Machine Learning Approach to Predict Tunnel FET Characteristics Using Random Forest Regression

We propose a physics-based analytical model that accurately captures the effects of buffer traps on dc characteristics of gallium nitride (GaN)-based high-electron-mobility transistors (HEMTs). The model is then semi-analytically extended to additionally include the transient behavior. Analytical formulations for the shift in the threshold voltage <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${(}{V}_{\text {OFF}}{)}$ </tex-math></inline-formula> and two-dimensional electron gas (2-DEG) density due to the presence of buffer traps in the steady state are presented. In pulsed operation, technology computer-aided design (TCAD) simulations indicate that a time-dependent negative potential (NP) is developed under the gate, resulting in a modified <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${V}_{\text {OFF}}$ </tex-math></inline-formula> and current collapse (CC). An expression for the modified <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${V}_{\text {off}}$ </tex-math></inline-formula> helps capture the pulsed current–voltage characteristics. The model captures the dependence of bias, time, temperature, trap concentration, capture cross section area, and activation energy of traps on the steady-state and transient characteristics. The model is implemented in Verilog-A in an existing compact model framework using a diode and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">RC</i> sub-circuit and validated using measured data and TCAD simulations. The modeling results are in excellent agreement with the experimental data and TCAD simulations. Since the model is physics-based, it requires fewer number of parameters compared to that in the existing models.

A machine-learning-based model for predicting the effective stiffness of precast concrete columns

Study of parasitic oscillation of a multi-chip SiC MOSFET circuit based on a signal flow graph model by TCAD simulation

Concerns regarding disinfection byproducts (DBPs) in drinking water persist, with measurements in water treatment plants (WTPs) being relatively easier than those in water distribution systems (WDSs) due to accessibility challenges, especially during adverse weather conditions. Machine learning (ML) models offer improved predictions of DBPs in WDSs. This study developed multiple ML models to predict Trihalomethanes (THMs), Haloacetic Acids (HAAs), Dichloroacetonitrile (DCAN), and N-nitrosodimethylamine (NDMA) in WDSs using data collected over 13years (2008-2020) from 113 water supply systems (WSS) in Ontario. Data were collected tri-monthly (four times/year) following Ontario's regulatory requirements. Four common ML models-linear regressor (LR), random forest regressor (RFR), support vector regressor (SVR), and artificial neural networks with multiple folds cross-validation (ANN-MV) and single fold validation (ANN-SV)-were trained and tested using different datasets. R2 values for training datasets of THMs, HAAs, DCAN, and NDMA models ranged from 0.533 to 0.976, 0.560 to 0.980, 0.602 to 0.993, and 0.449 to 0.858, respectively. For testing datasets, R2 ranged from 0.517 to 0.939, 0.437 to 0.945, 0.565 to 0.973, and 0.517 to 0.718, respectively. Among THMs, HAAs, and DCAN, ANN-SV models were identified as the best, followed by the RFR model, whereas for NDMA, SVR was the superior model, followed by the LR model. Some models reliably predicted DBPs, suggesting they could replace costly sampling and experimental analysis for DBPs in the WDSs, thereby enhancing DBPs control in WDSs and reducing human exposure and associated risks.

Personal identification of an individual has always been a major concern in forensic science. Reconstruction of the facial profile is considered as one of the final stages in the process of identification. Nevertheless, recent advancements in artificial intelligence (AI) and machine learning (ML) have demonstrated remarkable potential in predictive modelling and forensic applications. The current study uses customised machine learning models to predict facial dimensions based on dental and jaw parameters. A sample of 422 participants (201 males and 221 females) from a North Indian population was collected and analysed. Dental casts, anthropometric facial measurements and photographs of the participants were collected with informed consent. ML models such as Support Vector Regression (SVR), Random Forest Regression (RFR), Decision Tree Regression (DTR), and Linear Regression (LR) were trained using dental and jaw measurements as input features for the models. The results show that the ML models predicted the facial dimensions with an accuracy of 90–94% and a very low prediction error of 0.1–0.9 across all facial measurements. Among the models, SVR and LR models perform well, followed by RFR, whereas DFR yielded comparatively lower accuracy. The findings demonstrate that machine learning models (SVR, RFR, DTR, and LR) can be used as novel approach to predict facial dimensions from jaw and teeth parameters. These techniques can be combined with other facial reconstruction techniques to produce more precise and accurate outcomes. The reliability and accuracy in predicting the facial dimensions indicate that the results can be applied in the practical and real situations such as personal identification, forensic investigations, disaster victim identification cases, and archaeological remains where only jaw and teeth are available for examination. Integrating ML-based predictions with traditional facial reconstruction techniques could enhance the accuracy and reliability of forensic identification methodologies.

This work evaluates the potential impact of interface traps on the interface of oxide and semiconductor on a 2-D-based ion-sensitive field-effect transistor (ISFET) using a surface potential-based model and Technology Computer-Aided Design (TCAD) simulation calibrated for pH sensing applications. The electrolyte/oxide interface in the proposed 2-D ISFET model is modeled using the Guoy-Chapman–Stern technique and the site binding association model. The baseline field-effect transistor (FET) of the ISFET is modeled using Fermi-Dirac statistics to obtain surface potential, which is further used to derive a compact drain current expression. The proposed ISFET model development is carried out by accounting for the interface trap effect. The voltage and current sensitivity of ISFET with and without interface traps are calculated to demonstrate the impact of interface traps in 2-D material-based pH sensors. It is observed that voltage sensitivity remains close to the Nernst limit (59 mV/pH), and a decrease in current sensitivity from 34 to 2 nA/pH is observed, as the density of interface traps is varied from 0 to <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${5}\times {10} {^{{12}}}$ </tex-math></inline-formula>cm<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$^{-}2 $ </tex-math></inline-formula>eV<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$^{-}1 $ </tex-math></inline-formula>. Another sensitivity metric, the transconductance to drain current ratio, is analyzed for its variation with changes in the density of interface traps. The noise level of ISFET in the presence of interface traps is analyzed, and its influence on the minimum resolvable pH is demonstrated. The proposed model prediction closely matches the TCAD simulation data obtained from a calibrated TCAD simulation setup. The model is suitable for implementation in Verilog-A for ISFET-based circuit simulation, and the impact of interface traps on the response time of ISFET-based circuits is also demonstrated.

A functionally enhanced transistor is a potential candidate for further advancing electronics and Moore's law beyond the classical scaling. This chapter discusses these kinds of multifunctional transistors called reconfigurable field-effect transistor (RFET) and reconfigurable tunnel field-effect transistor (RTFET). The RFET works on the principle of Schottky barrier tunneling, and the RTFET works on the principle of band-to-band tunneling. Both devices can be configured as an n-type and p-type device based on the biasing. This chapter explains the working and performance comparison of RFET and RTFET in detail with the help of technology computer-aided design (TCAD) simulations. Further, the potential and current models of a single-gated RFET and double-gated RTFET are presented in this chapter. The presented analytical models are compared and verified with TCAD simulations. The potential in the channel regions of RFET and RTFET is modeled by solving a two-dimensional (2D) Poisson's equation. Because the working principle of both devices is different, two different formulas are utilized for modeling the current in the device. The current model for the RFET is developed by integrating Landauer's formula, whereas the current model for RTFET is obtained by integrating band-to-band generation rate over the tunneling volume. The procedure, technique, and assumptions followed to obtain the potential and current models of RFET and RTFET are detailed in this chapter.

This study compares various machine learning models to determine the method with the highest accuracy rate to predict the equilibrium scour depth of bridge abutments. Unlike in previous studies, hyperparameter optimization is focused on increasing model performance in small datasets, and feature importance ranking is analyzed. This study employs Multiple Linear Regression (MLR), Support Vector Regression (SVR), Decision Tree Regressor (DTR), Random Forest Regressor (RFR), XGBoost, and Artificial Neural Networks (ANNs) to predict the Dse. The dataset consists of 150 records with the following key hydraulic parameters: the flow depth (Y), abutment length (L), channel width (B), flow velocity (V), and median grain size (d50). The results show that DTR achieved the highest accuracy (R2 = 0.992, accuracy = 99.28%), followed by XGBoost (R2 = 0.990, accuracy = 99.21%) and ANNs (R2 = 0.981, accuracy = 98.77%). Traditional MLR exhibited lower accuracy (R2 = 0.806, accuracy = 81.14%), confirming the superiority of AI-based models. These findings highlight the effectiveness of machine learning in scour prediction, providing a reliable alternative to conventional methods. This study underscores the potential of AI in hydraulic engineering, facilitating efficient bridge design and maintenance strategies. Future research should explore real-time data integration and hybrid AI models for improved interpretability and robustness.

The Moderate Resolution Imaging Spectroradiometer (MODIS) level-3 chlorophyll-a (Chl-a) product is one of the widely used ocean colour products that is often used for water quality monitoring of marine ecosystems. However, this product includes a large amount of missing data due to high surface reflectance and cloudy conditions that inevitably affect its suitability for spatiotemporal analysis of water quality. The objective of this study was to compare four Machine Learning (ML) techniques including K-nearest neighbour (KNN), Support Vector Regression (SVR), Random Forest Regression (RFR), and Artificial Neural Network (ANN) with well-known Data Interpolation Empirical Orthogonal Function (DINEOF) method for spatiotemporal missing imputation of MODIS Chl-a. The Southern Caspian Sea, which has a high Chl-a concentration, was selected as the case study. A cross-validation approach ranging missing data ratio from 0.1 to 0.8 was implemented to investigate the optimal parameters of the models and compare their performance for missing imputation. The results indicated that all ML models, except KNN, outperformed the DINEOF method for missing imputation of MODIS Chl-a. The SVR with the highest accuracy and the lowest variability of errors had the best performance among the five competing models, while the KNN showed the worst performance. The main reason for the better accuracy of the SVR than the other models is its structural risk minimization procedure that leads to the better generalization of the SVR model. The current results showed that the ML techniques used in the current study, the SVR in particular, are able to produce reliable imputations of the MODIS Chl-a missing data and can be a useful tool in water quality monitoring of marine ecosystems.

This work evaluates machine learning (ML) and deep learning (DL) models for Photovoltaic (PV) power output prediction based on two real-world datasets. Eight models – Support Vector Regression (SVR), K-Nearest Neighbors (KNN), Linear Regression (LR), Random Forest Regression (RFR), Gradient Boosting Regression (GBR), Decision Tree (DT), one Dimensional Convolutional Neural Network (1DCNN) and Artificial Neural Network (ANN) – were evaluated using Root Mean Squared Error (RMSE), R-squared (R²), Mean Absolute Error (MAE), and Mean Squared Error (MSE). For Dataset1 (region of Cacak) 1DCNN achieved the optimal performance, with the reduced errors and the largest R² values at three spatial locations, and it was closely followed by RFR and GBR. Likewise, in Dataset2 (region of Kraljevo), 1DCNN, ANN, and RFR give the best results. Conventional models, such as LR and DT, performed poorly in both data tribes. The results highlight the ability of state-of-the-art ensemble and DL models in learning nonlinear patterns in solar energy data and thus, the importance of choosing the right prediction tools to provide accurate and reliable PV power forecasts. &nbsp;

A machine learning (ML) method was used to optimize the trap distribution of the charge trap nitride (CTN) to simultaneously improve its performance/reliability (P/R) characteristics, which are tradeoffs in 3-D NAND flash memories. Using an artificial neural network (ANN), we modeled the relationship between trap distributions and P/R characteristics. The ANN was trained using a large experimentally calibrated technology computer-aided design (TCAD) simulation dataset. The gradient descent method was adapted to optimize the trap distribution, achieving the best P/R characteristics based on the well trained ANN. Eventually, we found the best trap profile distributed in both space and energy. In particular, the energetic trap distribution had a larger impact on the P/R characteristics than that of the spatial trap distribution. Furthermore, in terms of the P/R characteristics, it was generally preferable to increase all inputs of the energetic trap distribution. However, the acceptor-like trap energy level (ETA) and its standard deviation (sigma EA) caused a tradeoff between P/R characteristics; therefore, ML was used to determine their optimal points. The proposed ML method allows the optimization of trap distribution to obtain the best P/R characteristics rapidly and quantitatively. Our findings could be used as a guideline for determining the physical properties of CTN in 3-D NAND flash cells.

Insulated gate bipolar transistors (IGBTs) are the key component in power electronics, and the intricate relationship between their performance and structural parameters poses a formidable challenge in the design process. This article proposes an automatic optimal design method for IGBT structural parameters to leverage the pretrained machine learning (ML) model to efficiently predict the initial IGBT device’s performance, followed by utilizing the differential evolution (DE) algorithm to automatically adjust structural parameters based on the disparity between predicted and expected device performance until the expected performance is achieved. The method is validated in the design of punch-through IGBTs (PT-IGBTs) and trench gate field-stop IGBTs (FS-IGBTs), and the performance of technology computer-aided design (TCAD) simulation of the designed device is similar to the target performance. In particular, the simulation results of the designed FS-IGBT are highly fitted to the datasheet of the commercial device, which verifies the generalizability and effectiveness of the method. In addition, comparative analyses with various algorithms show DE provides the fastest optimization and extraordinary robustness under the exact specifications. Crucially, the proposed design scheme aligns with semiconductor physics. The method simplifies IGBT design without the need for manual tuning and TCAD tool simulation.

Artificial neural networks for infectious diarrhea prediction using meteorological factors in Shanghai (China)

Global solar radiation prediction by ANN integrated with European Centre for medium range weather forecast fields in solar rich cities of Queensland Australia

There is a growing consensus that the physics-based model needs to be coupled with machine learning (ML) model relying on data or vice versa in order to fully exploit their combined strengths to address scientific or engineering problems that cannot be solved separately. We propose several methodologies of bridging technology computer-aided design (TCAD) simulation and artificial intelligence (AI) with its application to the tasks for which traditional TCAD faces challenges in terms of simulation runtime, coverage, and so on. AI-emulator that learns fine-grained information from rigorous TCAD enables simulation of process technologies and device in real-time as well as large-scale simulation such as full-pattern analysis of stress without high demand on computational resource. To accelerate atomistic molecular dynamics (MD) simulation, we have done a comparison study of descriptor-based and graph-based neural net potential, and also show their capability with large-scale and long-time simulation of silicon oxidation. Finally, we discuss the use of hybrid modeling of AI- and physics-based model for the case where physical equations are either fully or partially unknown.