Error Back-propagation Algorithm Research Articles

Comparative study of the effectiveness of time series and arima modeling x RNA approach in a Multi Layer Perceptron (MLP) model in predicting fouled in petroleum pre-processing heat exchangers

This work presents a comparison between two approaches based on Artificial Intelligence techniques to predict critical deposition in the battery of shell and tube heat exchangers, used in petroleum pre-processing. Deposition or incrustation, which occurs on the thermal exchange surfaces of heat exchangers, during operation, can reduce the efficiency of this equipment and generate an increase in the cost of refining for the petrochemical plant due to the modification of ideal operating parameters and changes. process variables, such as flow, pressure and temperature, input and output of heat exchangers, as well as thermodynamic parameters, such as fouling resistance coefficient and global heat transfer coefficient. The proposed methods for comparing results used the analysis and prediction of time series based mainly on integrated autoregressive moving average (ARIMA) and Deep Learning models and, on the other hand, an artificial neural network (ANN) model based on the Multi Layer architecture. Perceptron (MLP) with the Backpropagation error backpropagation algorithm. The data was collected in exchangers of a petroleum preheating battery in the years 2014 to 2021. The general objective of this work is to present the results of both approaches and compare, since they aim to predict the deposition in the heat exchanger of the shell and tube type, thus helping to minimize maintenance costs and increase the energy efficiency of the petrochemical plant, making operations safer and more efficient, bringing significant benefits to the petroleum industry.

Application programming interface for electron beam welding of Inconel 718 thin plates

In this paper the simulation of electron beam welding of Inconel 718 thin plates by a moving linear heat source is considered. Neural network models are developed and used for the description of the dependence of the molten pool geometry characteristics on the process parameters - electron beam power, welding velocity, and the plate thickness. Neural network models, based on a multi-layered feedforward neural network, trained with Levenberg-Marquardt error backpropagation algorithm are compared with estimated regression models. The neural networks are trained, tested and validated using a set of experimental data. The resulting models are implemented for developing of an application programming interface for electron beam welding of Inconel 718 thin plates, which is used for prediction, investigation and graphical optimization of the molten pool geometry characteristics of the obtained electron beam welds.

Synthesis of a measurement procedure for estimating the orientation of a small unmanned aerial vehicle under conditions of changing status of measurement results

The problem of processing measurement information with changing status of measurement results of micromechanical sensors of an intelligent on-board measuring system of a small unmanned aerial vehicle is considered. While vehicle is in the air, the measurement results status changes from confirmed to orienting, for example, due to sensor defects, degradation of the measuring channel, the appearance of false measurement output signals of micromechanical sensors due to vibrations and shocks caused by the movement of air masses. As a result, the probability of stability loss of a small unmanned aerial vehicle increases and it is necessary to raise the accuracy of estimating its orientation in conditions of changing status of measurement results. The measuring procedure of the Kalman structure is considered, the equations of which are determined with accuracy to the parameters of the transition and noise matrices of the state, as well as the perturbation vector, characterizing the measuring process. The parameter values are determined by a mathematical model for converting measuring information based on a dynamic mathematical model, which distinguishes the developed measuring procedure from a measuring procedure with a classical transition matrix. A neural network is used to find unknown parameters. A multilayer perceptron was selected as the basis of a neural network, for which an error back propagation algorithm is used to train. Based on the results of mathematical modeling and measurement experiment, it was found that the accuracy of the synthesized measuring procedure based on a dynamic mathematical model is higher than the accuracy of the measuring procedure of the Kalman structure with a classical transition matrix. The results of the study will be useful in the development of intelligent measurement procedures for on-board measurement systems operating under conditions of changing status of measurement results

Student sports data analysis and physical fitness evaluation based on convolutional neural networks

<p>In order to achieve the analysis of student sports data and physical fitness evaluation, the author proposes a method based on convolutional neural networks. A hybrid algorithm combining genetic algorithm and error backpropagation algorithm (BP) is used to train convolutional neural networks. The algorithm first uses genetic algorithm for global training, and then uses BP algorithm for local precise training. This overcomes the drawbacks of traditional BP networks such as long training time and frequent local atmospheric drift, and improves global circulation performance. A neural network model was established to display the relationship between the total physical activity score and multiple test scores of high school students by utilizing electrical networks to demonstrate the connectivity of the neural network. This model aims to evaluate the athletic performance of college students and compare the results with other experimental models. The results indicate that the neural network-based model for evaluating college student physical activity can reflect the differences in physical activity and scores among all students, making it a suitable standard for evaluating high school student physical activity. The fitting accuracy of deterministic neural network models is higher than that of multiple linear regression models, which means that neural network models better reflect the performance of the network. The accuracy of various indicators of student physical fitness and total score makes the model easy to operate, accurate to predict, and effective analysis is scientifically reasonable.</p>

Sparse-firing regularization methods for spiking neural networks with time-to-first-spike coding

The training of multilayer spiking neural networks (SNNs) using the error backpropagation algorithm has made significant progress in recent years. Among the various training schemes, the error backpropagation method that directly uses the firing time of neurons has attracted considerable attention because it can realize ideal temporal coding. This method uses time-to-first-spike (TTFS) coding, in which each neuron fires at most once, and this restriction on the number of firings enables information to be processed at a very low firing frequency. This low firing frequency increases the energy efficiency of information processing in SNNs. However, only an upper limit has been provided for TTFS-coded SNNs, and the information-processing capability of SNNs at lower firing frequencies has not been fully investigated. In this paper, we propose two spike-timing-based sparse-firing (SSR) regularization methods to further reduce the firing frequency of TTFS-coded SNNs. Both methods are characterized by the fact that they only require information about the firing timing and associated weights. The effects of these regularization methods were investigated on the MNIST, Fashion-MNIST, and CIFAR-10 datasets using multilayer perceptron networks and convolutional neural network structures.

Intelligent Vehicle Inspection Tool Design Based on Freeman Chain Code for Automatic Annotation of 3D Models

Autonomous vehicle are more and more widely used in daily life, and the requirements for their safety performance are higher and higher. As a tool for testing auto parts, intelligent inspection tools are crucial to the guarantee of automobile quality. However, traditional fixture design relies on manual drawing, which is inefficient and prone to errors. To solve this problem, this research uses Freeman chain code to determine the annotation object, uses case clustering method to annotate, and uses error back propagation algorithm to realize case knowledge classification learning, and designs intelligent vehicle inspection tool design technology based on Freeman chain code 3D automatic annotation method. The experimental results show that the geometric feature matching results are correct, and the difference in feature comparison results is significant, with a high accuracy rate. Meanwhile, the geometric similarity annotation method has a high accuracy rate, taking only 3 minutes to complete the annotation, which is 7 minutes longer than traditional manual annotation. The error backpropagation algorithm can accurately achieve feature classification, and the design time of size chain inspection tool deformation design is reduced by 214 minutes compared to manual reverse deformation design, significantly improving design efficiency. In summary, the proposed design method for automotive inspection tools can achieve automatic model annotation, improve design efficiency, and reduce design time.

Разработка искусственной нейронной сети для классификации ЭКГ

The paper presents a solution to the problem of developing an artificial neural network for classifying electrocardiograms in order to identify a patient's cardiovascular diseases. A literary review of the use of artificial neural networks for use in medicine in the determination of various cardiac pathologies in the condition of patients has been carried out. The most significant six parameters of the heart rhythm have been determined, which can be used in neural network modeling of the state of the human cardiovascular system. An analysis of the capabilities of well-known commercial programs for the classification of electrocardiograms was carried out, their advantages and disadvantages were identified. The capabilities of modern devices for taking cardiograms are described with an indication of the accuracy of their measurements. A study of three different structures of the desired neural network was carried out and the best option was chosen. The software package is made in Python in the Pycharm 2023.1 environment. The network was trained on the basis of the error backpropagation algorithm, the Adam method served as an optimizer. The standard deviation of the network output signal from the standard was used as a criterion. The Softmax function used in the output layer is used to convert the output of neurons into probabilities. Thus, a neural mathematical model for detecting heart diseases has been developed, which has low system requirements for a computer, which can be used to develop software for wearable electronics. An artificial neural network has the following metric: accuracy - 90%, sensitivity - 90%, specificity - 10%.

Forward-forward training of an optical neural network.

Neural networks (NNs) have demonstrated remarkable capabilities in various tasks, but their computation-intensive nature demands faster and more energy-efficient hardware implementations. Optics-based platforms, using technologies such as silicon photonics and spatial light modulators, offer promising avenues for achieving this goal. However, training multiple programmable layers together with these physical systems poses challenges, as they are difficult to fully characterize and describe with differentiable functions, hindering the use of error backpropagation algorithm. The recently introduced forward-forward algorithm (FFA) eliminates the need for perfect characterization of the physical learning system and shows promise for efficient training with large numbers of programmable parameters. The FFA does not require backpropagating an error signal to update the weights, rather the weights are updated by only sending information in one direction. The local loss function for each set of trainable weights enables low-power analog hardware implementations without resorting to metaheuristic algorithms or reinforcement learning. In this paper, we present an experiment utilizing multimode nonlinear wave propagation in an optical fiber demonstrating the feasibility of the FFA approach using an optical system. The results show that incorporating optical transforms in multilayer NN architectures trained with the FFA can lead to performance improvements, even with a relatively small number of trainable weights. The proposed method offers a new path to the challenge of training optical NNs and provides insights into leveraging physical transformations for enhancing the NN performance.

Modern synergetic neural network for imbalanced small data classification

Deep learning’s performance on the imbalanced small data is substantially degraded by overfitting. Recurrent neural networks retain better performance in such tasks by constructing dynamical systems for robustness. Synergetic neural network (SNN), a synergetic-based recurrent neural network, has superiorities in eliminating recall errors and pseudo memories, but is subject to frequent association errors. Since the cause remains unclear, most subsequent studies use genetic algorithms to adjust parameters for better accuracy, which occupies the parameter optimization space and hinders task-oriented tuning. To solve the problem and promote SNN’s application capability, we propose the modern synergetic neural network (MSNN) model. MSNN solves the association error by correcting the state initialization method in the working process, liberating the parameter optimization space. In addition, MSNN optimizes the attention parameter of the network with the error backpropagation algorithm and the gradient bypass technique to allow the network to be trained jointly with other network layers. The self-learning of the attention parameter empowers the adaptation to the imbalanced sample size, further improving the classification performance. In 75 classification tasks of small UC Irvine Machine Learning Datasets, the average rank of the MSNN achieves the best result compared to 187 neural and non-neural network machine learning methods.

Prediction of Women’s Fashion Product Expense in Hanoi City Using Artificial Neural Network

A person's fashion product expense plays an important role in building a garment company's product and pricing strategy. This article presents the results of predicting the fashion expenses of women aged 25 to 55 in Hanoi city. A linear multivariable model determined by the Bayesion Model Average (BMA) was applied to predict expenditures based on a set of 150 women survey data. The 3 layers feed-forward neural network model with 8 input neurons, 4 (+1) hidden layer neurons, and 1 output layer neuron, The Sigmoid activation function, trained by the error backpropagation algorithm has been established on software R specifically predicts the value of women's fashion expense with the above data set. The results show that the artificial neural network (ANN) has been set up allowing to prediction of fashion expenses accurately. The expenditure values predicted by the ANN and the actual value are linearly correlated with R2 = 0.987; The prediction results by ANN are more accurate than the defined multivariable linear model.

A Novel Design of Error Backpropagation Algorithm for Ingredient Mixing Process Tamarind Turmeric Herb

The goal of this study is to determine the best picture pattern for the tamarind turmeric herb. So far, the taste and color of tamarind turmeric herb have not been consistent, as they are impacted by maturity, and the amount of Turmeric used. The error backpropagation technique, which is commonly used in Content-based image retrieval systems, will be used to recognize image patterns. The main goal is to capture various portions of the tamarind turmeric herb during the extraction procedure. The camera is used to classify the tamarind turmeric herb product, process it into 5x5 pixels, and average the RGB value to obtain stable RGB values in each category, which are then fed into the Error Backpropagation algorithm. The most appropriate and fastest Error Backpropagation algorithm procedure will be found and implemented in a real-time computer. The first way will be to train the algorithm with ten data by changing neurons, layer, momentum, and learning rate, and the second technique will be to test the algorithm with ten data. The results of the training and testing procedure show that the two hidden layers can recognize 100% of inputs, with three input layers for R, G, and B values, ten neurons in the first and second hidden layers, and one output layer with Learning rate 0.5 and Momentum 0.6 as a parameter. Dark yellow is the best image pattern standard for tamarind turmeric herb, with RGB values in the range from 255, 103, 32 to 255, 128, 48.

BD-CVSA: A Broadband Direction Finding Method Based on Constructing Virtual Sparse Arrays

As the electromagnetic environment becomes more complex, radio direction finding systems need to cover a wide frequency range. However, the phase ambiguity of high frequency limits the development of broadband direction finding technology. The existing methods in solving the phase ambiguity problem heavily rely on designing the precise actual sparse array structure by theoretical derivation, which is expensive and time-consuming. To solve this problem, we propose a broadband direction finding method based on constructing virtual sparse arrays. First, we perform the convolutional operation on the direction feature matrix to achieve the weighted summation of the original baselines, which generates numerous virtual array elements and virtual baselines with different directions. The virtual array elements are combined to form virtual sparse arrays with different aperture sizes. Second, we optimize the parameters through the error backpropagation algorithm and obtain the optimal combination of virtual sparse arrays, which eliminates the phase ambiguity. Meanwhile, the virtual sparse arrays can maintain the advantages of sparse arrays, which can accomplish high-precision broadband direction finding. The experimental results demonstrate that the proposed method can solve the phase ambiguity problem effectively. It has high accuracy and strong robustness in simulation environment.

Optimal power distribution with mitigation of DC voltage fluctuations using artificial intelligent droop control scheme for MTDC grid

Optimal power share contribution with minimum DC voltage deviations is highly recommended in voltage source converter multi-terminal DC (VSC-MTDC). Hence, in this study, an artificial neural network based intelligent adaptive droop control proposes to enhance MTDC closed-loop system dynamics in terms of optimal power share and small redistribution of terminal DC voltages with negligible oscillations even in the presence of AC grid load demand change, admittance matrix uncertainty, outage of terminals and integration of wind farms. This is achieved via three-layer neural network and its weights are adaptively updated using error-back propagation algorithm. The asymptotic convergence of proposed control is derived using the Lyapunov’s stability theorem. Thus, the proposed intelligent adaptive droop control strategy not only guarantees to improve power share contribution with minimum voltage but also improved power quality and service continuity of MTDC grid. The robustness of proposed intelligent adaptive droop control is demonstrated on five terminal multi-terminal DC grid via MATLAB®.

Gas channel delineation utilizing a neural network and 3D seismic attributes: Simian Field, offshore Nile Delta, Egypt

Depositional architecture and hydrocarbon-bearing reservoir detection are considered the main challenges in the Nile Delta Basin, predominantly in the slope area. Channel fill deposits are very complicated because they are vertically stacked and pass through different depositional stages, forming complex depositional patterns. In the current study, an artificial neural network was used to detect gas and water-bearing reservoirs inside the complex depositional pattern and their spatial dimensions. Modern seismic reservoir characterization attribute analysis provides a good outcome in detecting reservoir fluids by mixing fluid-sensitive attributes with the well outcomes. Artificial Neural Network (ANN) algorithms for rock property prediction are considered to be a new emerging trend. The ANN algorithm has an advantage over other seismic reservoir delineation techniques because it has a good capability for building a nonlinear relationship between seismic data and petrophysical logs.Hence, it can be utilized to spatially anticipate reservoir fluids with a rational degree of accuracy. A back-error propagation algorithm, a supervised neural network method, was implemented on the Simian Field to anticipate the fluid-bearing reservoirs from the seismic-derived amplitudes. The ANN method showed unique results in the case study area by discriminating the gas-bearing intervals from the water-bearing intervals, although combining the assembly of the seismic attributes to supervise fluid classification. The method will help in the development plans of the field and track the volume of water and gas.

Grouting reinforcement strategy for tunnel sand layer based on BP neural network

Abstract Tunnel sand layer grouting reinforcement is a major difficulty in the current development of underground space. Finding a suitable method strategy for grouting reinforcement of the road sand layer to ensure the smooth implementation of the construction is imminent. In this paper, by building a BP neural network model, using signal forward propagation algorithm and error back propagation algorithm, back propagation of the error signal through the implied layer to the input layer, increased accuracy of calculations. To prove that BP neural network based on can effectively enhance the effect of tunnel grouting reinforcement, propose strategies for tunnel sand layer grouting reinforcement. Proven by simulation experiments: the effect of grouting reinforcement is influenced by the grouting material, grouting pressure, and the condition of the injected medium. The grouting parameters, grouting compressive strength and grouting age are the three major factors affecting the grouting reinforcement effect as deduced from the BP neural network input layer and implicit layer, a BP neural network model can be built to derive the parameters of these three major influencing factors. The calculation shows that, BP neural networks can provide specific data that can be relied upon for grout reinforcement, its effect prediction accuracy can reach 98%. It can be seen that BP neural network has practical application in tunnel sand layer grouting reinforcement strategy.

Artificial intelligence-based public safety data resource management in smart cities

Abstract With the development of urbanization, urban public safety is becoming more and more important. Urban public safety is not only the foundation of urban development, but also the basic guarantee for the stability of citizens’ lives. In the context of today’s artificial intelligence (AI), the concept of smart cities is constantly being practiced. Urban public safety has also ushered in some new problems and challenges. To this end, this article aimed to use AI technology to build an efficient public safety data resource management system in a smart city environment. A major goal of AI research was to enable machines to perform complex tasks that normally require human intelligence. In this article, a data resource management system was constructed according to the city security system and risk data sources, and the data processing method of neural network (NN) was adopted. Factors affecting urban public safety were processed as indicator data. In this article, the feedforward back-propagation neural network (BPNN) was used to predict the index data in real time, which has realized the management functions of risk monitoring and early warning of public safety data indicators. The BPNN model was used to test the urban risk early warning capability of the constructed system. BPNN is a multi-layer feed-forward NN trained according to the error back-propagation algorithm, which is one of the most widely used NN models. The results showed that the average prediction accuracy of the BPNN model for indicator prediction was about 89%, which was 16.1% higher than that of the traditional NN model. The average risk warning accuracy rate of the BPNN model was 90.3%, which was 16.5% higher than that of the traditional NN model. This shows that the BPNN model using AI technology in this article can more efficiently and accurately carry out early warning of risk and management of urban public safety.

Fault Diagnosis of Rotating Machinery Based on 1D–2D Joint Convolution Neural Network

Convolutional neural network has been widely used in fault diagnosis of mechanical devices. In particular, two-dimensional convolutional neural network requires manual selection of multi-scale transformation to transform vibration signal into two-dimensional structure. Although one-dimensional convolution neural network can directly use the vibration signal for convolution processing, it can't make full use of the nonlinear information in one-dimensional space. In order to make full use of the advantages of one-dimensional and two-dimensional convolutional neural networks, this paper develops a one-dimension in tandem with two-dimension joint convolutional neural network (1D-2D JCNN) for rotating machinery fault diagnosis. More specifically, one-dimensional convolution is employed to adaptively obtain the multi-scale feature vectors of the vibration signal, and these feature vectors are constructed into two-dimensional maps, and then these two-dimensional vectors are used as the input of the two-dimensional convolutions neural network. Take the cross-entropy loss function as the loss function, and use the error back propagation algorithm to optimize the filter parameters of the 1D-2D JCNN model to obtain the final fault diagnosis model. Using the motor bearing data set and the worm gearbox data set, the experimental results show the excellent classification performance of bearings and gears under different working conditions. The average diagnostic accuracy of 10 runs on CWRU bearing data set is 99.92%, and the variance is 3.96e-6. The average diagnostic accuracy of 10 runs on worm gear data set is 99.82%, and the variance is 7.82e-6. Compared with the traditional fault diagnosis model and the latest convolution neural network method, the 1D-2D JCNN shows better diagnosis performance.

Experimental and numerical studies on lithium-ion battery heat generation behaviors

Current predictions of battery HGR (heat generation rate) mainly rely on Bernardi’s empirical equations, which suffer from limitations of adaptability for thermal use. A novel scheme based on experiments and BPNN (BP (back propagation) neural network is a multilayer feedforward neural network trained according to the error back propagation algorithm) is proposed in this paper. In the experiments, the thermal generation behavior of 18650 batteries under 108 different operating conditions was investigated, and the Ta (ambient temperature) as well as ID (discharge current) were extracted as inputs to the BPNN model. The experimental results show that the heat generation performance of the battery is strongly influenced by Ta and ID. The higher Ta and ID, the more obvious this effect is. The maximum error of the battery HGR obtained by using Bernardi’s empirical equations can reach 45.2 %. The predicted battery HGR using BPNN has good accuracy with error control within 5%. Accordingly, this paper provides the technical basis for proposing a promising thermal management solution.

System for identifying the level of mobile gaming addiction in children and adolescents

This research paper describes a system for identifying the level of mobile gaming addiction in children and adolescents using data collected from mobile electronic devices. The system includes a mobile application that collects usage data on mobile games and uses a neural network model based on an "autoencoder" architecture to detect addiction levels. The neural network model includes the error back propagation algorithm and was trained on parameters selected from Chen Scale (CIAS) for addiction level detection. A prototype of the mobile application was created and adequate estimates of the neural network model quality were obtained. Future work will involve validation of the system with psychologists and further improvement of the neural network model by increasing the number of observations and adjusting the architecture. The results of this study demonstrate the potential of using neural networks and mobile technology to identify and monitor mobile gaming addiction in children and adolescents, which can aid in early intervention and prevention efforts.

A Dendritic Neuron Model Optimized by Meta-Heuristics with a Power-Law-Distributed Population Interaction Network for Financial Time-Series Forecasting

The famous McCulloch–Pitts neuron model has been criticized for being overly simplistic in the long term. At the same time, the dendritic neuron model (DNM) has been shown to be effective in prediction problems, and it accounts for the nonlinear information-processing capacity of synapses and dendrites. Furthermore, since the classical error back-propagation (BP) algorithm typically experiences problems caused by the overabundance of saddle points and local minima traps, an efficient learning approach for DNMs remains desirable but difficult to implement. In addition to BP, the mainstream DNM-optimization methods include meta-heuristic algorithms (MHAs). However, over the decades, MHAs have developed a large number of different algorithms. How to screen suitable MHAs for optimizing DNMs has become a hot and challenging area of research. In this study, we classify MHAs into different clusters with different population interaction networks (PINs). The performance of DNMs optimized by different clusters of MHAs is tested in the financial time-series-forecasting task. According to the experimental results, the DNM optimized by MHAs with power-law-distributed PINs outperforms the DNM trained based on the BP algorithm.