Usage Of Networks Research Articles

Optimization of Process Parameters in Laser Powder Bed Fusion of SS 316L Parts Using Artificial Neural Networks

Additive manufacturing is rapidly evolving and revolutionizing the fabrication of complex metal components with tunable properties. Machine learning and neural networks have emerged as powerful tools for process–property optimization in additive manufacturing. These techniques work well for the prediction of a single property but their applicability in optimizing multiple properties is limited. In the present work, an exclusive neural network is developed to demonstrate the potential of a single neural network in optimizing multiple part properties. The model is used to identify the optimal process parameter values for laser power, scan speed, and hatch spacing for the required surface roughness, relative density, microhardness, and dimensional accuracy in stainless steel parts. In-house-generated experimental data are used to train the model. The model has seven neurons in the hidden layer, which are selected using hyperparameter optimization. K-fold cross-validation is performed to ensure the robustness of the model, which results in a mean squared error of 0.0578 and R2 score of 0.59. The developed model is then used to predict the optimal process parameters corresponding to the user-required part properties. The model serves as a significant pre-processing step to identify the best parameters before printing, thus saving time and costs for repeated part fabrication. The study provides more insights into the usage of a single artificial neural network for the optimization of multiple properties of printed metal parts.

Air Pollution Monitoring via Wireless Sensor Networks: The Investigation and Correction of the Aging Behavior of Electrochemical Gaseous Pollutant Sensors

The continuously growing human activity in large and densely populated cities pollutes air and consequently puts public health in danger. This is why air quality monitoring is necessary in all urban environments. However, the creation of dense air monitoring networks is extremely costly because it requires the usage of a great number of air monitoring stations that are quite expensive. Instead, the usage of wireless sensor networks (WSNs) that incorporate low-cost electrochemical gas sensors provides an excellent alternative. Actually, sensors of this kind that are recommended for low-cost air quality monitoring applications may provide relatively precise measurements. However, the reliability of such sensors during their operational life is questionable. The research work presented in this article not only experimentally examined the correlation that exists between the validity of the measurements obtained from low-cost gas sensors and their aging, but also proposes novel corrective formulae for gas sensors of two different types (i.e., NO2, O3), which are aimed at alleviating the impact of aging on the accuracy of measurements. The following steps were conducted in order to both study and lessen the aging of electrochemical sensors: (i) a sensor network was developed to measure air quality at a place near official instruments that perform corresponding measurements; (ii) the collected data were compared to the corresponding recordings of the official instruments; (iii) calibration and compensation were performed using the electrochemical sensor vendor instructions; (iv) the divergence between the datasets was studied for various periods of time and the impact of aging was studied; (v) the compensation process was re-evaluated and new compensation coefficients were produced for all periods; (vi) the new compensation coefficients were used to shape formulae that automatically calculate the new coefficients with respect to the sensors’ aging; and (vii) the performance of the overall procedure was evaluated through the comparison of the final outcomes with real data.

CNNParted: An open source framework for efficient Convolutional Neural Network inference partitioning in embedded systems

Applications such as autonomous driving or assistive robotics heavily rely on the usage of Deep Neural Networks. In particular, Convolutional Neural Networks (CNNs) provide precise and reliable results in image processing tasks like camera-based object detection or semantic segmentation. However, to achieve even better results, CNNs are becoming more and more complex. Deploying these networks in distributed embedded systems thereby imposes new challenges, due to additional constraints regarding performance and energy consumption in the near-sensor compute platforms, i.e. the sensor nodes. Processing all data in the central node, however, is disadvantageous since raw data of camera consumes large bandwidth and running CNN inference of multiple tasks requires certain performance. Moreover, sending raw data over the interconnect is not advisable for privacy reasons. Hence, offloading CNN workload to the sensor nodes in the system can lead to reduced traffic on the link and a higher level of data security.However, due to the limited hardware-resources on the sensor nodes, partitioning CNNs has to be done carefully to meet overall latency requirements and energy constraints. Therefore, we present CNNParted, an open-source framework for efficient, hardware-aware CNN inference partitioning targeting embedded AI applications. It automatically searches for potential partitioning points in the CNN to find a beneficial workload distribution between sensor nodes and a central edge node. Thereby, CNNParted not only analyzes the CNN architecture but also takes hardware components, such as dedicated hardware accelerators and memories, into consideration to evaluate inference partitioning regarding latency and energy consumption.Exemplary, we apply CNNParted to three commonly used feed forward CNNs in embedded systems. Thereby, the framework first searches for several potential partitioning points and then evaluates the latter regarding inference latency and energy consumption. Based on the results, beneficial partitioning points can be identified depending on the system constraints. Using the framework, we are able to find and evaluate 10 potential partitioning points for FCN ResNet-50, 13 partitioning points for GoogLeNet, and 8 partitioning points for SqueezeNet V1.1 within 520 s, 330 s, and 140 s, respectively, on an AMD EPYC 7702P running 8 concurrent threads. For GoogLeNet, we determine two partitioning points that provide a good trade-off between required bandwidth, latency and energy consumption. We also provide insights into further interesting findings that can be derived from the evaluation results.

A New Robust Integral Reinforcement Learning Based Control Algorithm for Interleaved DC/DC Boost Converter

This article proposes a novel online integral reinforcement learning (IRL) based data-driven control algorithm for interleaved dc/dc boost converter. The proposed algorithm is independent of system model due to the usage of a three-layer neural network (NN). Furthermore, its controller gains are autonomously adjusted online through the value function based NN weights updating mechanism, which simplifies the controller gain tuning process. Compared to the conventional model-dependent control approaches, it provides superior control performance. Additionally, the proposed method contributes to significantly reduce the computational burden of classical IRL algorithm by removing the disturbance updating process. Experimental results are presented to verify the efficacy of the proposed algorithm.

Intelligent recognition and integration of grapical helements into virtual surrounding within augmented reality using hybrid convolutional neural networks.

In this paper analysis of modern augmented reality algorithms based on mobile devices was done. As a result, algorithmic shortcomings were identified and the usage of convolutional neural networks was proposed. Within the research the qualitative analysis of modern architectures of convolutional neural networks was carried out and their separate shortcomings at use in systems on the basis of processor architecture ARM was shown. As a result of this research it was found that to achieve the target accuracy and speed of the system it is important to use a hybrid convolutional neural network, which significantly improves the quality criteria of the system. The optimal structure and parameters for initialization and training of a hybrid convolutional neural network system used for augmented reality are obtained. The optimal training sample was formed and the use of pre-trained HCNN on another device of ARM architecture was described.

Isogeometric multi-patch topology optimization based on pix2pix

We present a novel approach that combines the power of pix2pix, an image-to-image translation framework, with the advanced capabilities of isogeometric multi-patch analysis for topology optimization. The proposed method adds the Nitsche’s methods into the advantages of Isogeometric analysis (IGA), thus gaining the ability to handle complex geometries by generating locally smooth and well-converged results. Additionally, the usage of generative adversarial network based pix2pix allows for a more efficient representation of the design space, reducing the computational cost of the optimization process. This approach has shown promising results in various numerical examples. This technique aims to improve the efficiency of conceptual design in complex engineering applications.

Usage of 3D U-net Convolutional Neural Network for the inspection of aerospace components

This work illustrates the use of deep learning methods applied on X-ray computed tomography (XCT) datasets to segment pores and fibres in reinforced composite components from the aeronautic industry by binary semantic segmentation. We first apply data pre-processing, and then employ a modified 3D U-Net, representing a convolutional neural network. Tweaking hyper-parameters, we have reached an optimal model for our datasets. One of the models has reached 99% segmentation accuracy when testing using a Dice function. In our experiments, pores and fibres in XCT datasets of aerospace components, more specifically of glass and carbon fibre reinforced composites, were segmented and analysed. In order to compare this modified 3D U-Net architecture with segmentation methods currently used in the industry, the datasets were also input to conventional Otsu thresholding. Our results shows that modified 3D U-Net performs better than Otsu thresholding, especially on the segmentation of small pores. Modified 3D U-Net also showed reasonable prediction accuracy when testing with an optimised model which was trained with a low number of dataset both for fibre and pore segmentation.

An Energy-Efficient Clustering and Fuzzy-Based Path Selection for Flying Ad-Hoc Networks

Flying Ad-hoc Networks (FANET) allow for an ad-hoc networking among Unmanned Aerial Vehicles (UAV), have recently gained popularity in a variety of military and non-militant applications. The existing work used the Glowworm Swarm Optimization (GSO) algorithm to create a self-organization depending on clustering technique for FANET. Owing to UAV increased mobility, network topology might vary over time, providing route discovery and maintenance is one of the most difficult tasks. And also, the network throughput is still more worsened by the network congestion. To solve this problem, the proposed work designed an energy efficient clustering and fuzzy-based path selection for FANET. In this work, initially, the clustering is performed using the UAV distance. For efficient communication and energy consumption, optimal selection of Cluster Head (CH) is performed by using Adaptive Mutation with Teaching-Learning-Based Optimization (AMTLBO) algorithm. To improve the optimal selection of CH nodes, the best fitness values are calculated. The fitness function depends on Link capacity, remaining energy and neighbor UAV distance. Next to that, nodes begin communications as well as transmit their information to their CH. Improved Fuzzy-based Routing (IFR) is introduced for improving the route discovery process. The goal is to find routes that have a high level of flying autonomy, minimal mobility, and a higher Received Signal Strength Indicator (RSSI). As a result, the energy usage of network is decreased, as well as the cluster’s lifespan is extended. Finally, an adaptive and reliable congestion detection mechanism is introduced to transmit the packets with congestion free path. The experimental result shows that the proposed AMTLBO system attains higher performance compared to the existing system in terms of energy usage, throughput, delay, overhead and packet delivery ratio.

Neural architecture search for energy-efficient always-on audio machine learning

Mobile and edge computing devices for always-on classification tasks require energy-efficient neural network architectures. In this paper we present several changes to neural architecture searches that improve the chance of success in practical situations. Our search simultaneously optimizes for network accuracy, energy efficiency and memory usage. We benchmark the performance of our search on real hardware, but since running thousands of tests with real hardware is difficult, we use a random forest model to roughly predict the energy usage of a candidate network. We present a search strategy that uses both Bayesian and regularized evolutionary search with particle swarms, and employs early stopping to reduce the computational burden. Our search, evaluated on a sound event classification dataset based upon AudioSet, results in an order of magnitude less energy per inference and a much smaller memory footprint than our baseline MobileNetV1/V2 implementations while slightly improving task accuracy. We also demonstrate how combining a 2D spectrogram with a convolution with many filters causes a computational bottleneck for audio classification and that alternative approaches reduce the computational burden but sacrifice task accuracy.

Rainfall prediction using time-delay wavelet neural network (TDWNN) model for assessing agrometeorological risk

In an agriculturally dependent nation like India, accurate and effective rainfall forecasting methods are crucial for assessing agrometeorological risk. Forecasting rainfall is perhaps one amongst the most arduous tasks in this context due to the prevalence of a non-linear pattern. One of the most promising and frequently employed approaches for forecasting rainfall data is the Time-Delay Neural Network (TDNN) model. TDNN's non-parametric, data-driven, and self-adaptive characteristics make it increasingly attractive for modelling nonlinear dynamics and generating nonlinear forecasts. Nevertheless, since the conventional TDNN uses the sigmoid activation function, there is always a chance that the training process may converge to local minima. This study addresses the usage of a Time-delay wavelet neural network (TDWNN), which employs a TDNN architecture with a hidden layer activation function derived from the orthonormal wavelet family, in order to circumvent this issue. TDWNN has been empirically demonstrated using annual rainfall data from two districts in the Indian state of West Bengal. According to the findings of this study, the TDWNN model is superior than the conventional TDNN method to assess agrometeorological risk.

FE{}^textrm{ANN}: an efficient data-driven multiscale approach based on physics-constrained neural networks and automated data mining

Herein, we present a new data-driven multiscale framework called FE{}^textrm{ANN} which is based on two main keystones: the usage of physics-constrained artificial neural networks (ANNs) as macroscopic surrogate models and an autonomous data mining process. Our approach allows the efficient simulation of materials with complex underlying microstructures which reveal an overall anisotropic and nonlinear behavior on the macroscale. Thereby, we restrict ourselves to finite strain hyperelasticity problems for now. By using a set of problem specific invariants as the input of the ANN and the Helmholtz free energy density as the output, several physical principles, e. g., objectivity, material symmetry, compatibility with the balance of angular momentum and thermodynamic consistency are fulfilled a priori. The necessary data for the training of the ANN-based surrogate model, i. e., macroscopic deformations and corresponding stresses, are collected via computational homogenization of representative volume elements (RVEs). Thereby, the core feature of the approach is given by a completely autonomous mining of the required data set within an overall loop. In each iteration of the loop, new data are generated by gathering the macroscopic deformation states from the macroscopic finite element simulation and a subsequently sorting by using the anisotropy class of the considered material. Finally, all unknown deformations are prescribed in the RVE simulation to get the corresponding stresses and thus to extend the data set. The proposed framework consequently allows to reduce the number of time-consuming microscale simulations to a minimum. It is exemplarily applied to several descriptive examples, where a fiber reinforced composite with a highly nonlinear Ogden-type behavior of the individual components is considered. Thereby, a rather high accuracy could be proved by a validation of the approach.

Reconfigurable Architecture for Noise Cancellation in Acoustic Environment Using Single Multiply Accumulate Adaline Filter

The creation of multiple applications with a higher level of complexity has been made possible by the usage of artificial neural networks (ANNs). In this research, an efficient flexible finite impulse response (FIR) filter structure called ADALINE (adaptive linear element) that makes use of a MAC (multiply accumulate) core is proposed. The least mean square (LMS) and recursive least square (RLS) algorithms are the most often used methods for maximizing filter coefficients. Despite outperforming the LMS, the RLS approach has not been favored for real-time applications due to its higher design arithmetic complexity. To achieve less computation, the fundamental filter has utilized an LMS-based tapping delay line filter, which is practically a workable option for an adaptive filtering algorithm. To discover the undiscovered system, the adjustable coefficient filters have been developed in the suggested work utilizing an optimal LMS approach. The 10-tap filter being considered here has been analyzed and synthesized utilizing field programmable gate array (FPGA) devices and programming in hardware description language. In terms of how well the resources were used, the placement and postrouting design performed well. If the implemented filter architecture is compared with the existing filter architecture, it reveals a 25% decrease in resources from the existing one and an increase in clock frequency of roughly 20%.

Machine Learning and Deep Learning Techniques for Spectral Spatial Classification of Hyperspectral Images: A Comprehensive Survey

The growth of Hyperspectral Image (HSI) analysis is due to technology advancements that enable cameras to collect hundreds of continuous spectral information of each pixel in an image. HSI classification is challenging due to the large number of redundant spectral bands, limited training samples and non-linear relationship between the collected spatial position and the spectral bands. Our survey highlights recent research in HSI classification using traditional Machine Learning techniques like kernel-based learning, Support Vector Machines, Dimension Reduction and Transform-based techniques. Our study also digs into Deep Learning (DL) techniques that involve the usage of Autoencoders, 1D, 2D and 3D-Convolutional Neural Networks to classify HSI. From the comparison, it is observed that DL-based classification techniques outperform ML-based techniques. It has also been observed that spectral-spatial HSI classification outperforms pixel-by-pixel classification because it incorporates spectral signatures and spatial domain information. The performance of ML and DL-based classification techniques has been reviewed on commonly used land cover datasets like Indian Pines, Salinas valley and Pavia University.

Detecting dementia from speech and transcripts using transformers

Alzheimer’s disease (AD) constitutes a neurodegenerative disease with serious consequences to peoples’ everyday lives, if it is not diagnosed early since there is no available cure. Alzheimer’s is the most common cause of dementia, which constitutes a general term for loss of memory. Due to the fact that dementia affects speech, existing research initiatives focus on detecting dementia from spontaneous speech. However, little work has been done regarding the conversion of speech data to Log-Mel spectrograms and Mel-frequency cepstral coefficients (MFCCs) and the usage of pretrained models. Concurrently, little work has been done in terms of both the usage of transformer networks and the way the two modalities, i.e., speech and transcripts, are combined in a single neural network. To address these limitations, first we represent speech signal as an image and employ several pretrained models, with Vision Transformer (ViT) achieving the highest evaluation results. Secondly, we propose multimodal models. More specifically, our introduced models include Gated Multimodal Unit in order to control the influence of each modality towards the final classification and crossmodal attention so as to capture in an effective way the relationships between the two modalities. Extensive experiments conducted on the ADReSS Challenge dataset demonstrate the effectiveness of the proposed models and their superiority over state-of-the-art approaches.

Strategies for reducing polarization in social networks

The widespread usage of social networks has intensified the phenomenon of opinion polarization. This polarization has a corrosive effect on the functioning of public opinions, societies, and democracies. Hence, it is vital to devise strategies for reducing polarization. In this paper, we propose a depolarization model based on the modified linear threshold model and the Hegselmann-Krause model. Using this model, we then devise three depolarization strategies (building heterogeneous edges, embedding local neutral opinion, and introducing global neutral opinion), and conduct depolarization experiments with varied parameters, including self-belief parameter, social influence moderating parameter, and the selection mode of new edges and active nodes. Our experiments with synthesized and real-world datasets have yielded some interesting results.

PTcomp: Post-Training Compression Technique for Generative Adversarial Networks

In a time of virtual spaces, the usage of generative adversarial networks is inevitable. Generative adversarial networks (GANs) are generative deep-learning models that can generate realistic data. GANs have been used in many applications like text-to-image, image-to-image, image synthesis, speech synthesis, etc. Its power lies in the diversity and novelty of the generated data. Despite their advantages, GANs are resource-hungry. GANs’ output resolution and high correlation make it more challenging to compress and fit on edge-devices storage and power budget. Hence, traditional compression techniques are not the best fit to use with GANs. Additionally, GANs training instability adds another dimension of difficulty. Therefore, compression techniques that require retraining are challenging for GANs. In this paper, we developed a weight clustering technique to compress GANs without the need for retraining, hence the name post-training compression technique (PTcomp). We also proposed a clustered-based pruning which adds more savings. Experiments on Cyclegan, Deep convolution gan (DCGAN), and Stargan using several datasets show the superiority of our technique against traditional post-training quantization. Our technique provides a 4x to 8x compression ratio with comparable quality to original models and 14% fewer mac operations due to pruning.

Advancements and Challenges in Energy-efficient 6G Mobile Communication Network

The arrival of 6G mobile communication networks is anticipated to revolutionize the technological landscape, bringing about profound innovations. This research paper explores the various technological advancements that will pave the way for the advent of 6G networks, with a particular focus on addressing energy consumption. It is widely recognized that energy efficiency plays a crucial role in the evolution of 6G networks. To enhance network performance, user experience, and resource management, the integration of Artificial Intelligence (AI) is expected to be a pivotal technology. AI-based solutions can effectively optimize energy usage and contribute to the overall efficiency of 6G networks. Furthermore, the incorporation of wireless communication systems, telecommunication, and the Internet of Things (IoT) will be integral to the infrastructure of 6G networks. The need for significant enhancements in 6G networks is also examined in this study. Ensuring the safety and protection of 6G networks from cyber threats becomes increasingly important due to the growing reliance on networked communication and the sensitive nature of transmitted information. Cutting-edge security methods such as homomorphic encryption and blockchain technology may be essential in this regard. Moreover, this research paper explores the impact of 6G networks on various domains and discusses the challenges that must be overcome to unlock the technology’s full potential. To ensure responsible adoption and usage of 6G networks, the development of new business models and regulatory frameworks may be necessary to support their implementation while addressing energy consumption concerns.

Deep Neural Network Based Cardio Vascular Disease Prediction Using Binarized Butterfly Optimization

In this digital era, Cardio Vascular Disease (CVD) has become the leading cause of death which has led to the mortality of 17.9 million lives each year. Earlier Diagnosis of the people who are at higher risk of CVDs helps them to receive proper treatment and helps prevent deaths. It becomes inevitable to propose a solution to predict the CVD with high accuracy. A system for predicting Cardio Vascular Disease using Deep Neural Network with Binarized Butterfly Optimization Algorithm (DNN–BBoA) is proposed. The BBoA is incorporated to select the best features. The optimal features are fed to the deep neural network classifier and it improves prediction accuracy and reduces the time complexity. The usage of a deep neural network further helps to improve the prediction accuracy with minimal complexity. The proposed system is tested with two datasets namely the Heart disease dataset from UCI repository and CVD dataset from Kaggle Repository. The proposed work is compared with different machine learning classifiers such as Support Vector Machine, Random Forest, and Decision Tree Classifier. The accuracy of the proposed DNN–BBoA is 99.35% for the heart disease data set from UCI repository yielding an accuracy of 80.98% for Kaggle repository for cardiovascular disease dataset.

Method for Fault Diagnosis and Speed Control of PMSM

In the field of fault tolerance estimation, the increasing attention in electrical motors is the fault detection and diagnosis. The tasks performed by these machines are progressively complex and the enhancements are likewise looked for in the field of fault diagnosis. It has now turned out to be essential to diagnose faults at their very inception; as unscheduled machine downtime can upset deadlines and cause heavy financial burden. In this paper, fault diagnosis and speed control of permanent magnet synchronous motor (PMSM) is proposed. Elman Neural Network (ENN) is used to diagnose the fault of permanent magnet synchronous motor. Both the fault location and fault severity are considered. In this, eccentricity fault may occur in the motor. To control the speed of the permanent magnet synchronous motor, Dolphin Swarm Optimization (DSO) algorithm is used. The proposed work is simulated by using MATLAB in terms of amplitude, speed and torque. The comparison graph of speed vs. torque obtained by the proposed method gives better result compared to the other existing techniques. The proposed work is also compared with Particle Swarm Optimization (PSO) and Elephant Herding Optimization (EHO) algorithm. The proposed usage of Elman Neural Network to detect the fault and the usage of Dolphin Swarm Optimization algorithm to control the speed of the permanent magnet synchronous motor gives better outcome.

An Extrapolation Approach for RF-EMF Exposure Prediction in an Urban Area using Artificial Neural Network

The prediction of the electric (E) field plays an important role in the monitoring of the radiofrequency electromagnetic field (RF-EMF) exposure induced by cellular networks. In this paper, we present an approach to extrapolate the E field in an urban area using artificial neural network. We first apply a moving average method over a sliding window to average out the EMF random fluctuations and remove the noise produced during the drive test recording of E field measurements along the route. Using public accessed datasets, i.e., cartoradio and OpenStreetMap, we then extract relevant features, including the ones that have a relation with the number of active antennas and those used by Bertoni-Walfisch propagation model. By applying the Gram-Schmidt Orthogonalization procedure, we select the best subset of the extracted features as inputs to the artificial neural network (ANN). In this work, two disjoint subsets are selected for the learning and testing phases to evaluate the performance of our proposal in the extrapolation of the E field measurements and to quantify the uncertainty generated by the proposed predictor due to the dynamicity usage of cellular networks and geolocalization inaccuracies.