike cellular computing, does not require a circuit diagram and the equation directly correlates the design of the cellular device. To our knowledge this is the first adaptation of ANN type architecture with engineered cells. This work may have significance in establishing a new platform for cellular computing, reversible computing and in transforming living cells as ANN-enabled hardware.

Development of artificial neural networks for an energy storage system generated power prediction

As energy utility firms expand their production outputs from renewable energy resources, interest in investment in an energy storage system (ESS) will increase in the coming years. This article determines the optimal number of hidden nodes and the maximum number of epochs to generate the best prediction results for this generated power estimation. Primary energy efficiency, energy ratio, and ambient air temperature were used to predict the generated power of an ESS. Experimental data sets for compressed air energy storage (CAES) were used to train a single hidden layer artificial neural network (ANN) model. The developed ANN models for CAES and compressed air storage with humidification (CASH) show quick convergence to the targeted error function. Based on eight different artificial neural network architectures, a 3−5−1 ANN architecture generated the best results. This architecture mean squared error was 0.00042. Using experimental data for CASH to validate the single hidden layer model, the ANN model performed satisfactorily.

Predicting the performance of a desulfurizing bio-filter using an artificial neural network (ANN) model

The aim of this study was to develop a model for predicting the performance of a desulfurizing bio-filter (BF), without requiring prior information about H2S biodegradation kinetics and mechanism. A single hidden layer artificial neural network (ANN) model was developed and validated using the gradient descent backpropagation (GDBP) learning algorithm coupled with a learning rate and a momentum factor. The ANN model inputs were gas flow rate, residence time, and axial position in the BF bed. The removal efficiency of H2S was the model output. Various structures for ANN model, differing in the number of hidden layer neurons, were trained and an early stopping validation technique, the K-fold cross-validation, was used to determine the optimal structure with the best generalization ability. The modeling results showed that there was a good agreement between the experimental data and the predicted values, with a determination coefficient (R&lt;sup&gt;2&lt;/sup&gt;) of 94%. This implies that the ANN model might be an attractive and useful alternative tool for forecasting the performance of desulfurizing BFs.

Moving ridge neuronal espionage network simulation for reticulum invasion sensing

PurposeThe paper aims to precise and fast categorization on to transaction evolves into indispensible. The effective capacity difficulty of all the IDS simulates today at below discovery amount of fewer regular barrage associations and therefore the next warning rate.Design/methodology/approachThe reticulum perception is that the methods which examine and determine the scheme of contact on unearths toward number of dangerous and perchance fateful interchanges occurring toward the system. Within character of guaran-teeing the slumberous, opening and uprightness count of to socialize for professional. The precise and fast categorization on to transaction evolves into indispensible. The effective capacity difficulty of all the intrusion detection simulation (IDS) simulates today at below discovery amount of fewer regular barrage associations and therefore the next warning rate. The container with systems of connections are reproduction everything beacon subject to the series of actions to achieve results accepts exists a contemporary well-known method. At the indicated motivation a hybrid methodology supported pairing distinct ripple transformation and human intelligence artificial neural network (ANN) for IDS is projected. The lack of balance of the situation traversing the space beyond information range was eliminated through synthetic minority oversampling technique-based oversampling have low regular object and irregular below examine of the dominant object. We are binding with three layer ANN is being used for classification, and thus the experimental results on knowledge discovery databases are being used for the facts in occurrence of accuracy rate and disclosure estimation toward identical period. True and false made up accepted.FindingsAt the indicated motivation a hybrid methodology supported pairing distinct ripple transformation and human intelligence ANN for IDS is projected. The lack of balance of the situation traversing the space beyond information range was eliminated through synthetic minority oversampling technique-based oversampling have low regular object and irregular below examine of the dominant object.Originality/valueChain interruption discovery is the series of actions for the results knowing the familiarity opening and honor number associate order, the scientific categorization undertaking become necessary. The capacity issues of invasion discovery is the order to determine and examine. The arrangement of simulations at the occasion under discovery estimation for low regular aggression associations and above made up feeling sudden panic amount.

Predicting particle collection performance of a wet electrostatic precipitator under varied conditions with artificial neural networks

The prediction of particle collection is important in designing and operating an electrostatic precipitator (ESP). In this work, a three–layer artificial neural network (ANN) was developed to predict the wet ESP performance under various conditions. The mean square error (MSE), standard deviation error (SDE), and variance account for (VAF) were used to evaluate the ANN model. The univariate analysis method was used to determine the contribution weight of each parameter. The MSE, SDE, and VAF for the trained ANN model were 0.0027%, 0.0362%, and 97.95%, respectively. The current decrease with operation time accounted for the inaccurate evaluation of wet ESP performance. The operating voltage, corona current, gas temperature, and residence time contributed 61.53%, 36.65%, 1.26%, and 0.56% to particle collection respectively. Furthermore, current and voltage dominated the particle collection in different regions. The research findings provide a valuable research approach to retrofit the design and operation of wet ESPs.

Integrating layered recurrent ANN with robust control strategy for diverse operating conditions of AGC of the power system

This study presents the structural, operational and control aspects of doubly fed induction generator (DFIG) based wind integrated power systems. The automatic generation control (AGC) of a meshed power system including DFIG-based wind turbines has been framed and investigations under various system perturbation are presented. The two-area system consisting of non-reheat thermal turbines with DFIG and interconnected through parallel AC/DC tie-lines is considered for the study. The system non-idealities such as governor lag and generation rate constraints are taken into consideration. An AGC strategy using a layered recurrent artificial neural network (ANN) is proposed in this work. The gains of the proposed AGC are obtained by effectively training the ANN using a set of reliable data obtained from a widespread range of operating system conditions using robust control strategy. The study also incorporates the design of AGC for the power system using the fuzzy logic concept and other AGC actions such as integral (I), proportional–integral (PI) and proportional–integral–derivative (PID) calculated via the means of particle swarm optimization (PSO). The results obtained with the proposed ANN created AGC are linked and demonstrated their superiority over fuzzy logic PI and traditional PSO-based I/PI/PID AGC strategies under numerous system operating conditions.

ELECTROSTATIC LENS SYSTEM DESIGN WITH THE ARTIFICIAL NEURAL NETWORKS

Yapay zekâ algoritmalarıyla son yıllarda birçok bilim dalında başarılı uygulamalar geliştirilmektedir. Deneysel veya benzetim programlarından elde edilen veriler söz konusu algoritmalarla işlenmektedir. Tasarlanan algoritma mimarileri ile veriler işlenerek tahmin ve sınıflandırma çalışmaları yapılmaktadır. Bu algoritmalardan, amaca ve veri kümesine uygun olan algoritmanın seçilmesi büyük önem taşımaktadır. Bu kapsamda, fizik alanındaki yenilikçi çalışmalarda yapay sinir ağı algoritması kullanmak yüksek performans değerleri elde etmeyi sağlamaktadır. Biyolojik nöronlardan esinlenen yapay sinir ağı (YSA), öğrenme yeteneğine sahip paralel hesaplama sistemidir. Bu çalışmada, üç katmanlı yapay sinir ağı kullanılarak beş elemanlı elektrostatik silindir lenslerin paralel demet modu belirlenmektedir. Çalışmada kullanılan veri kümesi, yüksek doğrulukta hesaplama yapabilen CPO(Yüklü Parçacık Optiği) programı yardımıyla elde edilmiştir. Verilerin analizi Matlab R2012b programı kullanılarak gerçekleştirilmiştir. Elde edilen sonuçlara göre, yapay sinir ağının fizik alanında paralel demet modunu belirlemede yüksek performans değerlerine sahip olan ve elektrostatik problem çözümlerinde sonlu fark ve sınır eleman metoduna alternatif bir metot olduğu ortaya konulmuştur. Oluşturulan YSA algoritması, test verilerinin %85,7’sini doğru olarak sınıflandırmıştır.

NeuroMem: Analog Graphene-Based Resistive Memory for Artificial Neural Networks

Artificial Intelligence (AI) at the edge has become a hot subject of the recent technology-minded publications. The challenges related to IoT nodes gave rise to research on efficient hardware-based accelerators. In this context, analog memristor devices are crucial elements to efficiently perform the multiply-and-add (MAD) operations found in many AI algorithms. This is due to the ability of memristor devices to perform in-memory-computing (IMC) in a way that mimics the synapses in human brain. Here, we present a novel planar analog memristor, namely NeuroMem, that includes a partially reduced Graphene Oxide (prGO) thin film. The analog and non-volatile resistance switching of NeuroMem enable tuning it to any value within the RON and ROFF range. These two features make NeuroMem a potential candidate for emerging IMC applications such as inference engine for AI systems. Moreover, the prGO thin film of the memristor is patterned on a flexible substrate of Cyclic Olefin Copolymer (COC) using standard microfabrication techniques. This provides new opportunities for simple, flexible, and cost-effective fabrication of solution-based Graphene-based memristors. In addition to providing detailed electrical characterization of the device, a crossbar of the technology has been fabricated to demonstrate its ability to implement IMC for MAD operations targeting fully connected layer of Artificial Neural Network. This work is the first to report on the great potential of this technology for AI inference application especially for edge devices.

N-hidden layer artificial neural network architecture computer code: geophysical application example

We provide a MATLAB computer code for training artificial neural network (ANN) with N+1 layer (N-hidden layer) architecture. Currently, the ANN application to solving geophysical problems have been confined to the 2-layer, i.e. 1-hidden layer, architecture because there are no open source software codes for higher numbered layer architecture. The restriction to the 2-layer architecture comes with the attendant model error due to insufficient hidden neurons to fully define the ANN machines. The N-hidden layer ANN has a general architecture whose sensitivity is the accumulation of the backpropagation of the error between the feedforward output and the target patterns. The trained ANN machine can be retrieved by the gradient optimization method namely: Levenberg-Marquardt, steepest descent or conjugate gradient methods. Our test results on the Poisson's ratio (as a function of compressional and shear wave velocities) machines with 2-, 3- and 4-layer ANN architectures reveal that the machines with higher number of layers outperform those with lower number of layers. Specifically, the 3- and 4-layer ANN machines have ≥97% accuracy, predicting the lithology and fluid identification in the oil and gas industry by means of the Poisson's ratio, whereas the 2-layer ANN machines poorly predict the results with as large error as 20%. These results therefore reinforce our belief that this open source code will facilitate the training of accurate N-hidden layer ANN sophisticated machines with high performance and quality delivery of geophysical solutions. Moreover, the easy portability of the functions of the code into other software will enhance a versatile application and further research to improve its performance.