Gaussian Process Regression Model Research Articles

Heteroscedastic Gaussian Process Regression for material structure–property relationship modeling

Uncertainty quantification is a critical aspect of machine learning models for material property predictions. Gaussian Process Regression (GPR) is a popular technique for capturing uncertainties, but most existing models assume homoscedastic aleatoric uncertainty (noise), which may not adequately represent the heteroscedastic behavior observed in real-world datasets. Heteroscedasticity arises from various factors, such as measurement errors and inherent variability in material properties. Ignoring heteroscedasticity can lead to lower model performance, biased uncertainty estimates, and inaccurate predictions. Existing Heteroscedastic Gaussian Process Regression (HGPR) models often employ complicated structures to capture input-dependent noise but may lack interpretability. In this paper, we propose an HGPR approach that combines GPR with polynomial regression-based noise modeling to capture and quantify uncertainties in material property predictions while providing interpretable noise models. We demonstrate the effectiveness of our approach on both synthetic and physics-based simulation datasets, including mechanical properties (effective stress) of porous materials. We also introduce an approximated expected log predictive density method for model selection, which eliminates the need for retraining the model during leave-one-out cross-validation, allowing for efficient hyperparameter tuning and model evaluation. By capturing heteroscedastic behavior, enhancing interpretability, and improving model selection, our approach contributes to the development of more robust and reliable machine learning models for material property predictions, enabling informed decision-making in material design and optimization.

Prediction of Remaining Useful Life for Lithium‐Ion Batteries Using Complete Ensemble Empirical Mode Decomposition with Adaptive Noise for Feature Analysis, and Bidirectional Long Short‐Term Memory Coupled with a Gaussian Process Regression Model

Accurately predicting the remaining useful life (RUL) of lithium‐ion batteries is a challenging task, with significant implications for managing battery usage risks and ensuring equipment stability. However, the phenomenon of capacity regeneration and the lack of confidence interval expression result in imprecise predictions. To tackle these challenges, this article proposes a novel method for predicting RUL by optimizing health features (HFs) and integrating multiple models. First, multiple HFs are collected from the charging curves, and the fusion HF is optimized by kernel principal component analysis. To eliminate local fluctuations caused by capacity regeneration effects, the complete ensemble empirical mode decomposition with adaptive noise is employed to decompose the fusion HF. Second, to address the issue of lacking confidence interval expression, a hybrid model is proposed by integrating bidirectional long short‐term memory neural network with Gaussian process regression for effectively capturing the lithium‐ion battery capacity‐declining trend and accurately predicting the RUL. Finally, the proposed model's effectiveness is validated by comparing it with several other models using National Aeronautics and Space Administration and Center for Advanced Life Cycle Engineering datasets. The results indicate that this model achieves a root mean square error of 0.0023 and a mean absolute error of 0.0058, demonstrating significant improvements in predictive accuracy for RUL with high reliability.

Combination of machine learning and COSMO-RS thermodynamic model in predicting solubility parameters of coformers in production of cocrystals for enhanced drug solubility

In this study, we develop predictive models for three target variables, denoted as δd, δp, and δh using a dataset with 86 features and 181 samples. The response parameters, which are Hansen solubility parameters, were correlated to input parameters via several machine learning techniques. The input features are molecular descriptors of coformers which are calculated based on COMSO-RS thermodynamic model and group contribution approach. The analysis includes outlier detection via Cook's distance, normalization with a min-max scaler, and feature selection through L1-based methods. Three regression models—Gaussian Process Regression (GPR), Passive Aggressive Regression (PAR), and Polynomial Regression (PR)—are employed, with hyperparameter optimization achieved using Transient Search Optimization (TSO). The results indicate that for δd, the PAR model outperforms others with an R2 score of 0.885, RMSE of 0.607, MAE of 0.524, and a maximum error of 1.294. The GPR model shows slightly lower performance with an R2 of 0.872, RMSE of 0.816, MAE of 0.579, and a maximum error of 2.755 for δd. The PR model performs on δd with an R2 of 0.814, RMSE of 0.923, MAE of 0.597, and a maximum error of 2.814. For δp, the GPR model provides the best performance, achieving an R2 score of 0.821, RMSE of 1.693, MAE of 1.391, and a maximum error of 3.457. The PAR model performs on δp with an R2 of 0.740, RMSE of 2.025, MAE of 1.980, and a maximum error of 6.609. Also, The PR model predicts δp with a R2 of 0.7, RMSE of 2.329, MAE of 2.02, and maximum error of 6.366. Similarly, for δh, the GPR model again shows superior performance with an R2 score of 0.983, RMSE of 1.243, MAE of 1.005, and a maximum error of 2.577. The PAR model also accurately predicts δh with a R2 of 0.924, RMSE of 2.713, MAE of 2.416, and maximum error of 6.307. Additionally, the PR model predicts δh with a R2 of 0.927, RMSE of 2.757, MAE of 2.334, and maximum error of 8.064. These results highlight the efficacy of the chosen models and optimization techniques in accurately predicting the specified outputs, demonstrating significant potential for application in relevant predictive modeling tasks.

Evaluation of machine tool dynamics in dry gear hobbing considering time-varying thermal effects

Effective stability analysis strongly depends on accurately predicting machine tool dynamics. However, the nonlinear thermal effects during the machining operations present a significant challenge to accurately modeling the machine tool dynamics. To this end, the evolution of the machine tool dynamics under the thermal growth during the dry gear hobbing process is investigated, and a rapid evaluation model of the temperature-dependent dynamics is developed. Firstly, a preliminary test was conducted to explore thermal effects on the machine tool dynamics. Then, the free vibrations at the spindle nose during the hobbing operations were filtered, and the corresponding modal parameters were estimated. Subsequently, the influence of temperature rise on the modal parameters is analyzed. The results show that the first- and third-order natural frequencies increase by 142 and 95 Hz, respectively, and the damping ratios for all modes are reduced with the machine tool temperature rise. Finally, the relationship between the feature temperatures and modal parameters is described using the multi-output Gaussian Process regression (MOGPR) model. The overall prediction accuracy is 95.6 %, and the relative errors in the test data set are within 10 %. The findings can provide an essential basis for the machining stability analysis of dry gear hobbing.

Predicting the Compression Capacity of Screw Piles in Sand Using Machine Learning Trained on Finite Element Analysis

Screw piles (often referred to as helical piles) are widely used to resist axial and lateral loads as deep foundations. Multi-helix piles experience complex interactions between the plates which depend on the soil properties, pile stiffness, helix diameter, and the number of helix plates among other factors. Design methods for these piles are typically highly empirical and there remains significant uncertainty around calculating the compression capacity. In this study, a database of 1667 3D finite element analyses was developed to better understand the effect of different inputs on the compression capacity of screw piles in clean sands. Following development of the numerical database, various machine learning methods such as linear regression, neural networks, support vector machines, and Gaussian process regression (GPR) models were trained and tested on the database in order to develop a prediction tool for the pile compression capacity. GPR models, trained on the numerical data, provided excellent predictions of the screw pile compression capacity. The test dataset root mean square error (RMSE) of 29 kN from the GPR model was almost an order of magnitude better than the RMSE of 225 kN from a traditional theoretical approach, highlighting the potential of machine learning methods for predicting the compression capacity of screw piles in homogenous sands.

Reliability‐based state parameter liquefaction probability prediction using soft computing techniques

The state parameter (ѱ) accounts for both relative density and effective stress, which influence the cyclic stress or liquefaction characteristic of the soil significantly. This study presents a ѱ‐based probabilistic liquefaction evaluation method using six soft computing (SC) techniques. The liquefaction probability of failure (PL) is calculated using the first‐order second moment (FOSM) method based on the cone penetration test (CPT) database. Then, six SC techniques, such as Gaussian process regression (GPR), relevance vector machine (RVM), functional network (FN), genetic programming (GP), minimax probability machine regression (MPMR) and multivariate adaptive regression splines (MARS), are used to predict PL. The performance of these models is examined using nine statistical indices. Additionally, plots such as regression plots, Taylor diagrams, error matrix and rank analysis are shown to assess the SC model's performance. Finally, sensitivity analysis is performed using the cosine amplitude method (CAM) to assess the influence of input parameters on output. The current study demonstrates that SC models based on state parameter predict PL effectively. RVM and MPMR models closely follow the GPR model in terms of performance, which is superior to the other models. Notably, two equations are generated using GP and MARS models to predict PL. The results of the sensitivity analysis reveal the magnitude of earthquake (Mw) as the most sensitive parameter. The outcomes of this research will offer risk evaluations for geotechnical engineering designs and expand the use of state parameter‐based SC models in liquefaction analysis.

Battery state-of-health estimation incorporating model uncertainty based on Bayesian model averaging

Accurately estimating the state-of-health (SOH) of lithium-ion batteries is crucial for efficient, reliable, and safe use. However, the degradation of these batteries involves complex and intricate failure mechanisms that cannot be fully captured by a single model. To tackle this challenge, we propose an SOH estimation method based on Bayesian Model Averaging (BMA), which effectively accounts for both parameter and model uncertainties by combining estimations from different model implementations. It provides both point-value estimates and the probability distribution estimates, delivering results in a fraction of a second. Evaluated on three open-source datasets, the maximum absolute error of SOH estimation is within 0.03, and the Continuous Ranked Probability Score is within 0.015. Compared to optimal individual models, the proposed BMA method reduces the prediction error of point estimates by half to two-thirds. Additionally, the prediction error for probability estimation decreases by an order of magnitude. Moreover, the comparison studies with respect to the Gaussian Process Regression model and the Quantile Regression Forests model demonstrate the applicability and superiority of proposed method. These results highlight the potential of the BMA method to advance battery SOH estimation and facilitate reliable battery management.

Lithium-ion battery state of health estimation using a hybrid model with electrochemical impedance spectroscopy

The State of Health (SOH) of lithium-ion batteries is crucial for maintaining their safety and reliability. Electrochemical impedance spectroscopy (EIS) provides extensive information about the battery state, but the data across high-, medium-, and low-frequency ranges are not independent. Practical challenges in obtaining full-frequency EIS data include long measurement times, low accuracy due to noise, and high costs. This paper proposes an SOH estimation method based on offline EIS and domain-adversarial neural network (DaNN) transfer. Initially, we use a feature search strategy on experimental EIS data to identify the optimal impedance feature subset through the root mean square error (RMSE) and comprehensive indicators. These features are then input into the DaNN for SOH estimation using transfer algorithms. Finally, based on feature correlation analysis, weights are allocated using a Gaussian process regression model for a secondary prediction, optimizing the SOH results. Experiments on laboratory calendar aging and publicly available cycling aging datasets demonstrate the method's effectiveness, achieving an RMSE of <1 %. The proposed method significantly improves SOH estimation accuracy and efficiency, addressing the limitations of traditional methods and reducing dependency on extensive measurement data. This advancement enhances the safety, reliability, and cost-effectiveness of battery applications.

Advancing nanomedicine production via green method: Modeling and simulation of pharmaceutical solubility at different temperatures and pressures

This study employs machine learning algorithms to model the mole fraction of paracetamol drug and the density of supercritical CO2 as a solvent. The tree-based models were employed to evaluate the impact of temperature and pressure on the fluctuations in drug solubility. The dataset utilized in this study has two input variables: temperature and pressure. Conversely, two factors were regarded as outputs: the solubility of drug and the density of solvent. To model the mole fraction and density, Gaussian Process Regression (GPR), Decision Tree (DT), and Linear Regression (LR) models were trained by some collected experimental data. These models were boosted with AdaBoost algorithm to enhance performance. The Firework Algorithm (FWA) was used to optimize the models’ hyper-parameters. All models performed well in predicting both outputs, but outcomes demonstrated that the ADA-GPR model, with R2 values of 0.995 and 0.996, respectively, achieved the highest accuracy for predicting both mole fraction and density. Also, the Explained Variance Scores of this model are 0.995 and 0.998 for mole fraction and density, respectively.

Evaluation of the probiotic, anti-microbial, anti-biofilm, and safety properties of Levilactobacillus brevis Lb13H

In this research, the probiotic potential, anti-microbial, anti-biofilm, and safety properties of Levilactobacillus brevis Lb13H were investigated. Also, Gaussian Process Regression (GPR) model was applied to predict some of experimental parameters. The strain demonstrated resistance to acidic pH, gastric and intestinal juices, and bile salts. Additionally, the strain exhibited 45.34% surface hydrophobicity, 36.55% auto-aggregation capacity, 26.30% co-aggregation, 10.20% adhesion to Caco-2 cells, 42.57% cholesterol removal, and 48.63% radical-scavenging properties. The probiotic strain was able to inhibit the adhesion of Listeria monocytogenes to intestinal cells by 31.30%, and its competitive inhibition percentage was 37.5%. Furthermore, the replacement of L. monocytogenes with the strain was measured at 22.10%. The most significant antimicrobial effect was observed against L. monocytogenes, while the least effect was noted against Rhizopus stolonifer. Haemolytic activity, DNase production, and biogenic amine production were not observed in the strain, which was also found to be sensitive to the antibiotics ciprofloxacin and chloramphenicol. The cell-free supernatant of L. brevis Lb13H effectively inhibited and degraded biofilm formation by L. monocytogenes, achieving effectiveness rates of 40.80% and 36.60% respectively. The results of modeling indicated that there is not any difference between actual and predicted data and GPR can predict the variables with high accuracy.

Combining evolutionary computation with machine learning technique for improved short-term prediction of UT1-UTC and length-of-day

Over the years, prediction techniques for the highly variable angular velocity of the Earth represented by Earth's rotation (UT1-UTC) and length-of-day (LOD) have been continuously improved. This is because many applications like navigation, astronomy, space exploration, climate studies, timekeeping, disaster monitoring, and geodynamic studies, all rely on predictions of these Earth rotation parameters. They provide early warning of changes in the Earth's rotation, allowing various industries and scientific fields to operate more precisely and efficiently. Thus, in our study, we focused on short-term prediction for UT1-UTC (dUT1) and LOD. Our prediction approach is to combine machine learning (ML) technique with efficient evolutionary computation (EC) algorithms to achieve reliable and improved predictions. Gaussian process regression (GPR) is used as the ML technique with genetic algorithm (GA) as the EC algorithm. GA is used for hyperparameter optimization of GPR model as selecting appropriate values for hyperparameter are essential to ensure that the prediction model can accurately capture the underlying patterns in the data. We conducted some experiments with our prediction approach to thoroughly test its capabilities. Moreover, two forecasting strategies were used to assess the performance in both hindcast and operational settings. In most of the experiments, the data used are the multi-technique combinations (C04) generated by International Earth Rotation and Reference Systems Service (IERS). In one of the experiments, we also investigated the performance of our prediction model on dUT1 and LOD from four different products obtained from IERS EOP 20 C04, DTRF20, JTRF20 and USNO. The prediction products are evaluated with real estimates of the EOP product with which the model is trained. The combined excitations of the atmosphere, oceans, hydrology, and sea level (AAM + OAM + HAM + SLAM) are used as predictors because they are highly correlated to the input data. The results depict the highest performance of 0.412 ms in dUT1 and 0.092 ms/day in LOD, on day 10 of predictions. It is worth noting that the later predictions were obtained by incorporating the uncertainty of the input data as weights in the prediction model, which was a novel approach tested in this study.Graphical

Broadband high gain performance MIMO antenna array for 5 G mm-wave applications-based gain prediction using machine learning approach

This paper presents the findings about implementing a machine learning (ML) technique to optimize the performance of 5 G mm wave applications utilizing multiple-input multiple-output (MIMO) antennas operating at the 28 GHz frequency band. This article examines various methodologies, including simulation, measurement, and the utilization of an RLC-equivalent circuit model, to evaluate the appropriateness of an antenna for its intended applications. In addition to its compact dimensions, the proposed design exhibits a maximum gain of 10.34 dBi, superior isolation exceeding 26 dB, and a broad bandwidth of 16.56 % Centered at 28 GHz and spanning from 25.905 to 30.544 GHz. Another supervised regression machine learning technique is utilized to predict the antenna's gain accurately. Machine learning (ML) models can be assessed by several measures, such as the variance score, R square, mean square error (MSE), mean absolute error (MAE), root mean square error (RMSE), and Mean Absolute Percentage Error (MAPE). Among the six machine learning models considered, it is seen that the Gaussian Process Regression (GPR) model exhibits the lowest error and achieves the highest level of accuracy in forecasting gain. The antenna under consideration has promising qualities for its intended use in high-band 5 G applications. This is evidenced by the modelling findings obtained from Computer Simulation Technology (CST) and Advanced Design System (ADS)and the measured and projected results derived using machine learning methodologies.

Gaussian process regression model based on stacking algorithm and application to stock price prediction

Gaussian process regression model based on stacking algorithm and application to stock price prediction

Gaussian process regression to predict dryout incipience quality of saturated flow boiling in mini/micro-channels

In the present study, a Gaussian process regression (GPR) model is developed to predict the dryout incipience quality for flow boiling in mini / micro – channels based on a consolidated database obtained from Purdue University Boiling and Two – Phase Flow Laboratory (PU – BTPFL) consisting of 997 points from 26 sources. The database includes 13 different working fluids over a wide range of operating conditions: hydraulic diameter (0.51 – 6.0 mm), mass velocities (29 – 2303 kg/ m2 s), liquid – only Reynolds number (125 – 53,770), boiling number (0.31 – 44.3 × 10-4), and reduced pressure (0.005 – 0.78). The inputs to the model were liquid – only Weber number (Wefo), reduced pressure (PR), boiling number (Bo), heated to frictional perimeter ratio(PH/PF), capillary number (Ca) and density ratio (ρg/ρf), and the output was dryout incipience quality. The database was randomly divided into training data to learn GPR kernel (covariance) parameters using maximum likelihood estimation, and test data to evaluate the prediction accuracy of the outputs based on mean absolute error (MAE). Six – different types of covariance functions were tested, and GPR model with automatic relevance detection (ARD) rational quadratic covariance function showed the best overall performance. A performance comparison of approved GPR model was made with an existing highly reliable universal correlation to predict dryout incipience quality for mini / micro – channels. Results show that the developed GPR model exhibits superior generalization ability with an overall MAE of 6.03 %, and a significant reduction of 51.76 % in MAE compared with the universal correlation. Overall, the GPR model was found to be a data efficient machine learning technique for predicting dryout incipience quality for flow boiling in mini / micro – channels based on a consolidated database.

Design of Experiment Is All You Need: Utilizing AI for High-Entropy Alloy Water Splitting Catalyst

High-entropy alloy (HEA) is a new group of material that have excited both fields of electrochemistry and computational chemistry, based on its intriguing concept and outstanding catalytic activity. Nevertheless, there exists a clear limitation to the 'trial-and-error' based research: its compositional space is too large. Therefore, AI-based high-throughput research is considered as an efficient way to optimize the composition of HEA electrocatalysts. To fully utilize AI's potential, the quality of data and rational Design of Experiment (DoE) is of importance. In this regard, we have 1) developed an automated one-step synthesis method of HEA with liquid handler and 2) made machine learning model with Gaussian process (GP) regression to minimize the overpotential (fig. 1a).DoE helps developing a new experiment by categorizing the variables and steps that are necessary. First step of DoE is to start with listing the variables. For the one-step synthesis of HEAs, independent variable was the composition of each elements, where the control variables were:A. Solvent (Amine/ethanol)B. Addition of CNT (O/X)C. Reaction temperature (high/low)D. Reaction time (long/short)E. Concentration of solution (5 mM / 10 mM)Each of control variables have two options and the selection is represented by writing an alphabet. For example, if amine (A) and high reaction temperature (C) condition was selected while the second option is used for the rest, it is written as AC. Based on confounding method theory, by conducting AC experiment, we also get information about BDE experiment by assuming a negative correlation to each other. It is written as: AC = -BDE. Similarly, an equivalent combinatorial list can be made for every possible combination (=25) , while the number of experiment is reduced from 32 to 16. Among these combinations, ABCE was found to be optimal synthesis condition (fig.1b).Afterward, we searched the optimal composition of HEAs, starting with grid searching the six-dimensional compositional field. MnFeCoNiCuMoPdPt - octonary alloy was the targeted alloy where the compositions of noble metal were fixed. Binomial distribution design was used to make the distance between neighboring points in 6-dimension compositional space equal (fig. 1c-d). 8C3 = 56 combinations were selected and consequently, the corresponding amounts of precursor solutions were transferred to 56 heat-inert vials and underwent heat treatment of 200 °C. Overpotential of 56 samples were measured and simple GP regression model was constructed based on these datasets.GP regression is composed of two parts: exploration and exploitation. During exploration, the optimal combination is recommended by searching the point that has the largest deviation. Interestingly, by comparing the Shapley Additive explanations (SHAP) value of each elements, it is clearly proved that Cu was harmful and Ni was beneficial to water splitting performance. Thus, Cu was screened out during the end of the exploration step. After finishing exploration, the next combination was recommended by searching near the point with least overpotential (exploitation step). By using GP regression, the optimal composition among HEAs was discovered in less than 200 experiments. It showed the overpotential less than 30mV for HER and less than 250mV for OER, making the overall water-splitting potential less than 1.5 V. In summary, a GP regression model showed synergistic effect with DoE, and HEA with the optimal composition showed an outstanding water splitting performance, reducing overall overpotential by 50% compared to the first step. Figure 1

Low-frequency chatter suppression for robotic milling using a novel MRF absorber

Robotic milling has a unique advantage for large and complex parts. However, it is extremely prone to low-frequency chatter due to the robot structure mode. In robotic milling, low-frequency chatter has a huge impact on machining quality and efficiency. In this paper, a novel MRF (Magnetorheological fluid) absorber is used to suppress low-frequency chatter during robotic milling based on the proposed low-frequency chatter suppression strategy. First, the GPR (Gaussian process regression) model is used to predict the impulse response function based on a few measured impulse response functions. Next, a cutting force model is developed considering the tool runout and the robot structure mode. The parameters and coefficients of the cutting force model are optimized by the measured cutting forces. Then, the low-frequency chatter frequency is predicted considering the modal coupling effect based on the predicted impulse response function and the simulated cutting force. Finally, an MRF absorber is designed based on the chatter frequency range. The controlled storage modulus of the MRF in the pre-yield region enables a wider frequency shift characteristic of the designed MRF absorber. The MRF absorber controls the input current based on the predicted low-frequency chatter frequency to achieve chatter suppression during robotic milling. Robotic milling experiments verified that the MRF absorber can effectively suppress low-frequency chatter. The experimental results show that the Energy Ratio Chatter Indicator of the industrial robot with MRF absorber is less than 0.1 under different milling conditions.

Kernel-based framework for improved prediction of discharge coefficient in vertically supported cylindrical weirs

ABSTRACT The present study represents the first use of kernel-based models to predict discharge coefficient (Cd) for two distinct types of cylindrical weirs, featuring vertical support and a 30-degree upstream ramp. For this purpose, kernel-based methods, including support vector machine, Gaussian process regression (GPR), Kernel extreme learning machine, and Kernel ridge regression, were used, as they offer notable advantages compared to other machine learning models, such as flexibility in handling various data patterns, robustness against overfitting, and effectiveness in high-dimensional data scenarios. The results indicated that the GPR model, with statistical metrics of R = 0.967, Nash–Sutcliffe efficiency (NSE) = 0.935, and root-mean-square error (RMSE) = 0.027, demonstrates superior accuracy in modeling the overall dataset collected from two distinct types of weirs. Through a conducted sensitivity analysis, it was identified that the upstream Froude number is pivotal in accurately predicting the Cd of a cylindrical weir. The modeling conducted for two distinct weir types revealed that a cylindrical weir with vertical support exhibits enhanced predictive capabilities (R = 0.997, NSE = 0.994, and RMSE = 0.007) for Cd. The findings indicate that the introduction of the upstream ramp alters hydraulic conditions, resulting in reduced modeling accuracy (R = 0.760, NSE = 0.529, and RMSE = 0.060).

A novel methodology for gamma-ray spectra dataset procurement over varying standoff distances and source activities

The adoption of machine learning approaches for gamma-ray spectroscopy has received considerable attention in the literature. Many studies have investigated the deployment of various algorithm architectures to a specific task. However, little attention has been afforded to the development of the datasets leveraged to train the models. Such training datasets typically span a set of environmental or detector parameters to encompass a problem space of interest to a user. Variations in these measurement parameters will also induce fluctuations in the detector response, including expected pile-up and ground scatter effects. Fundamental to this work is the understanding that 1) the underlying spectral shape varies as the measurement parameters change and 2) the statistical uncertainties associated with two spectra impact their level of similarity. While previous studies attribute some arbitrary discretization to the measurement parameters for the generation of their synthetic training data, this work introduces a principled methodology for efficient spectral-based discretization of a problem space. A signal-to-noise ratio (SNR) respective spectral comparison measure and a Gaussian Process Regression (GPR) model are used to predict the spectral similarity across a range of measurement parameters. This innovative approach effectively showcased its capability by dividing a problem space, ranging from 5 cm to 100 cm standoff distances and 5 μCi–100 μCi of 137Cs, into three unique combinations of measurement parameters. The findings from this work will aid in creating more robust datasets, which incorporate many possible measurement scenarios, reduce the number of required experimental test set measurements, and possibly enable experimental training data collection for gamma-ray spectroscopy.

Learning active flow control strategies of a swept wing by intelligent wind tunnel

An intelligent wind tunnel using an active learning approach automates flow control experiments to discover the aerodynamic impact of sweeping jets on a swept wing. A Gaussian process regression model is established to study the jet actuator’s performance at various attack and flap deflection angles. By selectively focusing on the most informative experiments, the proposed framework was able to predict 3,721 wing conditions from just 55 experiments, significantly reducing the number of experiments required and leading to faster and cost-effective predictions. The results show that the angle of attack and flap deflection angle are coupled to affect the effectiveness of the sweeping jet. Meanwhile, increasing the jet momentum coefficient can contribute to lift enhancement; a momentum coefficient of 3% can increase the lift coefficient by at most 0.28. Additionally, the improvement effects are more pronounced when actuators are placed closer to the wing root.

Efficient predictive modeling of resilient modulus in stabilized clayey soil using automated machine learning

Pavement subgrade design relies on the resilient modulus (Mr) to analyze structural response to vehicle-like loading. Adding stabilizers to the soil subgrade makes estimating Mr difficult and resource-intensive. This study uses an automated machine learning (ML) strategy to predict the Mr of stabilized clayey soil using recycled plastic waste. The proposed method automates model selection and hyperparameter tuning, making it a feasible alternative to tedious ML modeling and costly laboratory testing. From extensive laboratory investigation involving 3285 experimental data points, the automated ML model using Bayesian optimization evaluates ensembles, support vector machine (SVM), neural network (NET), decision trees, (TREE), and Gaussian process (GP) regression models, identifying the best model based on cross-validation mean squared error (MSE). Bayesian optimization explores hyperparameter spaces to find optimal configurations, enhancing the accuracy, scalability, and reliability of the prediction model. The optimization process yielded the best results for the ensemble least square boost (LSBoost) model with a cross-validation mean squared error (MSE) value of 6.723×10−29. The optimized ML model’s performance is measured using R2 and adjusted R2. The LSboost model’s R2 and adjusted R2 values of 0.9999 suggested overfitting, prompting further investigations using performance metrics like root mean squared error (RMSE), mean absolute error (MSE), and probability density function (PDF) for normalized absolute error (NAE) for training and testing datasets for the predictive ML model. The small RMSE and MAE (0.0049 and 0.0005) values and symmetrical NAE distribution of the proposed ML model demonstrate its high accuracy and generalization capabilities. The proposed model was subsequently tested on the new, unseen data and achieved predictions with an error rate of 0.24 %. This confirms the proposed ML model’s superiority over conventional deformation models, making it ideal for reliable geotechnical engineering applications.