Nonlinear Prediction Research Articles

Enhancing Mental Health Predictions: A Gradient Boosted Model for Sri Lankan Camp Refugees

This study explores the mental health challenges encountered by Sri Lankan camp refugees, a population often marginalized in mental health research, and analyzes a range of factors including socio-demographic characteristics, living conditions in camps, and psychological variables. In quantitative mental health research, linear regression serves as a conventional approach for assessing the influence of diverse factors on mental health outcomes. However, this method fails to accommodate non-linear relationships between mental health variables and predictors and relies on stringent model assumptions that often do not align with real-world conditions. This study introduces a model-agnostic, advanced machine learning/artificial intelligence (ML/AI) technique, glmboost, as a viable alternative to linear regression. The glmboost algorithm is capable of fitting non-linear prediction models while also conducting variable selection. Moreover, the coefficients obtained from the glmboost model retain the same interpretability as those derived from linear regression. While the glmboost model identifies several key factors including post-migration living difficulties, post-traumatic stress disorder, difficulty in sleeping, poor family functioning, and lower informal support from families as markers of declining mental well-being among the Sri Lankan refugees, the linear regression overlooks vital predictors such as family functioning and family support, highlighting the importance of utilizing advanced ML/AI techniques like glmboost to comprehensively capture complex relationships between predictor variables and mental health outcomes among refugee populations. Thus, by introducing a novel, data-driven approach to mental health risk assessment, this study paves the way for more precise and efficient analyses and interventions in refugee settings, with the potential for improved resource allocation and personalized support, thus revolutionizing mental health service delivery in challenging environments. Additionally, it contributes to the academic discussion on refugee mental health while emphasizing the pivotal role of advanced data analytics in addressing complex health issues within humanitarian contexts.

Non-linear Time Series Prediction using Improved CEEMDAN, SVD and LSTM

This study recommends a new time series forecasting model, namely ICEEMDAN - SVD - LSTM model, which coalesces Improved Complete Ensemble EMD with Adaptive Noise, Singular Value Decomposition and Long Short Term Memory network. It can be applied to analyse Non-linear and non-stationary data. The framework of this model is comprised of three levels, namely ICEEMDAN level, SVD level and LSTM level. The first level utilized ICEEMDAN to break up the series into some IMF components along with a residue. The SVD in the second level accounts for de-noising of every IMF component and residue. LSTM forecasts all the resultant IMF components and residue in third level. To obtain the forecasted values of the original data, the predictions of all IMF components and residue are added. The proposed model is contrasted with other extant ones, namely LSTM model, EMD - LSTM model, EEMD - LSTM model, CEEMDAN - LSTM model, EEMD - SVD - LSTM model, ICEEMDAN - LSTM model and CEEMDAN - SVD - LSTM model. The comparison bears witness to the potential of the recommended model over the traditional models.

Abstract PO1-07-01: Artificial intelligence-based prediction of Oncotype DX Score from whole slide images using human-interpretable features and breast biomarkers

Abstract Background The Oncotype DX Breast Recurrence Score assay (ODX) is a commonly used genomic test for patients with estrogen receptor (ER)-positive, HER2-negative, early-stage invasive breast cancer. While ODX predicts patients’ recurrence risk and benefit from chemotherapy, it is tissue and time-consuming, and expensive. Previous deep-learning or non-linear ODX prediction models achieved promising performance using whole-slide images (WSI) or with other covariates (e.g., ER, progesterone receptor (PR), HER-2, Ki-67 scores and tumor stage) but systematic quantification of the contribution of individual histological features to the ODX score remained challenging. Here, we extracted a rich set of human-interpretable features (HIFs) quantifying nuclear morphology and the distribution of cells and tissues in the tumor microenvironment. We used these HIFs, along with manually assessed ER, PR, HER-2 scores, and tumor stage, to predict ODX scores. We also explored the role of Ki-67 features in augmenting our model predictions Methods We developed machine learning models to extract cell, tissue, and nuclear HIFs from WSI stained with hematoxylin and eosin (H&E) and immunohistochemistry (IHC) against Ki-67. These models were deployed on 353 H&E WSI to extract 276 HIFs quantifying cell densities, tissue areas, relative cell counts, and nuclei size, shape and color. Hierarchical agglomerative clustering was applied to cluster these features into 11 groups. Univariate regression was performed to identify correlations between ODX scores and feature clusters. To predict ODX score using HIFs, a multivariable regression model was fitted to the data where regressors are representative features from each cluster, in addition to ER, PR, HER-2 scores and stage as covariates. The model was fitted on the training set (N=266 slides) and evaluated on the held-out test set (N=87 slides). To examine the utility of Ki-67 IHC features in ODX prediction, a Ki-67 model was deployed on a subset of 194 matched Ki-67 IHC WSI to extract count proportions and densities of Ki-67 positive cells in cancer epithelium and stroma. Results Significant positive correlations were identified between ODX scores and feature clusters corresponding to cancer cell density (p < 10-7), macrophage density (p < 10-4), immune cell density (p < 10-16), and variations in cancer nuclear size (p < 10-5) and color (p < 10-3). Significant negative correlations were observed between ODX scores and clusters related to fibroblast density (p < 10-3), and variations of non-cancer cell nuclear color (p = 0.02). Evaluation of our model’s ability to predict ODX scores revealed an association between predicted and observed scores (Pearson r = 0.74). The AUROC for model predictions of high/low classifications (with reference to a cutoff ODX score of 20) was 0.80. Ki-67 features were clustered together with H&E cancer cell density features and showed a stronger correlation with ODX scores compared to the H&E features alone. Combining Ki-67 cell density features with H&E features led to increased performance of the multivariable regression model on the test set compared to H&E features only (r = 0.62 vs. r = 0.59). However, a model combining Ki-67, H&E and covariates performed similarly to a model with H&E and covariates (r = 0.73), suggesting that Ki-67 features may not be needed for ODX prediction when H&E and covariate information is available. Conclusions Our model predicts ODX recurrence scores with comparable performance to other black-box approaches using WSI and breast biomarker status and without the need for, Ki-67 features. Using readily available information, our model has the potential to be a convenient screening tool for patient stratification, which may lead to better patient care, lower costs, and faster treatment. Citation Format: Nhat Le, John Van Arnam, Christian Kirkup, Ylaine Gerardin, Michael Drage, Neha Khaitan, Anurag Sharma, Georgi Galev, Nicholas Indorf, Emily Sansevere, Ilan Wapinski, Stephanie Hennek, Gloria Zhang. Artificial intelligence-based prediction of Oncotype DX Score from whole slide images using human-interpretable features and breast biomarkers [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO1-07-01.

Bilinear optimization of protein structure prediction: An exact approach via AB off-lattice model

Protein structure prediction (PSP) remains a central challenge in computational biology due to its inherent complexity and high dimensionality. While numerous heuristic approaches have appeared in the literature, their success varies. The AB off-lattice model, which characterizes proteins as sequences of A (hydrophobic) and B (hydrophilic) beads, presents a simplified perspective on PSP. This work presents a mathematical optimization-based methodology capitalizing on the off-lattice AB model. Dissecting the inherent non-linearities of the energy landscape of protein folding allowed for formulating the PSP as a bilinear optimization problem. This formulation was achieved by introducing auxiliary variables and constraints that encapsulate the nuanced relationship between the protein’s conformational space and its energy landscape. The proposed bilinear model exhibited notable accuracy in pinpointing the global minimum energy conformations on a benchmark dataset presented by the Protein Data Bank (PDB). Compared to traditional heuristic-based methods, this bilinear approach yielded exact solutions, reducing the likelihood of local minima entrapment. This research highlights the potential of reframing the traditionally non-linear protein structure prediction problem into a bilinear optimization problem through the off-lattice AB model. Such a transformation offers a route toward methodologies that can determine the global solution, challenging current PSP paradigms. Exploration into hybrid models, merging bilinear optimization and heuristic components, might present an avenue for balancing accuracy with computational efficiency.

Nonlinear Levitation-Guidance Coupling Force Prediction for HTS Pinning Maglev Under Arbitrary Motion Based on Gated Recurrent Unit

Nonlinear Levitation-Guidance Coupling Force Prediction for HTS Pinning Maglev Under Arbitrary Motion Based on Gated Recurrent Unit

Nonlinear fatigue life prediction model based on isodamage curve considering two‐block loading

AbstractA non‐linear life prediction model considering the loading history is proposed based on isodamage curve. The stress‐controlled low‐high block loading of low cycle fatigue (LCF) at high temperature, with various previous cycles, was performed on a nickel‐based superalloy. The damage evolvement of the prior loading was revealed using scanning electron microscopy under the low‐high block loading, especially the failure mechanism of coaxing effect. In addition, load history correlation factor was introduced to describe the influence of prior cycle. Based on the Ni‐based superalloy, the life prediction model of two‐step loading, including the low‐high block (LH) and high‐low block loading (HL), was proposed, agreeing well with the experimental results for different metals. Compared with other life prediction models, the proposed model demonstrated the higher prediction accuracy and wider applicability.

A horizontal vibration prediction method of high-speed elevator based on transferred-digital twin model

The vibrations of high-speed-elevators (HSEs) significantly impact the comfort of elevator rides, with horizontal vibrations (HV) being the most sensitive to bodies. Customized elevator products are characterized by high-speed capabilities, customization options, and complex operational conditions. These characteristics often lead to challenges such as difficulties in predicting HV, low prediction accuracy, and limited reusability of prediction models and data. In order to mitigate HV in elevators, its crucial to implement HV prediction during the design phase. To address this need, this paper proposes a method for predicting HSEHV based on a Transferred Digital Twin (TLDT) model. The HSE’s design is physically extracted, and an elevator Digital Twin (DT) model with a digital layer is constructed. This model includes product geometry models, dynamics models, and simulation models. A transfer learning (TL) model is developed to integrate simulation data generated by the digital layer with measurement data acquired from the data interface layer, resulting in high-capacity and high-fidelity DT data. Using the support vector regression (SVR) method, a high-dimensional nonlinear HSEHV prediction model is established, and is trained and optimized using DT data to achieve data-driven HSEHV prediction. The validity of this approach is confirmed using measured data from an elevator test tower. Results indicate that the SVR-based TLDT model achieves the highest accuracy and the lowest absolute error values for both peak-to-peak and A95 vibration acceleration when compared to other models such as NN, KNN, decision tree, and lasso models. Furthermore, by incorporating DT data, the model accuracies improve by 16.9%, 14.5%, 8.6%, 14.1%, and 8.6%, and 12.3%, 12.4%, 19.7%, 7.6%, and 8.7%, respectively, for the mentioned models. Finally, the HV of KLK2 elevator car system is predicted. When compared with simulation analysis and the conventional SVR prediction model, obtained results through the proposed method closely align with the measured values.

Improving the statistical downscaling performance of climatic parameters with convolutional neural networks

Abstract This study examines two downscaling techniques, convolutional neural networks (CNNs) and feedforward neural networks for predicting precipitation and temperature, alongside statistical downscaling model as a benchmark model. The daily climate predictors were extracted from the European Center for Medium-range Weather Forecast (ECMWF) ERA5 dataset spanning from 1979 to 2010 for Tabriz city, located in the northwest of Iran. The biases in precipitation data of ERA5 predictors were corrected through the empirical quantile mapping method. Also, two nonlinear predictor screening methods, random forest and mutual information were employed, alongside linear correlation coefficient. While these methods facilitate identification of dominant regional climate change drivers, it is essential to consider their limitations, such as sensitivity to parameter settings, assumptions about data relationships, potential biases in handling redundancy and correlation, challenges in generalizability across datasets, and computational complexity. Evaluation results indicated that CNN, when applied without predictor screening, achieves coefficient of determination of 0.98 for temperature and 0.71 for precipitation. Ultimately, future projections were employed under two shared socioeconomic pathways (SSPs), SSP2-4.5 and SSP5-8.5, and concluded that the most increase in temperature by 2.9 °C and decrease in precipitation by 3.5 mm may occur under SSP5-8.5.

Graph convolutional multi-mesh autoencoder for steady transonic aircraft aerodynamics

Calculating aerodynamic loads around an aircraft using computational fluid dynamics is a user’s and computer-intensive task. An attractive alternative is to leverage neural networks (NNs) bypassing the need of solving the governing fluid equations at all flight conditions of interest. NNs have the ability to infer highly nonlinear predictions if a reference dataset is available. This work presents a geometric deep learning based multi-mesh autoencoder framework for steady-state transonic aerodynamics. The framework builds on graph NNs which are designed for irregular and unstructured spatial discretisations, embedded in a multi-resolution algorithm for dimensionality reduction. The test case is for the NASA common research model wing/body aircraft configuration. Thorough studies are presented discussing the model predictions in terms of vector fields, pressure and shear-stress coefficients, and scalar fields, total force and moment coefficients, for a range of nonlinear conditions involving shock waves and flow separation. We note that the cost of the model prediction is minimal having used an existing database.

Nominal Differences in Acute Symptom Presentation and Recovery Duration of Sport-Related Concussion Between Male and Female Collegiate Athletes in the PAC-12

BackgroundSport-related concussion (SRC) is a heterogenous injury that often presents with varied symptoms and impairment. Recently, research has focused on identifying subtypes, or clinical profiles of concussion to be used in assessing and treating athletes with SRC. The purpose of this study was to investigate sex differences in clinical profiles, recovery duration, and initial symptom severity after SRC in a cohort of collegiate athletes in the Pacific-12 Conference (Pac-12).MethodsThis prospective cohort study examined post-SRC symptoms, recovery, and return-to-play times using data from the Pac-12 CARE Affiliated Program and Pac-12 Health Analytics Program. Clinical profiles reported by student-athletes were defined by the number (> 50%) of specific symptoms frequently reported for each profile. Generalized linear mixed models were used to examine associations among sex, clinical profiles, time-to-recovery, and return-to-play times.Results479 concussion incidents met inclusion criteria. The probabilities of initial presentation of each clinical profile, initial injury severity scores, and recovery times within a profile did not differ between sexes (p = 0.33–0.98). However, both males and females had > 0.75 probabilities of exhibiting cognitive and ocular profiles. Initial injury severity score was a strong nonlinear predictor of initial number of clinical profiles (p < 0.0001), which did not differ between sexes. The number of clinical profiles was also a nonlinear predictor of time-to-recovery (p = 0.03) and return-to-play times (p < 0.0001).ConclusionsInitial symptom severity was strongly predictive of the number of acute clinical profiles experienced post-SRC. As the number of clinical profiles increased, time-to-recovery and time to return-to-play also increased. Factors other than sex may be better associated with acute symptom presentation post-concussion as no sex differences were found in reported clinical profiles or recovery. Understanding the number and type of clinical profiles experienced post-SRC may help inform concussion diagnostics and management.

Experimental validation of momentum transport theory in the core of H-mode plasmas in the ASDEX Upgrade tokamak

This study employs the established momentum transport analysis at ASDEX Upgrade [Zimmermann et al., Nucl. Fusion 63, 124003 (2023)] to investigate the parametric variations of the momentum transport coefficients in the core of H-mode plasmas. These experimental results are compared to a comprehensive database of gyrokinetic calculations. Generally, good agreement between predicted and measured diffusive and convective transport coefficients is found. The predicted and measured Prandtl numbers correlate most dominantly with the magnetically trapped particle fraction. The experimentally inferred pinch numbers strongly depend on the logarithmic density gradient and magnetic shear, consistent with the theoretical predictions of the Coriolis pinch. The intrinsic torque from residual stress in the inner core is small, scales with the local logarithmic density gradient, and the data indicate a possible sign reversal. In the outer periphery of the core, the intrinsic torque is always co-current-directed and scales with the pressure gradient. This is consistent with prior experimental findings and global, non-linear gyrokinetic predictions. It suggests that profile shearing effects generate the intrinsic torque in the inner core. Toward the outer core, most likely, effects from E×B-shearing become more influential. These results offer the first comprehensive picture of this transport channel in the core plasma and contribute to validating the corresponding theoretical understanding. The derived scaling laws are used to construct a reduced momentum transport model, which has been validated against an additional dataset. This demonstrates that the model captures the essential contributions to momentum transport in the core of H-mode plasmas.

Ultra-fast tuning of neural network controllers with application in path tracking of autonomous vehicle

Neural network (NN) controllers have shown great potential in solving complex control or decision-making tasks. However, most of the NN controllers either rely on the availability of large datasets or require dense interactions with the environment, which hinders their application in real systems. In this paper, we introduce a model-based reinforcement learning (MBRL) algorithm, aimed at realizing ultra-fast tuning of deep NN controller from a small sample set of real-world data. The algorithm uses Gaussian processes (GPs) to model the unknown dynamics of real system and updates controller parameters through stochastic gradient descent. By using particle-based method for long-term predictions, the algorithm can easily incorporate online state estimators and filters into controller learning, which is conductive to learning from systems with partially measurable states and stochastic control delay. We apply the algorithm to calibrate a deep NN controller for the path tracking of a full-size autonomous vehicle (AV). Simulation and field test results show that the deep NN controller can be well calibrated after only one interaction with the environment and can achieve similar tracking performance to optimization-based methods such as nonlinear model prediction control (NMPC) in various test scenarios by combining with a feed-forward pure pursuit (PP) controller.

Beyond sleepless nights: Unraveling the complexity of alexithymia and suicide risk among university students.

There is a solid relationship between alexithymia and suicide risk. Nonetheless, the specific impact of alexithymia's distinct subscales on suicide risk has received the attention it deserves. This article presents a comprehensive exploration of suicide risk among university students, focusing on the interconnections among alexithymia, insomnia, and suicidal behavior. Three components of alexithymia including difficulties in describing emotions or feelings (DDF), difficulties in identifying emotions or feelings (DIF), and the externally oriented thinking were considered. The study involved 208 participants from a Persian university sample, examining the significance of incorporating both alexithymia and insomnia in suicide risk assessment and intervention planning. Insomnia was positioned as a pivotal mediator. A secure electronic link in the Telegram application was employed to collect the data. Both linear and nonlinear prediction models were used to explore potential associations among alexithymia, insomnia, and suicide risk. The study revealed substantial positive correlations between alexithymia and suicide risk, as well as between insomnia and suicide risk. Additionally, specific components of alexithymia exhibited noteworthy links to suicide risk. The inclusion of insomnia scores in suicide risk predictions is critical, as it greatly enhances the precision of risk assessments and facilitates the design of targeted and effective therapeutic interventions. The association between alexithymia and suicide risk showed a significant relationship (r=.29, p<.01). Moreover, a significant correlation was observed between alexithymia and insomnia (r=.32, p<.01). Additionally, insomnia exhibited a significant positive correlation with suicide (r=.35, p<.01). Interestingly, DDF and DIF showed positive correlations with suicide (r=.28, p<.01; r=.33, p<.01). The findings carry profound implications for suicide prevention efforts, providing valuable insights to safeguard the well-being and resilience of university students facing suicide risk challenges.

Impact of coronavirus pandemic on stock index: A polynomial regression with time delay

MotivationUnder contemporary market conditions in China, the stock index has been volatile and highly reflect trends in the coronavirus pandemic, but rare scientific research has been conducted to model the possible nonlinear relations between the two indicators. Added, on the advent that covid-related news in one time period impacts the stock market in another period, time delay can be an equally good predictor of the stock index but rarely investigated. ObjectivesTo contribute to filling the gaps identified in existing research, this study models relationship between the stock market index and coronavirus pandemic by leveraging volatility in the stock market and covid data through time delay and best degree in a polynomial environment. The resultant optimal time delay and best degree model is used to derive a high-accuracy prediction of stock market index. NoveltyIn line with the possible relations, the novelty of this study is that it proposes, validates and implements polynomial regression with time delay to model nonlinear relationship between the stock index and covid. MethodsThis study utilizes high-frequency data from January 2020 to the first week of July 2022 to model the nonlinear relationship between the stock index, new covid cases and time delay under polynomial regression environment. FindingsThe empirical results show that time delay and new covid cases, when modelled in a polynomial environment with optimal degree and delay, do present better representation of the nonlinear relationship such predictors have with stock index for China. Relative to results from the polynomial regression without delay, the empirical evidence from the model with delay show that an optimal time delay of 17 weeks makes it possible to predict the stock index at high accuracy and record improvements of 16-fold or higher. The representative delay model is used to project for up to 17 weeks for future trends in the stock index. ImplicationThe implication of the findings herein is that the prowess of the time delay polynomial regression is heavily dependent on instability in covid-related time trends and that researchers and decision-makers should consider modeling to cover for the unsteadiness in coronavirus cases to achieve better results.

Nonlinear Acoustic Behavior Prediction and Analysis of Straight Through Perforated Tube Silencer With Flow

Straight through perforated tube silencers are widely used in intake and exhaust noise control, and their acoustic behaviors are influenced by the gas flow and high amplitude sound remarkably. To investigate the acoustic attenuation performance of straight through perforated tube silencers in the presence of flow and high amplitude sound, an approach based on the three-dimensional (3D) time-domain computational fluid dynamics (CFD) simulation is employed and validated by comparing predictions and measurements. Transmission loss of the straight through perforated tube silencer under pure tone and multi-tone sound excitations is predicted. Results show that, under the pure tone sound excitations, the acoustic attenuations are affected remarkably by the high amplitude sound, and the nonlinear effects turn to weaken as flow velocity increases. With the sound pressure level (SPL) decreasing along the axis of the silencer, the acoustic nonlinearity in the orifice weakens gradually. The shedding vortices are restricted to the region near perforations because of the transport of flow. When the multi-tone sound excitation is enforced on the inlet, transmission loss of the silencer is different from the pure tone excitation at the same SPL of every compositional frequency.

The adaptive markets hypothesis: Insights into small stock market efficiency

ABSTRACT In this paper, we explore whether the adaptive markets hypothesis (AMH) describes the efficiency of the Finnish stock market better than the efficient markets hypothesis (EMH) does. Building on this, we also test how small market size and market liberalization impact the efficiency of the Finnish stock market and examine the relationship between market volatility and return in this market. We conduct this study by applying the subsample analysis and the rolling window analysis to the daily returns of the OMXH25 index and by measuring the efficiency through three linear and two nonlinear predictability tests. The results of our study strongly support the AMH. They also suggest that small market size alone does not make a market less efficient; opening a market to foreign investors improves its efficiency after a delay; and the correlation between market volatility and return varies over time in the Finnish stock market, being usually negative. These findings mostly contradict the traditional investment paradigm.

Matching Job Demands and Job Resources as Linear and Non-linear Predictors of Employee Vigor and Sustainable Performance

ABSTRACT The present study studied two prominent job stress models as predictors of employee vigor and employee sustainable performance. First, based on the matching principle of the Demand-Induced Strain Compensation (DISC) Model, we predicted that job resources moderate (i.e. strengthen) the linear relation of job demands with employee vigor and sustainable performance, particularly when the type of demands corresponds well with the type of resources. Second, based on the non-linearity principle of the Vitamin Model, we predicted that non-linear (i.e. curvilinear) relations between job demands or job resources with employee vigor and sustainable performance exist, particularly when the type of demands or resources matches well with the outcome. An online cross-sectional study was conducted among 1,175 Chinese employees working in different industries. Multivariate regression analyses revealed that (1) although we did not find moderating effects of job resources, “conceptual matches” were found in line with the matching principle of the DISC Model, and (2) curvilinear patterns were detected (although with a non-expected shape) in which so-called “non-linear matching” relations between either demands or resources and outcomes were present. This study highlights the importance of matching demands and resources in both linear and non-linear ways to predict employee vigor and employee sustainable performance. Finally, theoretical and practical implications of the findings were addressed.

Predictive Model of Restart-Up Pressure Drop after Shutdown for Heavy Oil-Water Ring Transportation Pipeline.

Aiming at the problem of restart-up for a heavy oil-water ring transportation pipeline due to instability and damage of the water ring, based on the self-developed design of a small indoor loop simulation experimental device and taking four kinds of ordinary heavy oil in the Lvda oilfield as the research object, the change trend of restart-up pressure drop with time is experimentally studied when the pipeline is restarted-up after shutdown at a constant water flow. On the basis of the regression analysis of the orthogonal restart-up experimental data of four factors (oil holdup, oil viscosity, standstill period, and water cleaning superficial velocity) and mixed levels by the statistical product and service solutions statistical analysis software, a multivariate nonlinear restart-up maximum pressure drop prediction model is established. Through analysis of the characteristics of each stage of the restart-up process, an exponential decay model of restart-up pressure drop with time is created. The research results show that the variations in restart-up pressure drop with time can be divided into two stages: the attenuation stage and the equilibrium stage. The predicted value of restart-up pressure drop with time is in good agreement with the measured one, and the goodness of fit is very close to 1. The maximum restart-up pressure drop rises along with the increase in oil holdup, oil viscosity, standstill period, and water cleaning superficial velocity. The restart-up time prolongs with the increase in oil holdup, oil viscosity, and standstill period but shortens with the increase in water cleaning superficial velocity.

A numerical research on transient transport of high-temperature particles associated with air gouging process

The high-temperature particles produced by air gouging processes are common pollutants in cleaning workshops, which could endanger workers’ health. In this research, a two-way coupled Eulerian-Lagrangian method based on discrete phase modelling (DPM) was used to investigate the transport characteristics of high-temperature particles generated from air gouging processes. The particle group in the movement was divided into core zone and boundary zone, and the kinetic mechanism of particles in different zones was analysed. The results show particles in the boundary zone are more susceptible to the vortex entrainment effect than particles in the core zone, and some particles could be moved from the vortex diffusion zone to the vortex recirculation zone. The smaller the particle size, the better the followability to the hot airflow. The slower the temperature decay of particles, the more significant the thermophoresis effect, which could enhance the diffusion behaviour of the small particles. By analysing the average residence time of particles in the breathing zone, the influence of different factors on the average vertical diffusion height [Formula: see text] of particles is summarized, and a nonlinear regression prediction model of [Formula: see text] is established. The results can contribute to health risk assessment and industrial ventilation design.

Next-generation non-linear and collapse prediction models for short- to long-period systems via machine learning methods

Since the 1960 s, a cornerstone of earthquake engineering has been estimating the non-linear response of structures based just on lateral strength, modal properties, and the anticipated seismic demand. Over the years, several studies have quantified this empirical relationship and integrated it within seismic design and assessment methodologies. These have been widely accepted for practical application and adopted in building codes worldwide. While these models work reasonably well, there are still areas in which improvements can be made, especially concerning their robust quantification of uncertainty. This is mainly due to the amount of data used to quantify these empirical relationships, the choice of functional forms during the fitting, the model fitting and testing process, and how the ground motion shaking intensity is characterised. This study tackles these issues via the non-linear analysis of single-degree-of-freedom oscillators to train several machine learning (ML) models. This was to examine the accuracy and applicability of such models within a seismic engineering context and explore potential gains in quantifying the non-linear response of structures via next-generation intensity measures, namely average spectral acceleration, Saavg. The results show that the Decision Tree and XGBoost models worked well across a broad range of periods, accurately predicting collapse and non-collapse responses. Appraising these with existing models showed a notable improvement all around. It indicates that the models based on data-driven ML approaches represent a positive step and can be seamlessly integrated with seismic analysis methodologies utilised worldwide. Subsequently, a Python-based library of these XGBoost ML models was developed and made available online to foster practical application.