Back-propagation Model Research Articles

Splitting behavior of polyvinyl alcohol fiber reinforced iron ore tailings concrete

To promote solid waste reutilization and reduce natural river sand consumption, iron ore tailings (IOT) are expected to be used in concrete production as a potential substitute for river sand under the premise of comparable material properties with ordinary concrete. In this study an IOT concrete reinforced by Polyvinyl Alcohol (PVA) fiber has been proposed and the slump and splitting behavior under various IOT substitution rates and PVA fiber contents has been experimentally investigated. Slump tests reveal that the optimal PVA fiber content is 0.1 % and IOT substitution rate is 60 % from the view of workability. Splitting tensile tests show a monotonous strength growth with PVA fiber content increase. The incorporation of IOT has increased concrete splitting tensile strength but the strength is smaller than that of normal concrete as soon as IOT substitution rates exceed 50 %. A regression model and Back Propagation (BP) neural network model have been adopted to establish splitting tensile strength prediction model. Compared to regression model, the BP model has exhibited better accuracy which is suitable to predict the splitting tensile strength of PVA-IOT concrete with PVA content less than 0.3 %.

Parameters prediction for Low-Plasticity ultrasonic rolling strengthening process of blades based on Few-Shot Genetic Bayesian-Back Propagation intelligent learning

This study aims to reduce titanium alloy blade surface roughness and enhance compressive residual stress (CRS) using the low-plasticity ultrasonic rolling strengthening process (URSP). Due to the URSP’s multifactorial complexity and characteristics of few and irregular test samples, this study proposes a Genetic Bayesian-Back Propagation neural network (GB-BP) for a few sample parameters optimization for low-plasticity URSP of the blade, which combines a back propagation (BP) neural network with genetic algorithm (GA) and Bayesian Optimization (BO). The proposed approach establishes a robust correlation between processing parameters, blade surface roughness, and CRS. Firstly, an orthogonal test with three factors, rolling depth, feeding speed, and rolling distance, is designed. The signal-to-noise ratio (S/N) analysis and mean value analysis identify rolling distance and depth as the primary factors influencing surface roughness and CRS. Furthermore, BP complement with BO is employed to predict blade surface roughness and CRS after URSP. Finally, GB-BP is proposed to predict parameters, and experimental validation demonstrates the superior predictive accuracy of the GB-BP network. Compared to traditional BP models, the mean error percentages for GB-BP prediction dropped by 51.286% for surface roughness and 69.818% for CRS. The root mean square error (RMSE) decreased by 51.864% for surface roughness and 71.982% for CRS. This network model can provide accurate parameters optimization for low-plasticity URSP of blades.

Prediction of rolling bearing performance degradation based on whale optimization algorithm and backpropagation model

Prediction of rolling bearing performance degradation based on whale optimization algorithm and backpropagation model

Predicting the low‐cycle fatigue life of Ti‐6Al‐4V alloy using backpropagation neural network optimized by the improved dung beetle algorithm

AbstractIn this study, we propose an innovative approach that enhances the performance of the backpropagation (BP) neural network in predicting the low‐cycle fatigue life of Ti‐6Al‐4V alloy by improving the dung beetle optimization (DBO) algorithm with the maximin Latin hypercube design (MLHD) strategy. To address the challenges posed by complex geometric components under different temperature conditions, this research employs finite element simulation to expand the limited experimental dataset and utilizes these data to further guide and optimize the MLHD_DBO_BP model. Test results indicate that the proposed MLHD_DBO_BP model significantly outperforms the traditional finite element method (FEM) and other neural network models in terms of fatigue life prediction performance. This research demonstrates the effectiveness of machine learning models that combine experimental and simulation data in predicting the low‐cycle fatigue life of Ti‐6Al‐4V alloy.

Segmented parabolic adjustment of the FAST reflector utilizing spatial coordinate rotation transformation

Abstract Since the middle of the twentieth century, the advent of radio telescopes has brought a whole new way and approach to astronomical observation. For Arecibo-type radio telescopes, the tuning optimization of the active reflecting surface (working paraboloid) is the main factor affecting the reflectance calibration. In this study, leveraging the transformation of spatial coordinates through rotation, we introduce an innovative optimization model specifically for the segmented paraboloid of the Five-hundred-meter Aperture Spherical radio Telescope (Hereinafter referred to as FAST) designed by China astronomer and scientist Nan Rendong. This research constructs the equation for an ideal paraboloid and adjusts the working paraboloid to fit within specified constraints such as the orientation of the target star, the adjustment limit of the actuator, and the spatial coordinates. The study employs a combination of coarse and fine grid searches to identify and record the optimal adjustment scheme of the main cable nodes at different angles and the corresponding 2226 actuator coordinates and telescoping length, based on which we build a back propagation model to continuously modify the adjustment scheme. A combination of geometric simulation and Monte Carlo tests were also used for verification. Furthermore, we delve into the impact of variations between adjacent nodes of the modulating actuators, as well as potential longitudinal and radial changes. Compared to the conventional conditioning model, the segmented solution idealized paraboloid we created increases the original reflection efficiency from 77.92% to 95.56% in the working area of 300 m aperture, it will contributes to enhancing the overall performance of FAST.

Utilizing Enterprise Economic Benefit Evaluation Methods in Edge Intelligent Neural Network Applications

The core of enterprise economic benefit evaluation lies in the development of a quantitative identification model. The Back Propagation (BP) neural network possesses robust parallel computing, adaptive learning, and error correction capabilities, which can effectively reveal the economic benefits of enterprises and their relationship with influencing factors. This study establishes an economic benefit evaluation model for express delivery enterprises based on the BP neural network. The model takes the annual profit rate of enterprises as the quantitative index of economic benefits and selects 13 factors, both external and internal, influencing the annual profit rate of express delivery enterprises as inputs for the BP neural network model. The economic benefit evaluation model based on BP neural network meets the requirement of objective mean square error in the 300th training cycle. The research results demonstrate that the BP model significantly saves computing time and enables rapid, comprehensive, and objective evaluation of the economic benefits of industrial enterprises.

The changes prediction on terrestrial water storage in typical regions of China based on neural networks and satellite gravity data

Accurate prediction of regional terrestrial water storage change (TWSA) is of great significance for water resources planning and management, and early warning of extreme climate disasters. Aiming at the problem that the conventional methods on prediction of TWSA time series are difficult to be accurate, the six typical regions are selected in China as examples, including the upper reaches of the Yangtze River (UYR), the southwest region (SWR), the Liaohe River Basin (LRB), the North China Plain (NCP), the Qinghai-Tibet Plateau (QTP), and the Pearl River Basin (PRB). The mascon product from GRACE/GRACE-FO provided by CSR is used to extract TWSA time series in six typical areas. The improved Back Propagation (BP) neural network, Long Short-Term Memory (LSTM) neural network and the latest Bidirectional LSTM (BiLSTM-attention) neural network model based on attention mechanism are proposed to predict and analyze the regional TWSA. In the experiment, the selection of the optimal model parameters such as the number of hidden layer nodes and the number of hidden units of the neural network model is tested and analyzed in detail. Meanwhile, the model prediction results are compared with the traditional least squares method and random forest (RF) prediction method. The root mean square error (RMSE), determination coefficient (R2), Nash–Sutcliffe efficiency coefficient (NSE) and mean absolute percentage error (MAPE) were used to evaluate the accuracy of the predicted results. The results show that the improved BP, LSTM and Bi-LSTM-attention neural network models all achieve higher prediction accuracy in UYR and SWR areas. RMSE is less than 2.641 cm, R2 is as high as 0.8 or more, NSE is above 0.6, and MAPE is within 0.1. Compared with the least square method, the RMSE of the predicted results from three neural network decreased by 0.998 cm, 0.700 cm and 0.7563 on average, and the R2 increased by 81.75%, 69.89% and 72% on average. Compared with RFML method, the RMSE from three neural network is reduced by 0.601 cm, 0.316 cm and 0.360, and R2 is increased by 38.20%, 24.60% and 27.06% on average. NSE and RMSE are improved to varying degrees in the above regions. It shows that the improved BP, LSTM and BiLSTM-attention model used can effectively predict TWSA. The research methods and results in this paper can provide important reference for the rational utilization of regional water resources and disaster risk assessment.

Efficient multi-objective rolling strategy of photovoltaic/hydrogen system via short-term photovoltaic power forecasting

Green hydrogen serves as an effective energy carrier for achieving sustainable development goals. However, the high production cost and intermittent renewable energy output cause obstacles to hydrogen production. Various electrolyzer control methods have been proposed to promote electrolyzers' efficiency. Traditional electrolyzer control methods may lead to significant differences in electrolyzer wear among large-scale hydrogen production systems. Hence, this paper proposes a multi-objective rolling strategy based on photovoltaic power forecasting to mitigate electrolyzer switch fluctuations, enhance hydrogen production efficiency, and better manage the electrolyzers. Initially, the trend of photovoltaic output is calculated based on the predicted values of the photovoltaic output from the particle swarm algorithm (PSO)-backpropagation (BP) model, and during the power ramp-up phase, some electrolyzers are activated in advance to expedite the cold start process and enhance the hydrogen production. Simultaneously, the electrolyzers' power distribution and operational status are adjusted in conjunction with the real-time power situation. Finally, a multi-objective rolling strategy is proposed to balance the operational indicators of the electrolyzers. The results indicate that the proposed method yields 4,952,642 Nm3 hydrogens over six months, which has increased by 5.37% and 9.00% compared to the simple start-stop and rotation strategy, respectively. Moreover, switch action times are reduced by 27.68% and 43.22%, respectively. These results demonstrate the effectiveness of the proposed method in significantly enhancing hydrogen production and improving the economic feasibility of photovoltaic/hydrogen systems.

Monitoring soil nutrients using machine learning based on UAV hyperspectral remote sensing

ABSTRACT Unmanned aerial vehicles (UAV) are rapidly evolving experimental platforms that play an important role in remote sensing. In this study, we investigated a machine learning method for monitoring soil nutrient content using UAV hyperspectral remote sensing. We employed machine learning techniques for feature extraction and soil hyperspectral information modelling. In contrast to traditional mathematical transformation methods, we adopted a combination of random forest and differential evolution algorithms to rank the weights of individual hyperspectral data, thereby obtaining a series of spectral feature subsets for soil organic matter, total nitrogen and available phosphorus and potassium. Furthermore, the analytic hierarchy process was used for weight analysis, and the characteristic bands of the four soil nutrients were successfully extracted. Next, a quantitative inversion model based on a back-propagation (BP) neural network was established to estimate soil nutrient content, with determination coefficients higher than 0.7 and 0.6 for the modelling and verification sets, respectively. The relative percent difference values were greater than 2, among which the highest was for available potassium, with determination coefficients of 0.95 and 0.84 for the modelling and verification sets, respectively. In addition, visualization distribution maps of soil nutrients were obtained by combining the BP model and original reflectance hyperspectral images, and the comparisons of content histograms showed a relatively consistent distribution between the sampling and inversion points. The results verified the effectiveness of the combined machine learning method for large-scale and high-precision monitoring and visualization of soil nutrient contents.

A Creep Model of Steel Slag–Asphalt Mixture Based on Neural Networks

To characterize the complex creep behavior of steel slag–asphalt mixture influenced by both stress and temperature, predictive models employing Back Propagation (BP) and Long Short-Term Memory (LSTM) neural networks are described and compared in this paper. Multiple stress repeated creep recovery tests on AC-13 grade steel slag–asphalt mix samples were conducted at different temperatures. The experimental results were processed into a group of independent creep recovery test results, then divided into training and testing datasets. The K-fold cross-validation was applied to the training datasets to fine-tune the hyperparameters of the neural networks effectively. Compared with the experimental curves, both the effects of BP and LSTM models were investigated, and the broad applicability of the models was proven. The performance of the trained LSTM model was observed by a 95% confidence interval around the fit errors, thereby the creep strain intervals for the testing dataset were obtained. The results suggest that the LSTM model had enhanced prediction compared the BP model for creep deformation trends of steel slag–asphalt mixture at various temperatures. Due to the potent generalization strength of artificial intelligence technology, the LSTM model can be further expanded for forecasting road rutting deformations.

A Multi-Objective Genetic Algorithm-Based Predictive Model and Parameter Optimization for Forming Quality of SLM Aluminum Anodes

Aluminum–air batteries are characterized as “green energy for the 21st century” due to their clear advantages in terms of high current discharge, high specific energy, low cost, and easy-to-obtain electrode materials. This study develops the SLM aluminum anode quality prediction model and evaluates its learning and training results using the BP neural network architecture. By altering the network topology of the SLM aluminum anode quality prediction model, we create a process parameter backpropagation model that takes advantage of the extremely adaptable capabilities of artificial neural networks. The quick and exact selection of process parameters meets the goals of density, self-corrosion current, and anode usage, hence improving the forming quality and processing efficiency of SLM aluminum anodes. The experimental results show that the process parameter backpropagation model’s parameter configurations match to the real densities and self-corrosion currents, which are somewhat higher than the specified target values. The maximum error rate for the aluminum anode forming quality prediction model is 8.23%. Furthermore, the actual anode utilization rate is somewhat lower than the projected target value, indicating that the backpropagation model can satisfy actual production needs.

Modeling and forecasting of coal price based on influencing factors and time series

The fluctuations of coal prices substantially impact carbon pricing, the flexibility in power generation at coal-fired plants, and the pricing strategies of upstream and downstream industries. Given limitations in the existing coal price forecasting research, this study introduces grey relational analysis to determine the relational grade between influencing factors and coal prices. Results indicate that crude oil production, CPI (Consumer Price Index), coal sales of key coal mines, and PMI (Purchasing Managers Index) exhibit the highest correlation with coal prices, with grey relational coefficients all exceeding 0.9. This study also applies principal component analysis to achieve data dimensionality reduction, resulting in a 2% decrease in average forecasting error. The influence-based forecasting model developed in this study demonstrates an average forecasting accuracy exceeding 90%. To render coal price forecast more practical, this study constructs several models based on time series to forecast the coal price for the subsequent 30 days. The results show that the BP (Back Propagation) neural network model demonstrates higher forecasting accuracy than the LSTM (Long Short-Term Memory) model and ARIMA (Autoregressive Integrated Moving Average) model for 30-day coal price forecasting. The forecast generated by BP model closely match the actual coal price trends, achieving a high average prediction accuracy exceeding 85%. The accurate forecasting model of coal price proposed in this study holds significant importance in characterizing carbon price fluctuations, enhancing power plant flexibility in electricity generation, and guiding coal procurement.

Injection rate measurements and Machine-Learning based predictions of ECN Spray A-3 piezoelectric injector

This study investigates the impact of parametric conditions on the rate of injection (ROI) and injection quantities of a new piezo-type Engine Combustion Network (ECN) Spray A-3 injector using a combination of experimental and model-based approaches. The experiments were conducted using a Zeuch-based HDA Moehwald injection rate measurement system to obtain ROI data. An artificial neural network (ANN) algorithm was employed to develop a predictive model for ROI data, accurately capturing injection behaviors and ROI patterns under a wide range of conditions. The temporal ROI data were trained using a Bayesian regularization backpropagation model with four input conditions: chamber temperature, chamber pressure, injection pressure, and injection duration. This study examined the influence of these conditions on the transient profiles of the ROI, quasi-steady ROI, injection duration, and total injection mass for both the experimental and predicted results. The model demonstrated good predictions that aligned well with the experimental measurements, indicating a significant increase in the ROI and injection quantities with an increase in injection pressure, while the effects of chamber temperature and pressure were less pronounced. The ANN model used in this study allows for the prediction of various ROI profiles without the need for additional experiments or computational fluid dynamics techniques, and complex physical modeling, thereby substantially reducing the cost and time involved in developing injection control systems for advanced internal combustion engines.

Predicting thermodynamic adhesion energies of membrane fouling in planktonic anammox MBR via backpropagation neural network model

Predicting thermodynamic adhesion energies was a critical strategy for mitigating membrane fouling. This study utilized a backpropagation (BP) neural network model to predict the thermodynamic adhesion energies associated with membrane fouling in a planktonic anammox MBR. Acid-base (ΔGAB), electrostatic double layer (ΔGEL), and Lifshitz–van der Waals (ΔGLW) energies were selected as output variables, the training dataset was collected by the advanced Derjaguin-Landau-Verwey-Overbeek (XDLVO) method. Optimization results identified “7-10-3″ as the optimal network structure for the BP model. The prediction results demonstrated a high degree of fit between the predicted and experimental values of thermodynamic adhesion energy (R2 ≥ 0.9278), indicating a robust predictive capability of the model in this study. Overall, the study presented a practical BP neural network model for predicting thermodynamic adhesion energies, significantly enhancing the prediction tool for adhesive fouling behavior in anammox MBRs.

Back propagation model for prediction of deposition parameters in plasma sprayed WC-based coatings

Back propagation model for prediction of deposition parameters in plasma sprayed WC-based coatings

Constitutive Model and Microstructure Evolution of Ti65 Titanium Alloy.

The hot deformation behavior and mechanism of Ti65 alloy with a bimodal microstructure were investigated by isothermal compression experiments conducted on the Thermecmastor-Z simulator equipment at temperatures ranging from 950 to 1110 °C and strain rates ranging from 0.01 to 10.0 s-1. The Arrhenius constitutive model, based on strain compensation, and Grey Wolf optimization-neural network with back propagation model (GWO-BP), were both established. The differences between the experimental and predicted value of flow stress were compared and analyzed using the two models. The results show that the prediction accuracy of GWO-BP in the two-phase region is higher than that of Arrhenius model. In the single-phase region, both methods demonstrated high prediction accuracy. Compared to the single-phase region, the flow stress of Ti65 alloy shows a higher degree of softening in the two-phase region. During deformation in the two-phase region, the initial lamellar α phase transformed from a kinked and elongated morphology to a globularized topography as the strain rate decreased. Boundary-splitting was the primary mechanism leading to the spheroidization process. The degree of recrystallization increased with the increase in strain rate during the deformation in the single-phase region, while dynamic recovery and strain-induced grain boundary migration were the main deformation mechanisms at a lower strain rate. Discontinuous dynamic recrystallization may be the dominant recrystallization mechanism under a high strain rate of 10 s-1.

Comparative study of linear and nonlinear calibration algorithm for extrapolation ability of near infrared spectroscopy quantitative analysis

The determination of the o-nitrotoluene (o-MNT) content in separation process of mononitrotoluene (MNT) is of interest, since it affects the purity of m-nitrotoluene (m-MNT) and p-nitrotoluene (p-MNT). In real-world applications, the calibration model inevitably requires dealing with complex extrapolation problems. Therefore, this study extracted the spectral features of the o-nitrotoluene based on the interval selection algorithm. The linear calibration method (partial least squares (PLS)) and nonlinear calibration methods (support vector machine (SVM), back propagation (BP), random forest (RF), extreme learning machine (ELM)) were used to build the calibration models based on o-nitrotoluene samples in different concentration ranges, and the prediction accuracy and robustness of the calibration model were compared. The results indicate that the effectiveness of different calibration methods is different when going from prediction accuracy to robustness. The prediction accuracy and robustness of RF models are not satisfactory. BP models, which are capable of producing very accurate results in terms of prediction accuracy, are not able to solve extrapolation problems. PLS model has more advantages in model prediction accuracy. ELM has shown the best behavior in terms of robustness of model, but is inferior to PLS in terms of prediction accuracy.

A novel intelligent risk prediction model for the effectiveness of CO2/N2–ECGD technology

AbstractTo solve the problems of high sampling requirements and low predictive accuracy resulting from the complexity, uniqueness, and randomness of predicting the risk of the CO2 and N2 injection to enhance coal seam gas drainage (CO2/N2–ECGD) technology. The principal component analysis (PCA) method to reduce the dimensionality of the factor data that contribute to the effect risk of the technology was adopted. And the particle swarm optimization (PSO) method was implemented to search for optimal hyperparameters in support vector machine (SVM) by particle search, as a solution to the traditional SVM hyperparameters optimization problem. A novel risk prediction model using machine learning algorithms for gas injection displacement technology was constructed. The prediction results were tested and compared with those of backpropagation (BP), Random Forest (RF), and Decision Tree (DT) models using data from 29 gas injection displacement field projects in China. The results demonstrated that the SVM model had greater accuracy in prediction than the other three models. Additionally, after PSO optimization and dimensionality reduction, the PCA–PSO–SVM model reached 100% prediction accuracy, while requiring less modeling and operation time. The study provided a reliable and reasonable model for predicting technical effects, along with a theoretical basis for risk management and prevention. First, the technology's influencing indicators were analyzed by examining its mechanisms. Second, we utilized the PCA method to reduce the dimensionality of the factor data that contribute to the risk of the technology's effects. Third, we implemented the PSO method to search for optimal hyperparameters in the SVM through particle search, as a solution to the traditional SVM hyperparameters optimization problem. Finally, the prediction results were tested and compared with those of BP, RF, and DT models using data from 29 gas injection displacement field projects in China. The SVM model was found to have greater accuracy in prediction than the other three models. After PSO optimization and dimensionality reduction, the PCA–PSO–SVM model achieved 100% prediction accuracy while requiring less modeling and operation time. The study presents a valid and reasonable model for predicting technical effects and a theoretical basis for risk management and prevention.

Data-driven hydraulic property analysis and prediction of two-dimensional random fracture networks

The large parameter space and uncertainty of the discrete fracture networks (DFNs) make the hydraulic characteristics have multiple possibilities. Statistical analysis of hydraulic characteristics based on a large amount of data is a standard practice and central to DFN methodology. The present study developed a two-dimensional (2D) DFN numerical seepage model based on Matlab and Comsol from modeling and parameter calculation to simulation to investigate the influence of parameters involved fracture network on seepage characteristics of DFNs. In total, 1728 2D DFN models were generated with increasing fracture density, fracture length, power-law exponent, fracture orientations, and fracture included angles, and 1728 × 3 = 5184 numerical simulations of fluid flow along the x-axis, y-axis, and center through the DFN models are conducted using a self-developed numerical code. The influence of fracture geometric features on connectivity and permeability are estimated, respectively. And to realize the prediction of permeability tensor by known geological data, a theoretical model based on the Snow model and a data mining model based on the back-propagation model are discussed. The analysis helps researchers and engineers can make informed decisions and develop effective strategies for addressing the challenges posed by uncertainty hydraulic property of the fractured rock masses.

Application of Machine Learning for Productivity Prediction in Tight Gas Reservoirs

Accurate well productivity prediction plays a significant role in formulating reservoir development plans. However, traditional well productivity prediction methods lack accuracy in tight gas reservoirs; therefore, this paper quantitatively evaluates the correlations between absolute open flow and the critical parameters for Linxing tight gas reservoirs through statistical analysis. Dominant control factors are obtained by considering reservoir engineering theories, and a novel machine learning-based well productivity prediction method is proposed for tight gas reservoirs. The adaptability of the productivity prediction model is assessed through machine learning and field data analysis. Combined with the typical decline curve analysis, the estimated ultimate recovery (EUR) of a single well in the tight gas reservoir is forecasted in an appropriate range. The results of the study include 10 parameters (such as gas saturation) identified as the dominant controlling factors for well productivity and geological factors that impact the productivity in this area compared to fracturing parameters. According to the prediction results of the three models, the R2 of Support Vector Regression (SVR), Back Propagation (BP), and Random Forest (RF) models are 0.72, 0.87, and 0.91, respectively. The results indicate that RF has a more accurate prediction. In addition, the RF model is more suitable for medium and high-production wells based on the actual field data. Based on this model, it is verified that the productivity of low-producing wells is affected by water production. This study confirms the model’s reliability and application value by predicting recoverable reserves for a single well.