Three-layer Back-propagation Neural Network Research Articles

Data transmission optimization based on multi-objective deep reinforcement learning

Abstract Simultaneously reducing network energy consumption and delay is a hot topic today. This paper addresses this issue by designing a novel multi-objective data transmission optimization algorithm based on deep reinforcement learning. A three-layer back propagation (BP) neural network is designed to improve the accuracy of environmental state prediction, by learning from historical state and action sequence data, which can help the agent make better decision for routing selection in complex network environment. Based on this, we use Q-Learning to find routing for transmission demands, aggregating more traffic through less links and routers, to reduce energy consumption and delay. To enhance the efficiency and robustness of the algorithm, a new reward mechanism is designed based on the traffic demand and the link state. The algorithm divides candidate links into three levels for path selection so that a better solution can be obtained on the basis of ensuring feasible solutions are obtained. Continuous updating of the Pareto set through multiple state steps approximates the optimal solution. We leverage the Euclidean distance to the reference point to measure the optimization effect of the two objectives. The simulation results show that this algorithm outperforms existing algorithms in reducing energy consumption and network delay.

Robust ELM-PID tracing control on autonomous mobile robot via transformer-based pavement crack segmentation

Pavement crack tracing is paramount to missions encompassing automated crack sealing for road maintenance. However, existing methods still face several challenges, including the incapability to precisely extract crack trajectories and the challenge of tuning control parameters within intricate backgrounds. To address these limitations, the ViT-S2T network and the ELM-PID control system are proposed for crack tracing. Specifically, the ViT-S2T consists of two branches. The transformer-based feature extraction module (TFEM) integrates multi-head attention mechanism and multi-layer perceptron to capture global contextual crack semantic features. The incoherent segmentation masks (ISM) employs a binary classifier to predict the coarsest irrelevant mask and further performs up-sampling fusion of higher-resolution features. Moreover, the Neural-PID control method is designed to track crack trajectories, combining Extreme Learning Machines (ELM) and Proportional Integral Derivative (PID). The ELM-PID controller utilizes a three-layer backpropagation neural network and proposes the ELM model for adaptive adjustment by predicting the tuning parameters of PID. This framework is applied to real-time visual tracing for edge AI. Extensive tests performed on three arduous datasets of DeepCrack, CFD, and S2TCrack, achieving a precision of 82.76% and mAP@0.5 of 75.63% and speed of 0.0513 m/s, demonstrating the superior and robust nature of our approach in pavement crack tracing.

Machine Learning-Based Remote Sensing Inversion of Non-Photosynthetic/Photosynthetic Vegetation Coverage in Desertified Areas and Its Response to Drought Analysis

Determining the responses of non-photosynthetic vegetation (NPV) and photosynthetic vegetation (PV) communities to climate change is crucial in illustrating the sensitivity and sustainability of these ecosystems. In this study, we evaluated the accuracy of inverting NPV and PV using Landsat imagery with random forest (RF), backpropagation neural network (BPNN), and fully connected neural network (FCNN) models. Additionally, we inverted MODIS NPV and PV time-series data using spectral unmixing. Based on this, we analyzed the responses of NPV and PV to precipitation and drought across different ecological regions. The main conclusions are as follows: (1) In NPV remote sensing inversion, the softmax activation function demonstrates greater advantages over the ReLU activation function. Specifically, the use of the softmax function results in an approximate increase of 0.35 in the R2 value. (2) Compared with a five-layer FCNN with 128 neurons and a three-layer BPNN with 12 neurons, a random forest model with over 50 trees and 5 leaf nodes provides better inversion results for NPV and PV (R2_RF-NPV = 0.843, R2_RF-PV = 0.861). (3) Long-term drought or heavy rainfall events can affect the utilization of precipitation by NPV and PV. There is a high correlation between extreme precipitation events following prolonged drought and an increase in PV coverage. (4) Under long-term drought conditions, the vegetation in the study area responded to precipitation during the last winter and growing season. This study provides an illustration of the response of semi-arid ecosystems to drought and wetting events, thereby offering a data basis for the effect evaluation of afforestation projects.

Comparative study of integer-order and fractional-order artificial neural networks: Application for mathematical function generation

This paper investigates the impact of fractional derivatives on the activation functions of an artificial neural network (ANN). Based on the results and analysis, a three-layer backpropagation neural network model with fractional and integer derivatives employed in the activation function and fractional gradient descent method backpropagation learning algorithm has been proposed. Specifically, three perceptrons have been proposed based on fractional and integer derivatives applied to activation functions and learning algorithms. They are fractional derivative activation function (FDAF) perceptron, integer derivative activation function (IDAF)-fractional derivative learning algorithm (FDLA) perceptron, and fractional derivative learning algorithm (FDLA) perceptron. The Riemann-Liouville (RL) fractional derivative, Grunwald-Letnikov (GL) derivative, Caputo-Fabrizio (CF) fractional derivative, Caputo (C) fractional derivative, and Atangana-Baleanu (AB) fractional derivative have been employed. The impact of these derivatives’ fractional-order (FO) is investigated in a wide range from 0.1−0.9 on testing mean square error (MSE) and time required to train the perceptron. FO-based and IO-based perceptrons are compared with the help of performance metrics such as testing MSE and the time required to train the models. The training and testing simulation results illustrate that Caputo-Fabrizio derivative-based perceptrons outperform other fractional derivatives. Also, the performance of the FO-based perceptron is better in terms of the least MSE.

Research on quantitative inversion characterization of high-definition electrical imaging logging in oil-based mud based on backpropagation neural network and multiple population genetic algorithm-Levenberg-Marquardt algorithm

Electrical imaging logging in oil-based mud (OBM) has been developed for some time and is gradually playing an important role in the description of deep carbonate and shale reservoirs. Quantitative characterization of reservoir rock parameters such as resistivity is one of the most innovative developments in this field. The development of this technology needs to address and resolve four core issues: a wide range of parameter variations, removal of mud-cake influence in low-resistivity formations, dielectric rollover in high-resistivity formations, and multifrequency dielectric dispersion effects. To address the aforementioned issues, the joint use of a backpropagation neural network (BPNN) and the multiple population genetic algorithm (MPGA)-Levenberg-Marquardt (LM) algorithm for high-resolution quantitative imaging is developed. First, the numerical simulation is used to calculate the well-logging response data under the influence of multiple parameters, thereby establishing a forward response database. Then, within the forward response database, the instrument response function is fitted using BPNN, to compress the data volume. Next, based on the fitted response function, an inversion method for three parameters, including reservoir rock resistivity, permittivity, and plate standoff, is established using the LM algorithm optimized with MPGA. The results indicate that the use of a three-layer BPNN enables rapid and accurate calculation of the electrical imaging logging response in OBM. The calculation of a single point only requires 0.1 ms with an accuracy of more than 99%. The MPGA-LM algorithm exhibits stronger stability and improved inversion accuracy, with a single point inversion time of only 2 ms, and contributes to the high-definition quantitative description of electrical imaging logging in OBM, which is important in characterizing formation structures, distinguishing formation fractures, etc.

Optimizing asphalt foaming using neural network

AbstractThis study uses a three-layer backpropagation neural network combined with particle swarm optimization to control the foamed bitumen in cold recycling technology. The foaming process of bitumen is non-linear and depends on dynamic temperature. By developing a neural network model, this study effectively captures the complex relationships between temperature, water content, air pressure, and the expansion ratio and half-life of foamed bitumen. The integration of particle swarm optimization enhances the accuracy and convergence of the neural network model by optimizing the initial weights. This optimization process improves the model's ability to predict and control the quality of foamed bitumen accurately. It serves as a valuable tool for the rapid development of high-quality cold asphalt design.

Water-exit dynamics of a ventilated underwater vehicle in wave environments with a combination of computational fluid dynamics and machine learning

A ventilated vehicle exiting water in a wave environment is a complex nonlinear process, and the mechanism by which the wave conditions influence this process remains poorly understood. This paper describes realistic simulations of a ventilated vehicle exiting a water body under various wave conditions. Comprehensive analysis is conducted for a range of distinct wave scenarios, and a machine learning-based method is developed for the rapid forecasting of vehicle-related parameters. A three-layer backpropagation neural network is constructed, and its prediction performance is verified. Subsequently, predictive and optimization procedures are employed to determine the optimal wave phase for the water exit of the vehicle. Different wave conditions are shown to significantly affect the evolution of the ventilated cavity as well as the kinematic and loading characteristics of the vehicle. The pitch angular velocity and angle at the moment when the head of the vehicle reaches the free surface exhibit a positive cosine trend under different wave conditions. No regularity of the pitch angular velocity at the moment when the tail reaches the free surface is evident. The neural network exhibits exceptional proficiency in predicting the motion parameters and load characteristics of the vehicle. The optimal point for the vehicle to exit the water is determined to be at a wave phase of 0.125π, while the most hazardous point occurs when the wave phase is 1.1875π.

Enhancing plasticity in optoelectronic artificial synapses: A pathway to efficient neuromorphic computing

The continuous growth in artificial intelligence and high-performance computing has necessitated the development of efficient optoelectronic artificial synapses crucial for neuromorphic computing (NC). Ga2O3 is an emerging wide-bandgap semiconductor with high deep ultraviolet absorption, tunable persistent photoconductivity, and excellent stability toward electric fields, making it a promising component for optoelectronic artificial synapses. Currently reported Ga2O3 optoelectronic artificial synapses often suffer from complex fabrication processes and potential room for improvement due to plasticity. To address the issue of low device plasticity and practical application scenarios, we present an amorphous Ga2O3 (α-GaOx) flexible optoelectronic artificial synapse. This synapse modulates light stimulus signals using electron/oxygen vacancies and optical stimulation and operates as a visual storage device for information processing. We investigate the improvement of the optoelectronic synapses' plasticity by controlling the number of oxygen vacancies via a plasma treatment method and demonstrate its effective application in a three-layer backpropagation neural network for handwritten digit classification. Under the same stimulus conditions, the synaptic weight of samples treated with Ar plasma exhibits a higher rate of change, with the current levels increasing by 2–3 orders of magnitude, achieving greater plasticity. The improved optoelectronic synapses achieved an accuracy of 93.34%/94%, demonstrating their potential as efficient computing solutions and insights for future applications in NC chips.

Research and assessment of high sensitivity intensity modulation vector curvature sensor based on cascaded bamboo-joint microcavity structure

This study presents the VC-BM, a novel vector bamboo joint cascaded microcavity curvature sensor, which utilizes light intensity correlation. The device is fabricated using arc fusion, connecting a tapered seven-core optical fiber with a multi-mode and hollow-core fiber. Meanwhile, mode-coupling theory optimizes its design for high-quality transmission spectra. The feasibility and sensitivity of vector curvature sensing via intensity modulation (VC-BM) have been demonstrated through theoretical simulations and experimental validation. Building upon this foundation, a novel envelope intensity differential modulation algorithm is proposed, which maintains curvature direction recognition and enhances sensitivity, reaching up to 243.57 ± 3.63 dB/m−1. Additionally, leveraging its advantages, the successful integration of a three-layer Backpropagation (BP) neural network has enabled predictive analysis of this curvature sensing device, thereby increasing its commercial value. Furthermore, the sensor's high stability and resistance to temperature interference make it an ideal solution for wearable micro-deformation monitoring.

INTERNET TRAFFIC ZONE IDENTIFICATION BY BACKPROPAGATION AND PROBABILISTIC NEURAL NETWORKS

The article proposes an approach based on the concept of Artificial Intelligence for the categorization of urban areas of Internet content by corporate customers. The applicability of different neural apparatus was analyzed as well as three-layer Backpropagation Neural Networks (BPN) and four-layer Probabilistic Neural Networks (PNN) as the most suitable for the purpose of the study were selected. The synthesis of BPN architectures for Internet traffic identification was carried out according to a different number of computing units in the hidden layers with hyperbolic tangent sigmoid, log-sigmoid and linear transfer functions. The variations of a set of specific criteria were examined as Accuracy, Mean-Squared Error, Mean Absolute Error, Correlation coefficients, etc. The selection of PNNs against the defined quality indicators was based on a stepwise increase of the spread indicator of the Kernel functions in a Radial-Basis (RB) structural layer by analogy similar to that applied to BPNs. In the research processes, high levels of neural recognition indicators were established in processing with the Incoming flows of Internet Packages in an Accuracy of over 90.00%.

Roughness prediction of end milling surface for behavior mapping of digital twined machine tools

Background: The quality of machined parts is considered as a relevant factor to evaluate the production performance of machine tools. For mapping the production performance into a digital twin machine tool, a virtual metrology model for surface roughness prediction, which affects products' mechanical capacity and aesthetic performance, is proposed in this paper. Methods: The proposed model applies a three-layer backpropagation neural network by using real-time vibration, force, and current sensor data collected during the end milling machining process. A grid search plan is used to settle down the number of neurons in the middle layer of the backpropagation neural network. Results: The experimental results indicate that the model with multiple signals as input performs better than it with a single signal. In detail, when the model input is the combination of force, vibration, and current sensor data, the prediction accuracy reaches the optimum with the mean absolute percentage error of 1.01%. Conclusions: Compared with the state-of-the-art convolutional neural network method with automatic feature extraction ability and other commonly used traditional machine learning methods, the proposed data preprocessing procedure integrated with a three-layer backpropagation neural network has a minimum prediction error.

Simulation analysis of carbon peak path in China from a multi-scenario perspective: evidence from random forest and back propagation neural network models.

China faces tough challenges in the process of low-carbon transformation. To determine whether China can achieve its new 2030 carbon peaking and carbon intensity reduction commitments, accurate prediction of China's CO2 emissions is vital. In this paper, the random forest (RF) model was used to screen 26 carbon emission influencing factors, and seven indicators were selected as key variables for prediction. Subsequently, a three-layer back propagation (BP) neural network was constructed to forecast China's CO2 emissions and intensity from 2020 to 2040 under the 13th Five-Year Plan, 14th Five-Year Plan, energy optimization, technology breakthrough, and dual control scenarios. The results showed that energy structure factors have the most significant impact on China's CO2 emissions, followed by technology level, and economic development factors are no longer the main drivers. Under the 14th Five-Year Plan scenario, China can achieve its carbon peaking on time, reaching 10,434.082 Mt CO2 emissions in 2030. Although the new commitment to intensity reduction (over 65%) under this scenario cannot be achieved, the 14th Five-Year Plan can bring about 73.359 and 539.710 Mt of CO2 reduction in 2030 and 2040 respectively, compared to the 13th Five-Year Plan. Under the technology breakthrough and dual control scenarios, China will meet its new commitments ahead of schedule, with the dual control scenario being the optimal pathway for CO2 emissions to peak at 9860.08 Mt in 2025. It is necessary for Chinese policy makers to adjust their current strategic planning, such as accelerating the transformation of energy structure and increasing investment in R&D to achieve breakthroughs in green technologies.

State of Health Estimation of Lithium-Ion Batteries Using a Multi-Feature-Extraction Strategy and PSO-NARXNN

The lithium-ion battery state of health (SOH) estimation is critical for maintaining reliable and safe working conditions for electric vehicles (EVs). However, accurate and robust SOH estimation remains a significant challenge. This paper proposes a multi-feature extraction strategy and particle swarm optimization-nonlinear autoregressive with exogenous input neural network (PSO-NARXNN) for accurate and robust SOH estimation. First, eight health features (HFs) are extracted from partial voltage, capacity, differential temperature (DT), and incremental capacity (IC) curves. Then, qualitative and quantitative analyses are used to evaluate the selected HFs. Second, the PSO algorithm is adopted to optimize the hyperparameters of NARXNN, including input delays, feedback delays, and the number of hidden neurons. Third, to verify the effectiveness of the multi-feature extraction strategy, the SOH estimators based on a single feature and fusion feature are comprehensively compared. To verify the effectiveness of the proposed PSO-NARXNN, a simple three-layer backpropagation neural network (BPNN) and a conventional NARXNN are built for comparison based on the Oxford aging dataset. The experimental results demonstrate that the proposed method has higher accuracy and stronger robustness for SOH estimation, where the average mean absolute error (MAE) and root mean square error (RMSE) are 0.47% and 0.56%, respectively.

A prediction model of wall shear stress for ultra-high-pressure water-jet nozzle based on hybrid BP neural network

ABSTRACT Two hybrid back-propagation neural network (BPNN) models optimized by two heuristic search algorithms, namely genetic algorithm (GA-BP) and particle swarm optimization (PSO-BP), are proposed in this paper to predict radial maximum wall shear stress instead of traditional computational fluid dynamics (CFD) methods. The two proposed models are trained and validated using a database of 150 radial maximum wall-shear-stress values obtained via CFD simulations. The best fit model is identified from various BPNN models based on the Akaike information criterion (AIC) and the Bayesian information criterion (BIC), balancing the trade-off between goodness-of-fit and model complexity. The model performance is evaluated by MAE, RMSE, regression coefficient (R), and Nash-Sutcliffe efficiency (NSE). The best fit model is a three-layered BPNN model consisting of a 4:4:1 topology. In almost all evaluation indicators, the two hybrid BPNN methods outperform three existing algorithms, namely classical BPNN, random forest (RF), and gradient boosting decision tree (GBDT). Both PSO-BP and GA-BP can provide a more precise assessment of radial maximum wall shear stress, with maximum errors being 5.81% and 8.24% respectively. The proposed PSO-BP prediction model is promising and has great feasibility in predicting the radial maximum wall shear stress of UHP water-jet nozzle in engineering applications.

Probabilistic Evaluation of Tunnel Boring Machine Penetration Rate Based on Case Analysis

Although geological parameters are known to affect the penetration rate (PR) of a tunnel boring machine (TBM), their relation to the probability of TBM PR has been rarely considered. In this article, a probabilistic evaluation model of TBM PR was proposed. Firstly, the marginal distributions of five geological parameters were confirmed by mathematical statistics. Then Copula theory was used to construct a five-dimensional joint probability distribution of the geological parameters in line with the marginal distributions. Next, the collected geological parameters were utilized to train a three-layer backpropagation neural network (BPNN) model for predicting the TBM PR. Finally, A Copula—BPNN coupled model was built for estimating the probability of TBM PR, and a Weibull distribution function of the predicted TBM PR was obtained through Monte Carlo simulation. Considering the uncertainty, correlation, and multi-factor influence, this paper realized the probabilistic evaluation of TBM PR. Discussion on the parameter uncertainty and independence shows that the variability of the geological parameters is necessary in TBM PR prediction. Quantitative probability estimation of the TBM PR can help with optimizing the driving parameters under different geological conditions to improve construction efficiency.

Predictive Study of Flow-Accelerated Corrosion Characteristic Parameters Based on the Neural Network

Corrosion of equipment by corrosive media is widespread in the processing of inferior crude oil. In hydroprocessing reactor effluent systems, corrosive media are very destructive to heat exchangers and air coolers during flow and cooling because of the high-temperature and -pressure environment. A fire and explosion in the air cooler or heat exchanger are highly likely when their tubes leak. Currently, there are no effective direct detection and prediction means to evaluate the corrosion risk in real time, creating significant hidden threats to the safe operation of the equipment. Therefore, this paper proposes a condition expansion method based on a Gaussian distribution. The distribution laws of characteristic corrosion parameters under various working conditions were studied, and the corrosion risk of the equipment was evaluated. A three-layer back-propagation neural network model is constructed to predict the characteristic corrosion parameters. After testing, the model is shown to have superior predictive accuracy and generalization performance. It can also meet the demand for real-time equipment corrosion prediction. The proposed method can serve an essential role in guiding engineers to take correct and timely prevention and control measures for different degrees of corrosion to reduce losses.

Intelligent Monitoring System Based on Noise-Assisted Multivariate Empirical Mode Decomposition Feature Extraction and Neural Networks.

Because of the nonlinearity and nonstationarity in the vibration signals of some rotating machinery, the analysis of these signals using conventional time- or frequency-domain methods has some drawbacks, and the results can be misleading. In this paper, a couple of features derived from multivariate empirical mode decomposition (MEMD) are introduced, which overcomes the shortcomings of the traditional features. A wind turbine gearbox and its bearings are investigated as rotating machinery. In this method, two types of feature structures are extracted from the decomposed signals resulting from the MEMD algorithm, called intrinsic mode function (IMF). The first type of feature vector element is the energy moment of effective IMFs. The other type of vector elements is amplitudes of a signal spectrum at the characteristic frequencies. A correlation factor is used to detect effective IMFs and eliminate the redundant IMFs. Since the basic MEMD algorithm is sensitive to noise, a noise-assisted extension of MEMD, NA-MEMD, is exploited to reduce the effect of noise on the output results. The capability of the proposed feature vector in health condition monitoring of the system is evaluated and compared with traditional features by using a discrimination factor. The proposed feature vector is utilized in the input layer of the classical three-layer backpropagation neural network. The results confirm that these features are appropriate for intelligent fault detection of complex rotating machinery and can diagnose the occurrence of early faults.

Study on control-oriented emission predictions of PPCI diesel engine with two-stage fuel injection

Model predictive control (MPC) is an effective technology used to improve the control accuracy of low temperature combustion engines. The main purpose of this study is to establish a control-oriented emission prediction model based on fuel injection parameters, which provides the basis for the development of MPC controller for PPCI engines. In this study, the combustion model and backpropagation (BP) neural network emission prediction model of a partially premixed compression ignition (PPCI) diesel engine were established, through which the real-time combustion and emissions can be realized by inputting two-stage fuel injection parameters. The Wiebe function linearization method was proposed for the combustion model to achieve high-precision reconstruction of combustion at different fuel injection parameters. The nitrogen oxide (NOx) and particulate matter (PM) emissions were predicted using the combustion characteristics calculated by the combustion model. The results show that the established combustion model can accurately reconstruct the heat release rate and cylinder pressure of PPCI combustion, and the maximum error between the calculated and test values of the combustion parameters is less than 9%. The three-layer BP neural network can achieve high-precision predictions of NOx and PM emissions, with a prediction accuracy greater than 95%. Under the 300th continuous transient seconds, the cumulative prediction error of NOx was 2%, and the cumulative prediction error of PM was 4.9%, indicating high accuracy. This provides the basis for developing MPC controllers for PPCI engines.

Individual Travel Knowledge Graph-Based Public Transport Commuter Identification: A Mixed Data Learning Approach

Commuters are the stable travel group for the public transportation (PT) service system. Accurately identifying the PT commuters is conducive to promoting PT service quality and development of urban sustainable transportation. This paper extracts individual PT travel chain information and constructs individual travel knowledge graphs of PT passengers based on the association matching algorithm and the theory of multilayer planning. A mixed dataset is formed by associating individual travel chains with travel survey data. Seven travel characteristic indicators regarding travel performance and spatiotemporal travel characteristics are extracted. The identification model of PT commuters is developed based on a three-layer backpropagation neural network (BPNN). The optimal model structure of neuron node number, transfer function, and learning rate are discussed quantitatively according to the minimization of model errors. The evaluation indexes of overall accuracy and kappa coefficient of the constructed model are 94.5% and 87.9% separately. The results indicate that the model identification accuracy is acceptable, and the proposed characteristic indicators and systematic modelling procedure are effective. Then, the model performance is compared with the other five machine learning models further. The results confirm that the proposed model has a better identification accuracy and viability, and the model performance will improve with the increase of the sample size.

Pattern Recognition of Growth Characteristics Based on UHF PD Signals of Electrical Tree

In insulation system of power equipment, electrical tree grown under the effect of partial discharges (PD) is one of the main degradation processes leading to failure of high voltage polymeric insulation. PD signals are measured and analyzed for condition monitoring of electrical insulation. In this paper, the combine detection system of Ultra-High Frequency (UHF) PD and electrical tree was used to test PE samples. Based on the energy amplitude variation, the average energy and the average energy distribution of PD signals to identify the growth process of electrical tree, twenty-eight PD characteristic parameters were used as input parameters, such as the energy percentage of 16-layer wavelet decomposition, the wavelet coefficient parameter and the time-frequency domain parameter, which entered the three-layer Back Propagation Neural Network (BPNN) of forty-five hidden neurons, and then the accuracy of pattern recognition during the growth process of electrical tree can reach 99.93% after 5000 steps of training cycle. The consistency of calculation results and experimental results presents that the calculation method can reveal the relationship between PD signals and the growth of electrical tree, which lays the foundation for the process analysis of electrical tree and insulation state monitoring.