Back-propagation Artificial Neural Network Model Research Articles

Static Globularization Behavior and Artificial Neural Network Modeling during Post-Annealing of Wedge-Shaped Hot-Rolled Ti-55511 Alloy.

The globularization of the lamellar α phase by thermomechanical processing and subsequent annealing contributes to achieving the well-balanced strength and plasticity of titanium alloys. A high-throughput experimental method, wedge-shaped hot-rolling, was designed to obtain samples with gradient true strain distribution of 0~1.10. The samples with gradient strain distribution were annealed to obtain the gradient distribution of globularized α phase, which could rapidly assess the globularization fraction of α phase under different conditions. The static globularization behavior under various parameters was systematically studied. The applied prestrain provided the necessary driving force for static globularization during annealing. The substructure evolution and the boundary splitting occurred mainly at the early stage of annealing. The termination migration and the Ostwald ripening were dominant in the prolonged annealing. A backpropagation artificial neural network (BP-ANN) model for static globularization was developed, which coupled the factors of prestrain, annealing temperature, and annealing time. The average absolute relative errors (AARE) for the training and validation set are 3.17% and 3.22%, respectively. Further sensitivity analysis of the factors shows that the order of relative importance for static globularization is annealing temperature, prestrain and annealing time. The developed BP-ANN can precisely predict the static globularization kinetic curves without overfitting.

Drill bit wear monitoring and failure prediction for mining automation

This article introduces a novel approach for tricone bit wear condition monitoring and failure prediction for surface mining applications. A successful bit health monitoring system is essential to achieve fully autonomous blasthole drilling. In this research in-situ vibration signals were analyzed in time-frequency domain and signals trend during tricone bit life span were investigated and introduced to support the development of artificial intelligence (AI) models. In addition to the signal statistical features, wavelet packet energy distribution proved to be a powerful indicator for bit wear assessment. Backpropagation artificial neural network (ANN) models were designed, trained and evaluated for bit state classification. Finally, an ANN architecture and feature vector were introduced to classify the bit condition and predict the bit failure.

An Improved Tropospheric Tomographic Model Based on Artificial Neural Network

Global navigation satellite systems (GNSS) tropospheric tomography can be used to build a three-dimensional water vapor field. In traditional tomography, the signals crossing from the four sides of the tomographic region are not utilized. To make the best use of these valuable side-crossing signals, an improved tomographic model based on back propagation artificial neural network (BP-ANN) is proposed. In the new tomographic model, the inside part of the slant wet delay (SWD) of the side-crossing signal is divided into two sections: the isotropic and anisotropic components. The former is estimated by the zenith wet delay multiplied by the mapping function multiplied by an isotropic scale factor using a BP-ANN model, and the latter is estimated by horizontal gradients of the SWD multiplied by an anisotropic scale factor using an empirical model. The new tomographic model is experimentally evaluated using the HK CORS network measurements for the period of 21 days from 1 to 21 August 2019. Statistical results show that the root mean square error (RMSE) of slant water vapor reconstructed from the improved model is reduced to 1.35 from 2.85 mm of the traditional model. Compared with the traditional/height factor models, the percentages of the reduction in the RMSE of the tomographic result derived from the new model are 16%/9% and 22%/16%, respectively, using radiosonde and ERA5 data as references. These results suggest a good performance of the new model for GNSS tropospheric tomography.

Prediction of secant shear modulus and damping ratio for an extremely dilative silica sand based on machine learning techniques

A better understanding of soil response to dynamic loads, including earthquakes, can result in safer designs that can reduce casualties. The damping ratio and shear modulus are critical parameters in soil dynamics, and several factors affect these parameters, including density and moisture content. The non-linear and multiple influences on these two parameters make their estimation difficult. Machine learning techniques are very powerful mapping tools with a remarkable capacity to perform nonlinear multivariate function approximations. In this study, to predict sand secant shear modulus and damping ratio from input variables, artificial neural networks (ANN) and classification and regression random forests (CRRF) were used as alternative estimators. The database was created using a series of simple shear tests that accurately assessed damping ratios and secant shear modulus to predict these two dynamic parameters. The input variables of the proposed predictive models included vertical stress, relative density and cyclic stress ratio, and its outputs included secant shear modulus and damping ratio. The Bayesian Regularization (BR) back-propagation ANN model produced correlation coefficient (R) and mean absolute error (MAE) values of 0.998 and 0.006, respectively, while CRRF models gave R and MAE values of 0.995 and 66.051, respectively. Additionally, sensitivity analysis of artificial intelligence (AI) models demonstrated that vertical stress and relative density played a vital role in predicting damping ratio, while all three parameters were important in predicting secant shear modulus. In this study, two developed artificial intelligence models were compared with existing literature models. According to the results, for test database, the existing models were able to predict the shear modulus and damping ratio with R of 0.911 and 0.918, respectively. However, the proposed ANN and CRRF models were able to predict shear modulus with R of 0.993 and 0.996, and damping ratios with R of 0.992 and 0.990, respectively. The results showed that ANNs and CRRFs were more robust than existing models for predicting damping ratio and shear modulus, as well as identifying the influence of input variables on sand dynamic properties.

Data-driven prediction of energy consumption of district cooling systems (DCS) based on the weather forecast data

Accurate prediction of energy consumption is the theoretical basis to achieve the low-carbon operation and maintenance of building HVAC systems. Existing studies either use only partial weather data resulting in low accuracy or need to build a three-dimensional building model which is difficult to achieve. The aim of this study is to develop a simple and convenient model for the energy consumption prediction of DCS using easily obtained weather data. To achieve the above purpose, this investigation first analyzes the linear correlation between the outdoor environmental parameters with the actual measured energy consumption data. It indicated that partial weather data and energy consumption data have the highest linear correlation. Then, the backpropagation artificial neural network (BP-ANN) algorithm was adopted to construct the energy consumption prediction model and the main parameters affecting its prediction performance were determined. Finally, the accuracy of the proposed BP-ANN model was tested. The results showed that the proposed energy consumption prediction model driven by weather data was suitable for the energy consumption prediction of a district comprising multiple buildings without requiring details of buildings and systems. In addition, the proposed method is not only helpful to the energy consumption prediction and intelligent operation and maintenance of DCS, but also can be transferred to any district comprising multiple buildings easily.

Genetic optimized Al–Mg alloy constitutive modeling and activation energy analysis

As an intelligent global optimization method, the genetic algorithm has tremendous potential for improving flow behavior modeling and analysis. Based on flow stress-true strain curves of Al–Mg AA5005 alloy under temperature 563∼773 K and strain rate 0.0003∼0.03s−1, a phenomenological model named Arrhenius-type (A-T) was established to describe the flow behavior. On this basis, the genetic optimized A-T (GA-T) model with higher precision was obtained by optimizing A-T parameters α, n, Q and lnA. To reduce the large computing power consumed by unnecessary complex topological network structure when conducting simulations by the back propagation artificial neural network (BP-ANN) model, a genetic optimized BP-ANN (GBP-ANN) model was designed through determining the initial values of weights, biases and hyper parameters. The presented GBP-ANN model inherits the advantage of the BP-ANN model’s high accuracy as well as maintaining the simplest structure. The statistical analysis demonstrates that the GBP-ANN model possesses the best flow behavior description ability among three established models. Moreover, the GBP-ANN also shows a better generalization performance than the GA-T model. Lastly, with the help of the GA-T model, the activation energy map was plotted to determine the desirable deformation condition analyze the deformation mechanism. Our work presents a combination of GBP-ANN model and genetic optimized Q analysis, thus shedding new light on high accuracy flow behavior modeling and deformation mechanism analysis.

Adaptive fuze-warhead coordination method based on BP artificial neural network

The appropriate fuze-warhead coordination method is important to improve the damage efficiency of air defense missiles against aircraft targets. In this paper, an adaptive fuze-warhead coordination method based on the Back Propagation Artificial Neural Network (BP-ANN) is proposed, which uses the parameters of missile-target intersection to adaptively calculate the initiation delay. The damage probabilities at different radial locations along the same shot line of a given intersection situation are calculated, so as to determine the optimal detonation position. On this basis, the BP-ANN model is used to describe the complex and highly nonlinear relationship between different intersection parameters and the corresponding optimal detonating point position. In the actual terminal engagement process, the fuze initiation delay is quickly determined by the constructed BP-ANN model combined with the missile-target intersection parameters. The method is validated in the case of the single-shot damage probability evaluation. Comparing with other fuze-warhead coordination methods, the proposed method can produce higher single-shot damage probability under various intersection conditions, while the fuze-warhead coordination effect is less influenced by the location of the aim point.

Intelligent prediction of wear life of automobile brake pad based on braking conditions

PurposeTo avoid braking accidents caused by excessive wear of brake pad, this study aims to achieve online prediction of brake pad wear life (BPWL).Design/methodology/approachA simulated braking test bench for automobile disc brake was used. The correlation and mechanism between the three braking condition parameters of initial braking speed, braking pressure and initial braking temperature and the tribological performance were analyzed. The different artificial neural network (ANN) models of wear loss were discussed. Genetic algorithm was used to optimize the ANN model. The structure scheme of the online prediction system of BPWL was discussed and completed.FindingsThe results showed that the braking conditions were positively correlated with the wear loss, but negatively correlated with the friction coefficient. The prediction accuracy of back propagation (BP) ANN model was higher. The model was optimized by genetic algorithm, and the average deviation of prediction results was 4.67%. By constructing the online monitoring system of automobile braking conditions, the online prediction of BPWL based on the ANN model could be realized.Originality/valueThe research results not only have important theoretical significance for the study of BPWL but also have practical value for guiding the maintenance and replacement of automobile brake pads and avoiding the occurrence of braking accidents.

Application of Artificial Neural Network to Predict the Properties of Permeable Concrete

The structure of permeable concrete has been the primary reason for its use in construction. Permeable concrete is composed of water, cement, aggregate, and little- to no-fines resulting in the presence of a significant number of voids. This makes permeable concrete an ideal solution to water accumulation issues as it acts as a drainage system. This study employs a feedforward backpropagation artificial neural network model that combines experimental laboratory data from previous studies with appropriate network architectures and training techniques. The purpose of the analysis is to develop a reliable functional relationship, based on water-cement ratio, aggregate-cement ratio, and density parameters, with which to estimate the compressive strength, porosity, and water permeability of permeable concrete. Multiple linear regression correlations are also established to predict and correlate these inputs and outputs. The two derived methods are then compared and discussed. The results reveal that ANN is better to anticipate the permeable concrete properties than regression analysis.

Testing Method for Intelligent Loading of Mining Emulsion Pump Based on Digital Relief Valve and BP Neural Network Control Algorithm

The mining emulsion pump is mainly used on a fully mechanized coal mining face, but it is rarely used on other occasions, so research on its loading test method is relatively limited. This paper proposes the application of a digital relief valve to the emulsion pump loading test. In addition, the small number of plungers in the emulsion pump will lead to large flow pulsation and pressure pulsation, and the nominal flow of different types of emulsion pumps varies greatly. These factors lead to the deficiency of a traditional PID control algorithm in control accuracy and efficiency. In order to improve control accuracy and efficiency, firstly, the influence of the flow rate of the tested pump and extension of the linear stepping motor shaft on the working pressure is studied. A backpropagation (BP) artificial neural network (ANN) model is used to fit a functional relationship between the three parameters. The flow rate of the tested pump and target pressure were provided as inputs to predict the extension of the linear stepping motor shaft, thereby realizing the remote intelligent control of the system pressure. Next, a BP ANN model is constructed, and its reliability is verified; the BP neural network algorithm and proportional-integral-derivative (PID) algorithm are compared through simulation. The simulation results show that the BP neural network algorithm has high control accuracy and small overshoot. Finally, two pumps with different flows are tested in a self-developed digital relief valve and test platform. The test results show that the proposed loading test method is intelligent and efficient, and it has high accuracy.

Prediction of gamma ray spectrum for 22Na source by feed forward back propagation ANN model

The radiation has been used in a variety of different fields since its discovery and thus its measurement becomes vital in these industries. Different type detector may be used to measure gamma rays depends on the purposes of measurements. Gamma ray energy spectrum is an important to determine either elemental analysing of a sample or radiation shielding purposes. On the other hand, Artificial Neural Network (ANN) may be used to predict and analysing of gamma-ray spectrum. In this study, gamma ray spectrum from 22Na source detected in NaI (Tl) detector was estimated by ANN. There have been installed ten different ANN models to find the network structure that produces the best predictive value for the gamma ray spectrum NaI (Tl) Detector. Estimation study has been continued with the ANN model with be possessed of lowest error value. ANN model was created by using energy, distance and gamma-rays energy spectrum (called Io) values. In the ANN model developed using the feed forward back propagation algorithm, were used artificial neurons two in the input layer, ten in the hidden layer and one in the output layer. For the case of present work, the experimental data was used 70% for education, 20% for validation and 10% for testing. The estimated values obtained with the ANN model were compared with the experimental results and a good correlation has been found between them (R2 = 0.99).

Peramalan Kurs Rupiah Terhadap Dolar Amerika Menggunakan Jaringan Saraf Tiruan

The Indonesian rupiah (IDR) exchange rate is used to gauge Indonesia's economic stability. Maintaining the IDR exchange rate's stability is critical since it has a direct impact on Indonesia's national monetary situation, particularly during the Covid-19 pandemic. Forecasting is one way to assess government policy in terms of lowering the exchange rate. The goal of this study is to use the backpropagation artificial neural network model to model and predict the IDR exchange rate. This study uses daily data on the US Dollar (USD) to Indonesian Rupiah (IDR) exchange rate from March 2020 to December 2021. The best BPNN model is BP (2,5,1) with 2 neurons in the input layer, 5 neurons in the hidden layer, and 1 neuron in the output layer. The accuracy of prediction of this model is very good with an RMSE value is 33,66 and MAPE value is 0,1796%.

Modeling of processing parameters for the cutting of rolled tungsten plates via abrasive water jet

Tungsten is a promising candidate for plasma-facing material in tokamaks. This study aims at determining whether the cut-surface roughness of rolled tungsten plates (RTPs) can be improved using an abrasive water jet (AWJ) through tests combined with modeling of the cutting-process parameters. Based on the key factors affecting the cut-surface roughness, the magnitude and type of test factors were increased, compared with the previous study. A batch test was designed and conducted to investigate the use of an AWJ to cut an RTP and obtain more comprehensive sample data. In addition, the multivariate linear regression method was employed to establish a multivariate quadratic response surface regression model of the relationship between the cutting-process parameters and the cut-surface roughness based on the roughness sample data. Moreover, a backpropagation (BP) artificial neural network (ANN) model was developed based on the cutting-process parameters to predict the cut-surface roughness. The two models were compared and experimentally validated, which showed that the BP ANN model exhibits relatively high accuracy. Thus, the established model can help to improve cut-surface quality and cutting efficiency.

A back propagation neural network model for accurately predicting the removal efficiency of ammonia nitrogen in wastewater treatment plants using different biological processes

Accurately predicting the water quality of treated water from a water treatment plant (WWTP) based on the obtained operating database is of great significance. However, it is difficult for common mechanistic models to work well. In this study, a back propagation artificial neural network (BPANN) model with high accuracy was developed to predict the denitrification efficiency based on a 1-year operating database. Standardized principal component analysis (PCA) methods were used to address the data, and the PCA processed data exhibited the best accuracy. In three WWTPs adopting the anaerobic/anoxic/oxic (A2O) process, the ammonia nitrogen removal efficiency of WWTPs was successfully predicted by using five variables: inlet flow rate, pH value, original ammonia nitrogen concentration, Chemical oxygen demand (COD) concentration, and total phosphorus concentration. Importantly, the obtained BPANN model can be effectively used for other widely used treatment processes, such as oxidation ditch (OD), sequencing batch reactor activated sludge process (SBR), membrane bioreactor (MBR), and cyclic activated sludge technology (CAST), by simply optimizing the training data ratios between 50/50 and 90/10. This is the first trial to set up a universal model for predicting the denitrification efficiency of WWTPs adopting common biological processes. The model could be used to choose the optimum treatment process in the new WWTP design or take action in advance to avoid the risk of excessive emissions when the already built WWTPs are subjected to sudden shocks.

Designing bulk metallic glasses materials with higher reduced glass transition temperature via machine learning

Bulk metallic glasses (BMGs) are considered as one of the most promising candidates in the field of alloy materials due to their outstandingly high yield strengths and large elastic limits at ambient conditions compared to their crystalline counterparts. The formation of BMGs is determined by reduced glass transition temperature (Trg) that represents the glass forming ability (GFA). Experimental search for BMGs with higher Trg is a challenging task due to the vast chemical space and lack of predictive guidelines. Here, we demonstrate a machine learning approach to guide experimental synthesis in search of BMGs with higher Trg. The relationships between the Trg and chemical compositions of BMGs were investigated by using three different machine learning algorithms including support vector regression (SVR), back propagation artificial neural network (BPANN) and gradient boosting regression (GBR). The best predictions from SVR, BPANN, and GBR models were recommended for experimental tests. Based on the leave-one-out cross-validation (LOOCV), the correlation coefficient (R) and root mean squared error (RMSE) between predicted and experimental Trg were 0.894 and 0.0116, respectively. Furthermore, virtual screening based on machine learning results was adopted to design BMGs with higher Trg. The average relative error of Trg from newly synthesized BMGs is 1.00%, which is satisfactory compared to that of the known experimental samples (3.91%). Among the newly synthesized BMGs, Mg0.65Ag0.19Cu0.06Gd0.1 owns the highest Trg of 0.6658, exceeding the highest Trg (0.6576) in the reported experiments before.

Changes in quality characteristics of shrimp (Penaeus chinensis) during refrigerated storage and their correlation with protein degradation

Shrimp are highly prone to spoilage during refrigerated storage. Systematic assessment of shrimp protein degradation and quality loss during refrigerated storage is essential for optimizing storage of shrimps. In this study, protein biochemical characteristics and quality indicators of shrimp during refrigerated storage at 4 °C were assessed, and back-propagation artificial neural networks (BP-ANNs) were used to elucidate correlations of quality characteristics with protein degradation. With increasing storage time, Ca2+-ATPase activity, total amino acids, sulfhydryl group values, elasticity, and stickiness decreased, whereas protein carbonyl group values, total volatile base nitrogen content, and biogenic amine content increased. Furthermore, SDS-PAGE and transmission electron microscopy revealed protein degradation and damage to shrimp muscle microstructure during storage. With protein degradation characteristics as input values and quality indicators as output values, BP-ANNs models were established to analyze their correlation. A high correlation coefficient of 0.9654 and low mean square error of 0.0069 were obtained, indicating that BP-ANN can reliably illustrate the correlation between protein degradation and shrimp loss. Our results provided basic data and technical support for the intelligent storage of shrimp.

Identification of novel serological autoantibodies in Chinese prostate cancer patients using high-throughput protein arrays.

Autoantibody (AAb) has a prominent role in prostate cancer (PCa), with few studies profiling the AAb landscape in Chinese patients. Therefore, the AAb landscape in Chinese patients was characterized using protein arrays. First, in the discovery phase, Huprot arrays outlined autoimmune profiles against ~ 21,888 proteins from 57 samples. In the verification phase, the PCa-focused arrays detected 25 AAbs selected from the discovery phase within 178 samples. Then, PCa was detected using a backpropagation artificial neural network (BPANN) model. In the validation phase, an enzyme-linked immunosorbent assay (ELISA) was used to validate four AAb biomarkers from 196 samples. Huprot arrays profiled distinct PCa, benign prostate diseases (BPD), and health AAb landscapes. PCa-focused array depicted that IFIT5 and CPOX AAbs could distinguish PCa from health with an area under curve (AUC) of 0.71 and 0.70, respectively. PAH and FCER2 AAbs had AUCs of 0.86 and 0.88 in discriminating PCa from BPD. Particularly, PAH AAb detected patients in the prostate-specific antigen (PSA) gray zone with an AUC of 0.86. Meanwhile, the BPANN model of 4-AAb (IFIT5, PAH, FCER2, CPOX) panel attained AUC of 0.83 among the two cohorts for detecting patients with gray-zone PSA. In the validation cohort, the IFIT5 AAb was upregulated in PCa compared to health (p < 0.001). Compared with BPD, PAH and FCER2 AAbs were significantly elevated in PCa (p = 0.012 and 0.039). We have demonstrated the first extensive profiling of autoantibodies in Chinese PCa patients, identifying novel diagnostic AAb biomarkers, especially for identification of gray-zone-PSA patients.

Hot deformation behaviors of AZ91 magnesium alloy: Constitutive equation, ANN-based prediction, processing map and microstructure evolution

The hot deformation behavior in an AZ91 magnesium alloy was studied using Arrhenius model and Back-Propagation Artificial Neural Network (BP-ANN) approaches. The hot compression tests of AZ91 alloy were performed on the MMS-300 simulator at deformation temperatures and strain rates in the range of 473–623 K and 0.001 s −1 - 1 s −1 , respectively. The results indicated that the BP-ANN model has higher accuracy than Arrhenius model and the correlation coefficient is as high as 0.99341. The processing map is divided into three parts: machinable area (A), instability area I (INS I) (B) and instability area II (INS II) (C). The instability mode of areas B and C gradually changes from cracks to holes with the increase of temperature or the decrease of strain rate. The stable hot-deformation conditions with peak efficiency of power dissipation were concentrated in the range of 523 K, 0.01 s −1 , 573 K, 0.01 s −1 , 623 K, 0.001–1 s −1 , respectively. With the increase of temperature, the deformation mechanism changed from twinning to dynamic recrystallization (DRX), which is discontinuous dynamic recrystallization (DDRX). Due to DRX,< 0001 > // CD texture intensity increased with the increase of temperature, and then decreased. ● BP-ANN is used to predict stress and strain. ● Instability mode: crack and hole. ● Deformation mechanism: twinning, slip and DRX.

Assessment of maize seed vigor under saline-alkali and drought stress based on low field nuclear magnetic resonance

To detect maize seed vigor under salt-alkaline and drought stress conditions, transverse relaxation time, physiological index, and electron microscopy images of germinating seeds were studied under different stress conditions. The results showed that the water in germinating maize seeds exist in bound water (T21), semi-bound water (T22), and free water (T23), in addition parabolic water (T24). The Pearson correlation analysis was performed with T2 relaxation parameters and seed vigor parameters, and this analysis yield nine optimized parameters. To predict vigor levels under different stress conditions, an error backpropagation artificial neural network model was developed, wherein the T2 chirality parameter was used as the input value, and the seed germination indices of different stress levels were used as the output values. The model could predict the environmental stress level of maize seed growth. The optimized parameter set showed a prediction accuracy of 92.50%, thus outperforming the T2 relaxation information model without parameter optimization (75.01%). The proposed method can collect data during the germination of maize seeds without any interference and achieve large-scale prediction of seed development status by small sample collection. With the stress environment was aggravated, the physiological structure of seed cells was changed, and the cell structure was destroyed and the ability of water absorption was disappeared. This analysis provides theoretical support and a reference basis for maize planting and production.

Flow Behavior of AA5005 Alloy at High Temperature and Low Strain Rate Based on Arrhenius-Type Equation and Back Propagation Artificial Neural Network (BP-ANN) Model.

To realize the purpose of energy saving, materials with high weight are replaced by low-weight materials with eligible mechanical properties in all kinds of fields. Therefore, conducting research works on lightweight materials under specified work conditions is extremely important and profound. To understand the relationship of aluminum alloy AA5005 among flow stress, true strain, strain rate, and deformation temperature, hot isothermal tensile tests were conducted within the strain rate range 0.0003–0.03 s−1 and temperature range 633–773 K. Based on the true stress-true strain curves obtained from the experiment, a traditional constitutive regression Arrhenius-type equation was utilized to regress flow behaviors. Meanwhile, the Arrhenius-type equation was optimized by a sixth-order polynomial function for compensating strain. Thereafter, a back propagation artificial neural network (BP-ANN) model based on supervised machine learning was also employed to regress and predict flow stress in diverse deform conditions. Ultimately, by introducing statistical analyses correlation coefficient (R2), average absolute relative error (AARE), and relative error () to the comparative study, it was found that the Arrhenius-type equation will lose accuracy in cases of high stress. Additionally, owning higher R2, lower AARE, and more concentrative value distribution, the BP-ANN model is superior in regressing and predicting than the Arrhenius-type constitutive equation.