Data-driven Artificial Neural Network Research Articles

Deep machine learning approaches for battery health monitoring

The performance and longevity of batteries can be measured through battery parameters, like: state of charge, state of health, and remaining useful life. Therefore, accurate estimation of these parameters is essential for developing efficient battery management systems. Recently, a variety of smart model-based and data-driven artificial neural network methods have been evolved for efficient management of battery systems. However, there is space for improving the accuracy and reliability of the proposed methods, particularly in the prediction and forecasting of battery performance metrics. In this study, a deep machine learning based forecasting technique has been proposed to better estimate the battery performance parameters. Available open source data-sets on three distinct battery technologies comprising of battery cycling profiles and environmental conditions, are adopted to construct and validate models for predicting battery health. A common data-set is developed by combining the open source data, which is later supplied as input to the predictive analytics and time series analysis. Encoders and decoders are employed for the extraction of important and relevant feature information from the data-set. Three different network model-based architectures are developed to predict the state of charge of Lithium-ion batteries. The proposed models are observed to accurately capture the dynamic behavior of the battery during discharge, which is confirmed by the analysis of the recorded voltage, current, and temperature data, and the precise estimation of state of charge. Furthermore, a novel forecasting method based on the time-series analysis is also introduced. Mean average error values less than 0.2% are observed when neural basis expansion analysis for interpretable time series architecture in conjunction with encoder–decoder-based feature extraction is employed.

Thermomechanical problem of functionally graded spherical shells based on homogenization schemes: Data-driven volume fraction optimization with material uncertainties

Mechanical analysis and optimal design for functionally graded materials (FGMs) structures are significant. The present paper analyzes the parameter uncertainties problem for the thermomechanical response of the FGMs spherical shells with volume fraction power distribution. Here, the interval random uncertainty model is recommended to describe the uncertainties of each component of material parameters in FGMs, and a data-driven artificial neural network (ANN) fast calculation for FGMs spherical shells thermomechanical problems with uncertainties analysis is established and trained. The thermomechanical coupling response analysis and volume fraction optimization analysis of the FGMs spherical shell with material uncertainties are realized for zirconia-titanium FGMs, with the maximum normalized Von Mises stress as the structural safety index and the ANN as the fast calculation method. In addition, the Von Mises stress of the FGMs spherical shell and a ZrO2-Ti alloy double-layer spherical shell under different loads is compared. The results show that material parameter uncertainties, especially those in Young's modulus, thermal expansion coefficient, and uniaxial yield limit significantly affect structural safety. The structural safety index λ calculated based on the volume fraction optimization design of FGMs spherical shell is significantly lower than that of the ZrO2-Ti alloy double-layer. The ANN method greatly improves the computational efficiency for uncertain problems and maintains high accuracy, compared with the statistical analysis after many calculations by randomly selecting material parameters. This study provides a path for the safety design of FGMs structures with uncertain parameters.

Augmentation of scarce data—A new approach for deep-learning modeling of composites

High-fidelity full-field micro-mechanical modeling of the non-linear path-dependent materials demands a substantial computational effort. Recent trends in the field incorporates data-driven Artificial Neural Networks (ANNs) as surrogate models. However, ANNs are inherently data-hungry, functioning as a bottleneck for the development of high-fidelity data-driven models. This study introduces a novel approach for data augmentation, expanding an original dataset without additional computational simulations. A Recurrent Neural Network (RNN) was trained and validated on high-fidelity micro-mechanical simulations of elasto-plastic short fiber reinforced composites. The obtained results showed a considerable improvement of the network predictions trained on expanded datasets using the proposed data augmentation approach. The proposed method for augmentation of scarce data may be used not only for other kind of composites, but also for other materials and at different length scales, and hence, opening avenues for innovative data-driven models in materials science and computational mechanics.

Land cover change analysis of buffer areas in New Capital City of Nusantara, Indonesia: A cellular automata approach on satellite imageries data

The proposed plan to move Indonesia's capital city to the New Capital City of Nusantara in East Kalimantan Province undoubtedly requires careful efforts to ensure food supply for the population. Population migration to the new capital may pose a food security challenge. To address this fundamental issue, one of the most crucial approaches is to establish buffer areas that can support the food needs of the new capital. The currently existing official Area Sampling Frame survey conducted by the government to assess food vulnerability faced several limitations, including weather conditions, field terrain variations, and high cost. In this study, we propose the utilization of remote sensing satellite imagery data in buffer areas to analyze changes and predict future land cover, which can provide valuable data for assessing food availability. We investigate the integration of a Cellular Automata method with the two most popular analytical methods of classical Logistic Regression and data-driven Artificial Neural Networks, known as CA-LR and CA-ANN, to identify and map land cover changes in the new capital buffer zones. Our findings reveal that both combined methods, CA-LR and CA-ANN, yield fairly promising results, with correctness and kappa statistic values exceeding 80%. Prediction results indicate that buffer areas are predominantly covered by trees, while built-up areas are still limited. The flooded vegetation cover, including rice fields, is predicted to decrease by 2024. This should be a matter of concern for stakeholders, considering the construction of the new capital city is still ongoing and the number of migrants is expected to keep rising.

A Data-Driven Artificial Neural Network Approach to Software Project Risk Assessment

In the realm of software project management, predicting and mitigating risks are pivotal for successful project execution. Traditional risk assessment methods have limitations in handling complex and dynamic software projects. This study presents a novel approach that leverages artificial neural networks (ANNs) to enhance risk prediction accuracy. We utilize historical project data, encompassing project complexity, financial factors, performance metrics, schedule adherence, and user-related variables, to train the ANN model. Our approach involves optimizing the ANN architecture, with various configurations tested to identify the most effective setup. We compare the performance of mean squared error (MSE) and mean absolute error (MAE) as error functions and find that MAE yields superior results. Furthermore, we demonstrate the effectiveness of our model through comprehensive risk assessment. We predict both the overall project risk and individual risk factors, providing project managers with a valuable tool for risk mitigation. Validation results confirm the robustness of our approach when applied to previously unseen data. The achieved accuracy of 97.12% (or 99.12% with uncertainty consideration) underscores the potential of ANNs in risk management. This research contributes to the software project management field by offering an innovative and highly accurate risk assessment model. It empowers project managers to make informed decisions and proactively address potential risks, ultimately enhancing project success.

Comparative evaluation of daily streamflow prediction by ANN and SWAT models in two karst watersheds in central south Texas

Abstract This work compares the accuracy of streamflow estimated by a data-driven artificial neural network (ANN) and the physically based soil and water assessment tool (SWAT). The models were applied in two small watersheds, one highly urbanized and the other primarily covered with evergreen forest and shrubs, in the San Antonio region of central south Texas, where karst geologic features are prevalent. Both models predicted daily streamflow in the urbanized watershed very well, with the ANN and SWAT having the Nash–Sutcliffe coefficient of efficiency (NSE) values of 0.76 and 0.72 in the validation period, respectively. However, both models predicted streamflow poorly in the nonurban watershed. The NSE values of the ANNs significantly improved when a time series autoregressive model structure using historical streamflow data was implemented in the nonurban watershed. The SWAT model only achieved trivial performance improvement after using the SWAT-CUP SUFI-2 calibration procedure. This result suggests that an ANN model may be more suitable for short-term streamflow forecasting in watersheds heavily affected by karst features where the complex processes of rapid groundwater recharge and discharge strongly influence surface water flow.

Experimental Investigation on the Effect of Graphene Oxide in Higher Alcohol Blends and Optimization of Injection Timing Using an ANN Method.

The use of alternative fuels in diesel engines has become more widespread due to a number of factors, including dwindling petroleum supplies, increasing prices for conventional fossil fuels, and environmental worries about pollutants and greenhouse gas emissions from internal combustion engines. Efficiency and emissions need to be appropriately balanced. Alcohols act as oxygenated fuels similar to octanol, offering a number of benefits over traditional fuels and can boost efficiency, enhance combustion, and reduce air pollution. Therefore, the research aimed to enhance the performance and combustion characteristics of a diesel and octanol blend using graphene oxide (GO) nanoparticles as a fuel additive in a single-cylinder diesel engine while reducing emissions. Research findings will contribute significantly to improving the physical and chemical properties of diesel and octanol blends, thereby mitigating the challenges of limited petroleum reserves and environmental concerns. A range of different blends of diesel and octanol were prepared on a volume/volume basis in proportions of D70OCT30, D60OCT40, and D50OCT50, and then GO was added as a fuel additive to the abovementioned blends in varied proportions (40, 60, and 80 ppm) resulting in nine blends. These blends were analyzed in terms of various performance, combustion, and emission characteristics, and the obtained results helped to shed light on the impact of GO as a fuel additive. The results indicated that the fuel blend D70OCT30GO0.006 yielded the highest values. Furthermore, it is highly imperative that we develop a model that can be used to predict engine behavior and its stability without having to run an engine. For this, a data-driven artificial neural network (ANN) model was developed to predict the optimized injection timing for better combustion and reduced emission. The efficiency and prediction capabilities of the model were compared to the experimental data, which indicated that the ANN model had a better prediction score. The injection timing of the engine was optimized from 21 °CA to 21.5 °CA, which increased the efficiency by 1%. The research findings showed significantly improved physical and chemical properties of the blends, thereby mitigating the challenges of limited petroleum reserves and environmental concerns.

Development of artificial neural network models for indirect evaporative coolers in multi-climate regions based on field measurement

Data-driven artificial neural networks (ANN) based on data drivers with their powerful self-learning capability and high adaptability are gaining increasing attention for the application in modeling indirect evaporative coolers (IEC). However, most ANN models of IEC in existing studies are either limited to a specific climate region or conventional IEC configurations. In this paper, a multi-region back-propagation (BP) neural network model for predicting the performance of advanced dew-point IECs is developed. Operational data for ANN model development were collected through field measurements in IEC projects in three typical climate zone cities in China (Dunhuang, Yulin, and Fuzhou), covering arid, moderately wet, and humid regions. A comparative study of three single-region neural network models was conducted in terms of both convergence characteristics and statistical performance metrics. Each model contains two input variables (inlet air temperature, inlet air humidity) and one output variable (wet-bulb efficiency). The results show that the model fits best in Yulin, followed by Dunhuang and Fuzhou, with total correlation coefficients of 0.9918, 0.9477, and 0.8946, respectively. The predicted values of wet-bulb efficiency are in good agreement with the actual operating data, and the deviation of almost all predicted values is within ±10%. Practical application The application values of IEC ANN models are as follows. First, the IEC-ANN model can predict IEC performance adaptively based on dynamic operational data. Therefore, it can provide the best operating strategy and design parameters for different situations. In addition, it can provide a fast response to guide the system operation when auxiliary control is required. Most importantly, this new approach requires only a limited number of tests rather than exhaustive experimental studies or dealing with complex mathematical models, and future manufacturers can use neural network technology to evaluate the performance of dew-point indirect evaporative coolers, thus saving engineering budget and time.

Prediction of Frictional Moment of Cylindrical Roller Bearing Using Experimental Data-Driven Artificial Neural Networks

Abstract Accurate prediction of the frictional moment of the bearing contributes to the correct determination of the power loss in drivetrains and the antifriction design of bearings. This paper investigates a method for accurately predicting the frictional moment of the cylindrical roller bearing (CRB) under a wide range of operating conditions. The complex relationship between the bearing frictional moment and multiple operating parameters such as the shaft speed, roller–raceway contact load, cage slip ratio and lubricating property is established using an experimental data-driven artificial neural network (ANN) model. To provide actual data for training and testing the ANN model, the frictional moment and multiple operating parameters of the test CRB are synchronously measured under many test conditions. Compared with the prediction results from conventional physical models, the experimental data-driven ANN model reveals a higher prediction performance of the frictional moment.

Data-driven artificial neural network for elastic plastic stress and strain computation for notched bodies

An integration of artificial neural network (ANN) and finite element (FE) analysis was developed to predict elastic–plastic stress and strains at notch locations based on a linear elastic FE analysis solution. Two FE models were created for different materials and multiaxial load cases to produce hypothetical elastic and elastic–plastic stress and strain datasets. One model was based on an elastic state, and the other was based on an elastic–plastic state. The ANN was trained with the elastic stress data from the linear elastic FE model as input and the elastic–plastic stress–strain data from the nonlinear elastic–plastic FE model as output. The dataset was divided into three groups: training, verification, and testing data. The ANN was trained using the training data, evaluated using the verification data, and tested for generalizability using the testing data. The results showed that the proposed methodology can predict elastic–plastic stress and strain fields for notched bodies under multiaxial loadings accurately and efficiently using only the elastic FE analysis solution.

A probabilistic framework for source localization in anisotropic composite using transfer learning based multi-fidelity physics informed neural network (mfPINN)

The practical application of data-driven frameworks like deep neural network in acoustic emission (AE) source localization is impeded due to the collection of significant clean data from the field. The utility of the such framework is governed by data collected from the site and/or laboratory experiment. The noise, experimental cost and time consuming in the collection of data further worsen the scenario. To address the issue, this work proposes to use a novel multi-fidelity physics-informed neural network (mfPINN). The proposed framework is best suited for the problems like AE source detection, where the governing physics is known in an approximate sense (low-fidelity model), and one has access to only sparse data measured from the experiment (high-fidelity data). This work further extends the governing equation of AE source detection to the probabilistic framework to account for the uncertainty that lies in the sensor measurement. The mfPINN fuses the data-driven and physics-informed deep learning architectures using transfer learning. The results obtained from the data-driven artificial neural network (ANN) and physics-informed neural network (PINN) are also presented to illustrate the requirement of a multi-fidelity framework using transfer learning. In the presence of measurement uncertainties, the proposed method is verified with an experimental procedure that contains the carbon-fiber-reinforced polymer (CFRP) composite panel instrumented with a sparse array of piezoelectric transducers. The results conclude that the proposed technique based on a probabilistic framework can provide a reliable estimation of AE source location with confidence intervals by taking measurement uncertainties into account.

Two methods for studying the response and the reliability of a fractional stochastic dynamical system

In this paper, we use stochastic averaging method and data-driven artificial neural network to study the response and reliability of a fractional stochastic dynamical system. Firstly, a generalized Harmonic transformation is introduced to get an approximate expression for the fractional derivative. Based on this procedure, stochastic averaging method is utilized to get Fokker–Planck Kolmogorov equation for the system response and Backward Kolmogorov equation for the system reliability. An exact stationary solution of amplitude response is obtained by solving Fokker–Planck Kolmogorov equation, in the meanwhile, numerical results regarding to the reliability by Crank–Nicolson difference method are derived too. We propose an algorithm of data-driven artificial neural network and apply it for solving these two equations simulatively, in order to verify the correction and accuracy of the analytical methods. All results illustrate that data-driven artificial neural network is highly effective to achieve the system response and reliability. The advantage of this deep learning algorithm mainly includes the meshless structure, boundary sample data collection, small amount of training data and unconstrained optimization. In addition, the increase of the fractional order can enhance the system response and leads to stochastic bifurcation.

A data-driven artificial neural network model for the prediction of ground motion from induced seismicity: The case of The Geysers geothermal field

Ground-motion models have gained foremost attention during recent years for being capable of predicting ground-motion intensity levels for future seismic scenarios. They are a key element for estimating seismic hazard and always demand timely refinement in order to improve the reliability of seismic hazard maps. In the present study, we propose a ground motion prediction model for induced earthquakes recorded in The Geysers geothermal area. We use a fully connected data-driven artificial neural network (ANN) model to fit ground motion parameters. Especially, we used data from 212 earthquakes recorded at 29 stations of the Berkeley–Geysers network between September 2009 and November 2010. The magnitude range is 1.3 and 3.3 moment magnitude (Mw), whereas the hypocentral distance range is between 0.5 and 20 km. The ground motions are predicted in terms of peak ground acceleration (PGA), peak ground velocity (PGV), and 5% damped spectral acceleration (SA) at T=0.2, 0.5, and 1 s. The predicted values from our deep learning model are compared with observed data and the predictions made by empirical ground motion prediction equations developed bySharma et al. (2013)for the same data set by using the nonlinear mixed-effect (NLME) regression technique. For validation of the approach, we compared the models on a separate data made of 25 earthquakes in the same region, with magnitudes ranging between 1.0 and 3.1 and hypocentral distances ranging between 1.2 and 15.5 km, with the ANN model providing a 3% improvement compared to the baseline GMM model. The results obtained in the present study show a moderate improvement in ground motion predictions and unravel modeling features that were not taken into account by the empirical model. The comparison is measured in terms of both theR2statistic and the total standard deviation, together with inter-event and intra-event components.

Developing Data-driven Artificial Neural Network for a High Throughput Retrieval of Aerosol Optical Depth and Surface Temperature of Mars

In this work, we aim to develop artificial neural network (ANN) techniques to reproduce the retrieval results of physical quantities from spacecraft observations of solar system bodies using radiative transfer methods. The particular application here is the retrieval of dust optical depth, water ice optical depth, and surface temperature on Mars using daytime observations obtained by the Thermal Emission Spectrometer on board the Mars Global Surveyor. Compared against the results obtained from traditional radiative transfer retrieval techniques, our ANN successfully recovered the three quantities using daytime observations. The principal advantage of these machine-learning algorithms is their complete automation and high throughput. Therefore, the algorithms presented here would be useful for very large data sets and would make practical the sampling of many different approximations or boundary conditions related to a given observation data set and retrieval problem.

A Single Subject, Feasibility Study of Using a Non-Contact Measurement to "Visualize" Temperature at Body-Seat Interface.

Measuring temperature changes at the body-seat interface has been drawing increased attention from both industrial and scientific fields, due to the increasingly sedentary nature from daily leisure activity to routine work. Although contact measurement is considered the gold standard, it can affect the local micro-environment and the perception of sitting comfort. A non-contact temperature measurement system was developed to determine the interface temperature using data gathered unobtrusively and continuously from an infrared sensor (IRs). System performance was evaluated regarding linearity, hysteresis, reliability and accuracy. Then a healthy participant sat for an hour on low/intermediate density foams with thickness varying from 0.5–8 cm while body-seat interface temperature was measured simultaneously using a temperature sensor (contact) and an IRs (non-contact). IRs data were filtered with empirical mode decomposition and fractal scaling indices before a data-driven artificial neural network was utilized to estimate the contact surface temperature. A strong correlation existed between non-contact and contact temperature measurement ( > 0.85) and the estimation results showed a low root mean square error (RMSE) (<0.07 for low density foam and <0.16 for intermediate density foam) and high Nash-Sutcliff efficiency (NSE) values (≈1 for both types of foam materials).

Post necking evaluation of the tensile test using artificial neural networks

This paper introduces a new method for evaluating the tensile test using finite element simulations and data-driven artificial neural networks. For this purpose, a synthetic data set was generated by finite element simulations using LS-DYNA. Artificial neural networks of two different topologies were trained and tested on parts of this synthetic data set. The networks use geometry information of the necking area as input data to predict a correction factor to convert the stress obtained from the tensile test to the equivalent flow stress. The best models are evaluated on the test set and good results are achieved.

Likelihood of Transformation to Green Infrastructure Using Ensemble Machine Learning Techniques in Jinan, China

Rapid urbanization influences green infrastructure (GI) development in cities. The government plans to optimize GI in urban areas, which requires understanding GI spatiotemporal trends in urban areas and driving forces influencing their pattern. Traditional GIS-based methods, used to determine the greening potential of vacant land in urban areas, are incapable of predicting future scenarios based on the past trend. Therefore, we propose a heterogeneous ensemble technique to determine the spatial pattern of GI development in Jinan, China, based on driving biophysical and socioeconomic factors. Data-driven artificial neural networks (ANN) and random forests (RF) are selected as base learners, while support vector machine (SVM) is used as a meta classifier. Results showed that the stacking model ANN-RF-SVM achieved the best test accuracy (AUC 0.941) compared to the individual ANN, RF, and SVM algorithms. Land surface temperature, distance to water bodies, population density, and rainfall are found to be the most influencing factors regarding vacant land conversion to GI in Jinan.

Seismic velocity model building using neural networks: Training data design and learning generalization

Data-driven artificial neural networks (ANNs) offer several advantages over conventional deterministic methods in a wide range of geophysical problems. For seismic velocity model building, judiciously trained ANNs offer the possibility of estimating high-resolution subsurface velocity models. However, a significant challenge of ANNs is training generalization, which is the ability of an ANN to apply the learning from the training process to test data not previously encountered. In the context of velocity model building, this means learning the relationship between velocity models and the corresponding seismic data from a set of training data, and then using acquired seismic data to accurately estimate unknown velocity models. We have asked the following question: What types of velocity model structures need to be included in the training process so that the trained ANN can invert seismic data from a different (hypothetical) geologic setting? To address this question, we create four sets of training models: geologically inspired and purely geometric, with and without background velocity gradients. We find that using geologically inspired training data produces models with well-delineated layer interfaces and fewer intralayer velocity variations. The absence of a certain geologic structure in training models, however, hinders the ANN’s ability to recover it in the testing data. We use purely geometric training models consisting of square blocks of varying size to demonstrate the ability of ANNs to recover reasonable approximations of flat, dipping, and curved interfaces. However, the predicted models suffer from intralayer velocity variations and nonphysical artifacts. Overall, the results successfully determine the use of ANNs in recovering accurate velocity model estimates and highlight the possibility of using such an approach for the generalized seismic velocity inversion problem.

Data-driven prediction of flame temperature and pollutant emission in distributed combustion

The flame temperature and pollutant emission (of NO and CO) characteristics in distributed combustion were examined using data-driven artificial neural network (ANN) approach. Experimental results collected from swirl-assisted distributed combustion using methane fuel at an equivalence ratio 0.9 were utilized for dataset preparation. The distributed combustion condition was created in a swirl-assisted burner (at thermal intensity of 5.72 MW/m3-atm) by diluting the main airstream with carbon dioxide. Experimental results of exhaust NO and CO concentrations and adiabatic flame temperature (AFT) derived from Chemkin-Pro® simulation were selected as the target output. The ANN model was developed with inlet airflow rates and O2 concentrations as the input, and pollutant emission and AFT as the output variables. The ANN possessed one hidden layer with variable number of neurons (10, 15, and 20). Tangent sigmoid and Log sigmoid transfer functions were tested along with feed forwards and cascade forward network schemes having Levenberg–Marquardt backpropagation training algorithms. Different learning rates of model training were investigated to determine the optimized training model. The best model (with learning rate 0.2) was selected based on the mean square error (MSE) and the strength of correlation (R) predicted for individual outputs. The model predicted very well the target outputs in both conventional swirl combustion and distributed combustion regime for wider applicability in a range of applications. A very strong correlation was predicted by the current ANN model for the overall data as indicated by the R value of 0.9842. Furthermore, the results obtained with Bayesian Regularization backpropagation method demonstrated better prediction than those from Levenberg–Marquardt algorithm.

Hybrid Data-Driven Modeling Methodology for Fast and Accurate Transient Simulation of SiC MOSFETs

To enable fast and accurate models of SiC MOSFETs for transient simulation, a hybrid data-driven modeling methodology of SiC MOSFETs is proposed. Unlike conventional modeling methods that are based on complex nonlinear equations, data-driven artificial neural networks (ANNs) are used in this article. For model accuracy, the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$I$</tex-math></inline-formula> – <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$V$</tex-math></inline-formula> characteristics are measured in the whole operation region to train the ANN. The ANN model is then combined with behavior-based equations to model the cutoff region and to avoid overfitting the ANN. In addition, the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$C$</tex-math></inline-formula> – <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$V$</tex-math></inline-formula> characteristics are modeled by ANNs with a logarithmic scale for accuracy. The proposed model is implemented and simulated in SPICE simulator SIMetrix. The simulation results are compared with experimental results from a double pulse tester to validate the proposed modeling methodology. The model is also compared with the Angelov model created by the Keysight MOSFET modeling software. The comparison results show that the proposed model is more accurate than the Angelov model. Besides, when compared to the Angelov model, the proposed model requires 30% less computation time when simulating a double pulse tester. In addition, the proposed modeling method also has better adaptability to model different types of SiC MOSFETs.