Root Mean Square Error Values Research Articles

Multiple faults detection in doubly-fed induction generator wind turbine using artificial neural network

The development of fault detection methods in wind turbine (WT), especially for single fault detection, is continuously increasing. However, the rapid growth of fault detection in WT leads to another challenge where multiple faults can occur. The single fault detection method in WT is no longer reliable, especially when multiple faults occur simultaneously. Therefore, multiple faults detection in doubly-fed induction generators (DFIG) WT was proposed using an artificial neural networks (ANN) model. These multiple faults include internal and external stator faults happening simultaneously. Internal stator faults cover inter-turn short circuit faults and open circuit faults, while external stator faults cover loss of excitation and external short circuit faults. The performance of the developed multiple faults detection model was measured using accuracy and the root mean square error (RMSE) value. The results show that the developed model performs well with high accuracy and a low RMSE value. Thus, the developed model can accurately detect the coexistence of multiple faults in DFIG WT.

Precision forecasting of fertilizer components’ concentrations in mixed variable-rate fertigation through machine learning

Accurate monitoring of fertilizer concentration and variable irrigation components is crucial for achieving precision irrigation through variable-rate fertigation. However, technological and cost constraints pose challenges in effectively managing mixed variable-rate fertigation. This study addresses these challenges by integrating machine learning (ML) with easily monitored physical parameters, such as electrical conductivity (EC), pH, and temperature, to predict fertilizer solution components and concentrations in mixed variable-rate fertigation. Cubic spline interpolation (CSI) was used to enhance the training dataset. Six ML algorithms—Multivariate Linear Regression (MLR), Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Extremely Randomized Trees (ETs), Multilayer Perceptron (MLP), and Extreme Gradient Boosting (XGB)—were used to develop prediction models, with their performance evaluated using the coefficient of determination (R2), root mean square error (RMSE), and Akaike information criterion (AIC). The Extended Fourier Amplitude Sensitivity Test (EFAST) assessed the sensitivity of the ML models to physical parameters. All of the ML models, except for MLR, particularly SVM, demonstrated superior performance in predicting fertilizer solution components’ concentrations, with R2 values between 0.989 and 0.997 and RMSE values between 0.089 and 0.210. The CSI significantly enhanced model performance, resulting in larger R2 values and smaller RMSE values. The AIC results and sensitivity analysis confirmed the exceptional performance of SVM, emphasizing its suitability for predicting fertilizer components using easily measured physical parameters. The developed ML models offer valuable insights for decision-makers managing irrigation and fertilization in mixed variable-rate fertigation.

Stock Market Prediction

The prediction of stock market trends is a challenging yet critical task in the financial sector, given its significant implications for investors, traders, and financial institutions. This research leverages the Long Short-Term Memory (LSTM) algorithm, a type of recurrent neural network (RNN), to develop a robust model for forecasting stock prices. The study utilizes historical stock market data sourced from Yahoo Finance, accessed via the yfinance package in Python. The primary objectives are to preprocess the data, implement the LSTM model, and evaluate its performance against traditional models such as Random Forest and Linear Regression. Data preprocessing involved handling missing values, normalizing the dataset, and transforming it into sequences suitable for LSTM training. The model's architecture includes multiple LSTM layers designed to capture temporal dependencies in the data. The study evaluates the model's performance using metrics such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and prediction accuracy. Comparative analysis shows that the LSTM model outperforms both Random Forest and Linear Regression models, with lower MSE and RMSE values and higher accuracy in predicting stock prices. This research discovered that LSTM's ability to retain long-term dependencies makes it particularly effective for stock market prediction, where historical trends and patterns significantly influence future prices. The results indicate that the LSTM model provides more reliable and precise predictions, which can enhance decision-making in trading and investment. This research highlights the potential of advanced neural network architectures in financial forecasting, offering a valuable tool for investors aiming to optimize their strategies and mitigate risks. The significance of this study lies in its practical application in the financial industry, demonstrating that machine learning models, particularly LSTM, can substantially improve the accuracy of stock market predictions. Future research could explore the integration of additional features, such as macroeconomic indicators and sentiment analysis, to further enhance model performance. This study underscores the importance of continuous innovation and the adoption of sophisticated algorithms to navigate the complexities of financial markets.

NDVI estimation using Sentinel-1 data over wheat fields in a semiarid Mediterranean region

ABSTRACT Annual crop monitoring is a key parameter for managing agricultural strategies. Several studies have relied on remote sensing products such as the normalized difference vegetation index (NDVI) as a vegetation dynamic metric. However, the dependence of optical data on weather conditions limits its availability. In this study, we reconstruct the NDVI time series of wheat fields using the moving averages of the Sentinel-1 normalized VH/VV cross-polarization ratio (IN) and the interferometric coherence in VV polarization over wheat selected fields in a semiarid site in Tunisia during two seasons, from 2018 to 2020. The crop cycle is divided into two periods: before and after the heading phase, which occurs in approximately the middle of March. Due to the volume-scattering impact, the second phase is divided into the ripening and maturation phase (NDVI ≥0.4) and senescence phase (NDVI <0.4). To estimate the NDVI values, different methods are used: curve-fitting equations and machine learning regressors such as the random forest (RF) and the support vector regressor (SVR). Low root mean square error (RMSE) values characterize NDVI estimation during the first period. In the second period, the RMSE values reach 0.06 when the NDVI is lower than 0.4. When the NDVI values exceed 0.4 in the second period, lower accuracy marks the NDVI estimation using the curve-fitting equations as a function of IN or coherence. Relative low accuracy characterizes the regression algorithms’ estimations when NDVI ≥ 0.4 compared to their performance during the aforementioned periods. The proposed approach was tested on different wheat fields. The NDVI estimations are characterized by RMSE values varying between 0.12 and 0.19. The use of RF and SVR outperformed the curve-fitting methods with an RMSE equal to 0.12. The present findings revealed the high accuracy of the proposed approach to estimate the missing values of wheat fields NDVI values during the vegetation development period until heading and the senescence phase. The presence of the mutual effect of the vegetation water content and its volume complicated the NDVI estimation using the C-band data.

Prediction of Optimum Dosage of Coagulant in Water Treatment Plant: A Comparative Study between Artificial Neural Network and Random Forest

Raw water, sourced directly from natural water bodies, is unsuitable for direct consumption due to the presence of various impurities. Therefore, it undergoes treatment at a Water Treatment Plant (WTP) before being supplied to the public. Preliminary treatment involves the removal of floating matter, through screening, while heavier particles settle out by gravity, fine particles remain in suspension, causing turbidity. Effective removal of these suspended particles requires coagulation to form flocs and facilitate the settling. Determining the optimal coagulant dosage is crucial, as both underdoing and overdosing of coagulant can lead to ineffective treatment and increased costs. Conventionally optimum dosage of coagulant is determined by performing jar test. This study focuses on predicting the optimum coagulant dosage using two soft computing techniques: Artificial Neural Network (ANN) and Random Forest (RF). The Input parameters for model development include turbidity, pH, temperature, and alkalinity of raw water from the Parvati Water Treatment Plant, Pune. In this study Four models were developed, namely Model A (Turbidity), Model B (pH, Alkalinity, Temperature, Turbidity), Model C (pH, Alkalinity, Temperature), and Model D (Alkalinity and Turbidity). These models were trained using ANN and RF. Predictions of optimum coagulant doses were made for the testing dataset, and model accuracy was evaluated using Scatter plots, Root Mean Squared Error (RMSE) and Coefficient of Correlation (R). Results indicate that RMSE values of ANN Models are comparatively lower than RF. Comparing among Models A, B, C, and D, Model B and Model D exhibit better performance, with lower RMSE values.

Comparison Accuracy of CHIRPS, GSMaP V7, and GSMaP V8 Satellite Rainfall Estimation in Kalimantan

The application of satellite product rainfall estimates (SPREs) is growing in hydrometeorology due to limited rainfall measurement. This study aims to compare the accuracy of three SPRE, namely Climate Hazards Group InfraRed Precipitation with Station (CHIRPS), Global Satellite Mapping of Precipitation (GSMaP) Moving Vector with Kalman Filtering (GSMaP-MVK), and near-real-time (GSMaP-NRT) versions 7 and 8, against daily and monthly rainfall measurements from eighteen gauges in Kalimantan from December 2021 to May 2023. Continuous validation includes root mean square error (RMSE), relative bias (RB), and correlation coefficient (CC), and categorical validation consists of a probability of detection (POD), false alarm ratio (FAR), and critical success index (CSI) were used to assess the accuracy of SPREs. The results showed that GSMaP-MVK version 8 has the highest accuracy on a daily scale with an RMSE value of 14.31 mm/day, while the CHIRPS has the highest accuracy on a monthly scale with an RMSE of 81 mm/month. GSMaP version 8 is better than GSMaP version 7, with a difference in RMSE, CC, and RB at 14.2%, 9.7%, and 84%. Categorical validation showed that GSMaP version 8 was 2.13%, higher in POD, 3.95% in CSI, and 10.2% in FAR compared to GSMaP version 7. Keywords: Accuracy, CHIRPS, GSMaP, Kalimantan, Rain-gauge.

Enhancing voxel-based dosimetry accuracy with an unsupervised deep learning approach for hybrid medical image registration.

Deformable registration is required to generate a time-integrated activity (TIA) map which is essential for voxel-based dosimetry. The conventional iterative registration algorithm using anatomical images (e.g., computed tomography (CT)) could result in registration errors in functional images (e.g., single photon emission computed tomography (SPECT) or positron emission tomography (PET)). Various deep learning-based registration tools have been proposed, but studies specifically focused on the registration of serial hybrid images were not found. In this study, we introduce CoRX-NET, a novel unsupervised deep learning network designed for deformable registration of hybrid medical images. The CoRX-NET structure is based on the Swin-transformer (ST), allowing for the representation of complex spatial connections in images. Its self-attention mechanism aids in the effective exchange and integration of information across diverse image regions. To augment the amalgamation of SPECT and CT features, cross-stitch layers have been integrated into the network. Two different 177 Lu DOTATATE SPECT/CT datasets were acquired at different medical centers. 22 sets from Seoul National University and 14 sets from Sunway Medical Centre are used for training/internal validation and external validation respectively. The CoRX-NET architecture builds upon the ST, enabling the modeling of intricate spatial relationships within images. To further enhance the fusion of SPECT and CT features, cross-stitch layers have been incorporated within the network. The network takes a pair of SPECT/CT images (e.g., fixed and moving images) and generates a deformed SPECT/CT image. The performance of the network was compared with Elastix and TransMorph using L1 loss and structural similarity index measure (SSIM) of CT, SSIM of normalized SPECT, and local normalized cross correlation (LNCC) of SPECT as metrics. The voxel-wise root mean square errors (RMSE) of TIA were compared among the different methods. The ablation study revealed that cross-stitch layers improved SPECT/CT registration performance. The cross-stitch layers notably enhance SSIM (internal validation: 0.9614vs. 0.9653, external validation: 0.9159vs. 0.9189) and LNCC of normalized SPECT images (internal validation: 0.7512vs. 0.7670, external validation: 0.8027vs. 0.8027). CoRX-NET with the cross-stitch layer achieved superior performance metrics compared to Elastix and TransMorph, except for CT SSIM in the external dataset. When qualitatively analyzed for both internal and external validation cases, CoRX-NET consistently demonstrated superior SPECT registration results. In addition, CoRX-NET accomplished SPECT/CT image registration in less than 6s, whereas Elastix required approximately 50s using the same PC's CPU. When employing CoRX-NET, it was observed that the voxel-wise RMSE values for TIA were approximately 27% lower for the kidney and 33% lower for the tumor, compared to when Elastix was used. This study represents a major advancement in achieving precise SPECT/CT registration using an unsupervised deep learning network. It outperforms conventional methods like Elastix and TransMorph, reducing uncertainties in TIA maps for more accurate dose assessments.

Improved forecasting of the compressive strength of ultra‐high‐performance concrete (UHPC) via the CatBoost model optimized with different algorithms

AbstractThis paper focuses on the applicability of CatBoost models constructed using various optimization techniques for improved forecasting the compressive strength of ultra‐high‐performance concrete (UHPC). Phasor particle swarm optimization (PPSO), dwarf mongoose optimization (DMO), and atom search optimization (ASO), which have been very popular recently, are preferred as optimization algorithms. A comprehensive and reliable data set is used to develop the CatBoost models, which include 785 test results with 15 input features. The performance of the CatBoost models (PPSO‐CatBoost, DMO‐CatBoost, and ASO‐CatBoost) optimized with different algorithms is thoroughly assessed by means of various statistical metrics and error analysis to determine the model with the best forecasting capability, and this model is compared with the models obtained from previous studies. In addition, Shapley additive exPlanations (SHAP) analysis is used to ensure the interpretability of the forecasting models and to overcome the “black box” problem of machine learning (ML) models. The obtained results demonstrate that all CatBoost models outstandingly forecast the compressive strength of UHPC. Among these models, the DMO‐CatBoost model stands out compared to the other models in various performance metrics, such as high coefficient of determination (R2) values, low root mean squared error (RMSE), mean absolute percentage error (MAPE), and mean absolute error (MAE) values, along with a smaller error ratio. In other words, the RMSE, R2, MAPE, and MAE values of the DMO‐CatBoost model for the training set are 3.67, 0.993, 0.019, and 2.35, respectively, whereas those for the test set are 6.15, 0.978, 0.038, and 4.51. Additionally, the performance ranking of the algorithms used to optimize the hyperparameters of the CatBoost model is as follows: DMO &gt; PPSO &gt; ASO. On the other hand, SHAP analysis showed that age, fiber dosage, and cement dosage significantly influence the compressive strength of UHPC. These findings can guide structural engineers in the design and optimization of UHPC, thus assisting them in developing strategies to improve the strength properties of the material. Finally, based on the best forecasting model developed in this work, a graphical user interface has been developed to easily forecast the compressive strength of UHPC in practical applications without additional tools or software.

Predictive Modelling on Competitor Analysis Performance by using Generalised Linear Models and Machine Learning Approach

Competitive analysis in digital and technology is trending in the business field. However, the field of digital and technology in the business world is vast and challenging to analyse. The purpose of this research is first to identify the success factor which represent the company performance. Second, is to identify the significant services provided by the company to their business user. Then, based on the first and second objectives, a predictive modelling is developed to produce the best solution to their business user. The research is implementing a case study from Telecommunication Company and using data science life cycle methodology. The statistical modelling that is used to develop the competitor’s analysis model is generalised linear model (GLM) which integrated with machine learning approach. Furthermore, the synthetic data set is created by using Gamma Distribution, Gaussian Distribution and Poisson Distribution due to some data from the case study is confidential. The synthetic data set is based on existing real data which are from Telecommunication Company sentiment analysis data, were used to investigate the performance of the proposed model. The machine learning technique is used to get the accuracy of the significant GLM which has been developed. The accuracy is tested by using the error rates of the machine learning technique which are Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and R-squared. This research discovered that the business solution to the significant service for the business user and discovered the best statistical model to be used for the business solution. The results show that the Gumbel distribution is the best fit model for the synthetic dataset where the values of RMSE is 1.0574, MAE is 0.9168 and R-squared is 0.3994, and the significant success factor that has been identified by using the GLM is advertising success factor. The model developed can be improved with another type of data set and different sizes of data. Hence, further studies and real-world data are required for better validation.

Computation of Fuzzy Linear Regression Model using Simulation Data

Regression analysis is a powerful technique for determining the causal influence on population outcome, although it is susceptible to outliers. It also oversimplifies the real-world data and problem, as data is rarely linearly separable. This paper analyses the simulation data of a fuzzy linear regression model (FLRM) model, which can aid in minimizing the interference of unwanted information, hence improving the precision of results. The aim of the fuzzy linear regression model (FLRM) is used to determine the best prediction model with the lowest measurement error. The model provides a fundamental mathematical and statistical framework for acknowledging the data imprecision. This study has applied the fuzzy linear regression model (FLRM) to evaluate the data in 15 rows of respondent models. The study implemented measurement error of cross-validation technique which is mean square error (MSE) and root mean square error (RMSE), to enhance data accuracy. Microsoft Excel and Matlab were applied to obtain the result. The simulation result indicates that comparing models with two measurement errors should be used to determine the optimal results. MSE and RMSE value with six parameters of a degree of fitting H-value have been calculated in the study. It concludes that a degree of fitting H-value of 0.0 is proven as a good model with the lowest MSE and RMSE measurement errors.

Spatial Weight Matrix Comparison of SAR-X Model using Casetti Approach

The Spatial Autoregressive Exogenous (SAR-X) model with the Casetti approach is used to describe the influence of location and exogenous variables in the description and prediction of spatial observations, namely, people's habits and behavior towards culture in Java Island. The SAR-X model with the Casetti approach is characterized by a spatial weight matrix that describes the coordinates of the region at each location. The spatial weight matrix is determined outside the model. This study examines the spatial weight matrix determined based on rook contiguity, bishop contiguity, queen contiguity, inverse distance and inverse distance squared, and compares the application of the spatial weight matrix to the SAR-X model with the Casetti approach for the description and prediction of people's habits and behavior towards culture in Java Island. The description and prediction results obtained are measured using the Root Mean Square Error (RMSE) value. The results of data processing show that the best spatial weight matrix in the SAR-X model with the Casetti approach to community habits and behavior in Java Island is the inverse distance squared spatial weight matrix, supported by the calculation of the minimum RMSE value and the coefficient of determination above 60%.

Application of a hybrid algorithm of LSTM and Transformer based on random search optimization for improving rainfall-runoff simulation

Flood forecasting using traditional physical hydrology models requires consideration of multiple complex physical processes including the spatio-temporal distribution of rainfall, the spatial heterogeneity of watershed sub-surface characteristics, and runoff generation and routing behaviours. Data-driven models offer novel solutions to these challenges, though they are hindered by difficulties in hyperparameter selection and a decline in prediction stability as the lead time extends. This study introduces a hybrid model, the RS-LSTM-Transformer, which combines Random Search (RS), Long Short-Term Memory networks (LSTM), and the Transformer architecture. Applied to the typical Jingle watershed in the middle reaches of the Yellow River, this model utilises rainfall and runoff data from basin sites to simulate flood processes, and its outcomes are compared against those from RS-LSTM, RS-Transformer, RS-BP, and RS-MLP models. It was evaluated against RS-LSTM, RS-Transformer, RS-BP, and RS-MLP models using the Nash–Sutcliffe Efficiency Coefficient (NSE), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Bias percentage as metrics. At a 1-h lead time during calibration and validation, the RS-LSTM-Transformer model achieved NSE, RMSE, MAE, and Bias values of 0.970, 14.001m3/s, 5.304m3/s, 0.501% and 0.953, 14.124m3/s, 6.365m3/s, 0.523%, respectively. These results demonstrate the model's superior simulation capabilities and robustness, providing more accurate peak flow forecasts as the lead time increases. The study highlights the RS-LSTM-Transformer model's potential in flood forecasting and the advantages of integrating various data-driven approaches for innovative modelling.

Advancing Stock Market Predictions with Time Series Analysis including LSTM and ARIMA

Predicting stock market prices accurately is a major task for investors and traders seeking to optimize their decision-making processes. This research focuses on the comparative analysis of advanced machine learning (ML) techniques, particularly, the Long Short-Term Memory (LSTM) model and Autoregressive Integrated Moving Average (ARIMA) model for predicting stock market prices. The study enforces thorough data collection and preprocessing to ensure the quality and reliability of the historical stock price data, forming a robust foundation for the predictive models. The core contribution of this paper lies in its systematic and comparative analysis of these two models. A range of performance metrics, including Mean Absolute Error (MAE) and Root Mean Square Error (RMSE), are employed to assess and contrast the predictive accuracy and efficiency of the LSTM and ARIMA models. The research findings indicate that the ARIMA model, contrary to expectations, outperforms the LSTM model in this study, achieving lower RMSE and MAE values. Specifically, the ARIMA model demonstrates a Test RMSE of 4.336 and a Test MAE of 3.45926, indicating its superior predictive accuracy compared to the LSTM model. Furthermore, the study sets its findings against the backdrop of existing literature by comparing the performance of its models with those reported in previous research. This comparison shows better results achieved by our stock market prediction models. By addressing limitations observed in prior studies and demonstrating practical applicability, this research contributes to advancing stock market prediction methodologies, offering valuable insights for investors and traders.

Characterizing the Exponential Profile of W' Recovery Following Partial Depletion.

The aim of this study was to characterize W' recovery kinetics in response to a partial W' depletion. We hypothesized that W' recovery following partial depletion would be better described by a biexponential than by a monoexponential model. Nine healthy men performed a ramp incremental exercise test, three to five constant load trials to determine critical power and W', and ten experimental trials to quantify W' depletion. Each experimental trial consisted of two constant load work bouts (WB1 + WB2) interspersed by a recovery interval. WB1 was designed to evoke a 25% or 75% W' depletion (DEP25% + DEP75%). Subsequently, participants recovered for 30, 60, 120, 300 or 600 s, and then performed WB2 to exhaustion in order to calculate the observed W' recovery (W'OBS). W'OBS data were fitted using monoexponential and biexponential models, both with a variable and a fixed model amplitude. Root mean square error (RMSE) and Akaike information criterion (AICc) were calculated to evaluate the models' goodness-of-fit. The biexponential model fits were associated with overall lower RMSE values (0.4-5.0%) compared to the monoexponential models (2.9-8.0%). However, ΔAICc resulted in negative values (-15.5 and -23.3) for the model fits where the amplitude was free, thereby favoring the use of a monoexponential model for both depletion conditions. For the model fits where the amplitude was fixed at 100%, ΔAICc was negative for DEP25% (-15.0), but positive for DEP75% (11.2). W'OBS values were strongly correlated between both depletion conditions (r = 0.92), and positively associated with V̇O2peak, CP and GET (r = 0.67-0.77). The present study results did not provide evidence in favor of a biexponential modeling technique to characterize W' recovery following partial depletion. Moreover, we demonstrated that fixed t values were insufficient to model W' recovery across different depletion levels, and that W' recovery was positively associated with aerobic fitness. These findings underline the importance of employing variable and individualized t values in future predictive W' models.

Shear strength model of FRP-reinforced concrete beams without shear reinforcement

The current design codes for evaluating the shear strength of concrete beams reinforced with fiber-reinforced polymer (FRP) bars predominantly rely on a semi-empirical approach through test data calibration without a robust theoretical background. This approach may lead to either overestimation or underestimation, resulting in significant scattering in strength predictions. This study proposed a modified approach to enhance the existing unified shear-strength model developed for steel-reinforced concrete (RC) beams based on the compression zone failure mechanism, for predicting the shear strength of FRP-RC beams without shear reinforcement. The proposed model integrates strain and stress approaches, accounting for the distinctive characteristics associated with FRP reinforcement. Additionally, the overall shear resistance capacity of the FRP-RC beam is considered as the sum of contributions from the intact concrete of the compression zone and the aggregate interlock mechanism. The reliability of the proposed model is verified by comparing its predictions with a comprehensive database comprising 264 FRP-RC beam specimens without shear reinforcement reported by 41 research groups. Furthermore, the accuracy of the proposed model was assessed through a comparative analysis with existing evaluation methods, encompassing machine learning-based models, empirical models, theoretical models, and design codes. The results indicate that the proposed model satisfactorily predicts the shear strength of FRP-RC beams within a wide range of design parameters. In comparison with machine learning-based models, the proposed theoretical model exhibits a similar level of accuracy in terms of the mean shear strength ratio and coefficient of determination (R2) values but with less scattering, indicated by lower coefficient of variation (COV) and root mean square error (RMSE) values. In addition, the proposed model outperforms existing empirical and theoretical models, as well as current design codes, by significantly enhancing prediction accuracy, demonstrated by lower COV, RMSE, and higher R2 values. Finally, a parametric study was conducted to investigate the effects of the key parameters on the shear strength of FRP-RC beams using both the proposed model and representative existing evaluation methods.

Tool Wear Prediction Based on LSTM and Deep Residual Network

To improve the accuracy and efficiency of tool wear predictions, this study proposes a tool wear prediction model called LSTM_ResNet which is based on the long short-term memory (LSTM) network and the Residual Network (ResNet). The model utilizes LSTM layers for processing, where the first block and loop blocks serve as the core modules of the deep residual network. The model employs a series of methods including convolution, batch normalization (BN), and Rectified Linear Unit (ReLU) to enhance the model’s expression and prediction capabilities. The performance of the LSTM_ResNet model was evaluated using experimental data from the PHM2010 datasets and two different depths (64 and 128 layers), training both LSTM_ResNet models for 200 epochs. The 64-layer model’s root mean square error (RMSE) values are 3.36, 4.35, and 3.59, and the mean absolute error (MAE) values are 2.42, 2.85, and 2.21; using 128 layers, the RMSE values are 3.66, 3.99, and 3.77, and the MAE values are 2.49, 2.73, and 3.01. The results indicate that the 64-layer LSTM has smaller average errors, suggesting that compared to other common network structures, the LSTM_ResNet network has a higher performance. This research provides an effective solution for tool wear prediction and helps to improve the technical level of tool wear prediction in China.

Experimental and simulated evaluation of inverse model for shallow underground thermal storage

This study employs an inverse grey-box (IGB) modeling approach, which combines measured data and the physics of systems to predict the performance of shallow underground thermal energy storage (UTES) with top and side insulation. A simplified IGB model of the shallow UTES was developed using thermal network analysis. Experimental studies were conducted for shallow vertical and horizontal UTES configurations. Detailed models representing the field experiments were developed using the TRNSYS simulation tool and calibrated with the measured data from the experimental setup. The 4 Resistance 2 Capacitance (4R2C) IGB model was trained and tested in MATLAB using field experimental data and the data from the calibrated TRNSYS model. In comparing experimental data with the TRNSYS model for UTES, the prediction of outlet water temperature showed good agreement, with root mean square error (RMSE) and coefficient of variation of RMSE (CVRMSE) values of 0.94 °C and 3.16 % for vertical, and 0.99 °C and 2.97 % for horizontal configurations. The IGB model also aligned well with the experimental data, showing CVRMSE of 7.91 % and 3.17 % for vertical and horizontal systems, respectively. Sensitivity analysis revealed that model performance improves with longer training durations and closer testing times to training periods. However, a convergence point of 20 weeks of training data was achieved for making long-term performance predictions.

SWAN: A multihead autoregressive attention model for solar wind speed forecasting

High speed streams of solar wind can impact Earth’s magnetic environment. Current state of the art geomagnetic effect forecasting methods can predict perturbations up to two hours in advance by using solar wind measurements made in the L1 Lagrange point. By forecasting these solar wind parameters, it should be possible to extend the time range for which said methods can be used.We present Solar Wind Attention Network (SWAN), an attention-based autoregressive model for the forecasting of daily average bulk speed values. We apply a novel automatic design method to create its architecture by developing a pipeline of hyperparameter optimization techniques as a proof of concept. Usage of this pipeline leads to a robust, reliable architecture which we empirically verify the convergence of in a variety of problem configurations regarding solar wind speed forecasting.An improvement in terms of root mean squared error (RMSE) is found when compared to WindNet and convolutional neural networks (CNNs) expanding it, even when SWAN does not incorporate images of the solar disk into the input. SWAN’s RMSE of 68.79 ± 1.05 km s−1 with a lead time of three days compares to RMSE values in the order of 80.28 ± 3.05, 76.30 ± 1.87 and 71.93 ± 0.40 km s−1. From these results, we discuss the possibility of incorporating both autoregressive and image information as input to enhance predictive capability of newly developed models. Furthermore, we analyze the resulting model’s behavior to characterize its main error sources, and find its performance to be at its best in scenarios not involving newly formed coronal holes or coronal mass ejections. It is found that this behavior matches well with a priori expectations derived from the underlying physics. Finally, we briefly discuss the need for an in-depth study of different evaluation metrics to provide a more thorough understanding of modeling results on the field.

The impact of base design and restoration type on the resin consumption, trueness, and dimensional stability of dental casts additively manufactured from liquid crystal display 3D printers.

To evaluate the effects of two base types and three restoration designs on the resin consumption and trueness of the 3D-printed dental casts. Additionally, the study explored the dimensional stability of these 3D-printed dental casts after 1 year of storage. Various types of reference dental casts were specifically designed to represent three types of dental restoration fabrications, including full-arch (FA), long-span (LS), and single-unit (SU) prostheses. The reference casts were digitized with a dental laboratory scanner and used to create flat and hollow base designs (N = 18) for the 3D-printed study casts. The 3D-printed study casts were digitized and evaluated against their corresponding references immediately after 3D printing and again after 1 year of storage, with the trueness quantified using the root mean square error (RMSE) at both time points. Volumes of resin used were recorded to measure resin consumption, and the weights of the 3D-printed study casts were also measured. The data were analyzed using two-way ANOVA and a Tukey post hoc test, α = 0.05. Volumetric analysis showed the flat-base design had significantly higher resin consumption with weights for the FA group at 42.51 ± 0.16g, the LS group at 31.64 ± 0.07g, and the SU group at 27.67 ± 0.31g, as opposed to 26.22 ± 1.01g, 22.86± 0.93g, and 20.10 ± 0.19g for the hollow designs respectively (p < 0.001). Trueness, assessed through two-way ANOVA, revealed that the flat-base design had lower RMSE values indicating better trueness in the LS (54 ± 6 µm) and SU (59 ± 7 µm) groups compared to the hollow-base design (LS: 73 ± 5, SU: 99 ± 11 µm, both p < 0.001), with no significant difference in the FA group (flat-base: 50 ± 3, hollow: 47 ± 5 µm, p= 0.398). After 1 year, the flat-base design demonstrated superior dimensional stability in the LS (flat base: 56 ± 6 µm, hollow base: 149 ±45 µm, p < 0.001) and SU groups (flat base: 95 ± 8 µm, hollow base: 183 ±27 µm, p < 0.001), with the FA group showing no significant difference in the base design (flat base: 47 ± 9, hollow base: 62 ± 12 µm, p = 0.428). The hollow-base design group showed lower resin consumption than the flat-base design group. However, the flat-base designs exhibited superior trueness and less distortion after 1 year of storage. These findings indicate that despite the higher material usage, flat-base designs provide better initial accuracy and maintain their dimensional stability over time for most groups.

Development and comparison of machine learning models for in-vitro drug permeation prediction from microneedle patch

The field of machine learning (ML) is advancing to a larger extent and finding its applications across numerous fields. ML has the potential to optimize the development process of microneedle patch by predicting the drug release pattern prior to its fabrication and production. The early predictions could not only assist the in-vitro and in-vivo experimentation of drug release but also conserve materials, reduce cost, and save time. In this work, we have used a dataset gleaned from the literature to train and evaluate different ML models, such as stacking regressor, artificial neural network (ANN) model, and voting regressor model. In this study, models were developed to improve prediction accuracy of the in-vitro drug release amount from the hydrogel-type microneedle patch and the in-vitro drug permeation amount through the micropores created by solid microneedles on the skin. We compared the performance of these models using various metrics, including R-squared score (R2 score), root mean squared error (RMSE), and mean absolute error (MAE). Voting regressor model performed better with drug permeation percentage as an outcome feature having RMSE value of 3.24. In comparison, stacking regressor have a RMSE value of 16.54, and ANN model has shown a RMSE value of 14. The value of permeation amount calculated from the predicted percentage is found to be more accurate with RMSE of 654.94 than direct amount prediction, having a RMSE of 669.69. All our models have performed far better than the previously developed model before this research, which had a RMSE of 4447.23. We then optimized voting regressor model’s hyperparameter and cross validated its performance. Furthermore, it was deployed in a webapp using Flask framework, showing a way to develop an application to allow other users to easily predict drug permeation amount from the microneedle patch at a particular time period. This project demonstrates the potential of ML to facilitate the development of microneedle patch and other drug delivery systems.