Model Building Research Articles

Data Models for Casting Processes – Performances, Validations and Challenges

Abstract Data-driven models with their associated data learning and training schemes can be utilised for the light metal casting processes. This paper presents the basis of data model building processes along with data training and learning exercises for vertical direct chill casting and high pressure die casting (HPDC) applications. The concepts of efficient database building, data translations and sampling, as well as real-time model building and validations are briefly discussed. Rigorous performance studies were additionally carried out for two real-world case studies. Different combinations of data solvers and interpolators are adapted for the model building techniques, while machine learning schemes are used for data trainings.

Deep convolutional neural network-based identification and biological evaluation of MAO-B inhibitors

Parkinson's disease (PD) is one of the most prominent motor disorder of adult-onset dementia connected to memory and other cognitive abilities. Individuals with this vicious neurodegenerative condition tend to have an elevated expression of Monoamine Oxidase-B (MAO-B) that catalyzes the oxidative deamination of aryalkylamines neurotransmitters with concomitant reduction of oxygen to hydrogen peroxide. This oxidative stress damages mitochondrial DNA and contributes to the progression of PD. To address this, we have developed a deep learning (DL)-based virtual screening protocol for the identification of promising MAO-B inhibitors using Convolutional neural network (ConvNet) based image classification technique by dealing with two unique kinds of image datasets associated with MACCS fingerprints. Following model building and prediction on the Maybridge library, our approach shortlisted the top 11 compounds at the end of molecular docking protocol. Further, the biological validation of the hits identified 4 compounds as promising MAO-B inhibitors. Among these, the compound RF02426 was found to have >50 % inhibition at 10 μM. Additionally, the study also underscored the utility of scaffold analysis as an effective way for evaluating the significance of structurally diverse compounds in data-driven investigations. We believe that our models are able to pick up diverse chemotype and this can be a starting scaffold for further structural optimization with medicinal chemistry efforts in order to improve their inhibition efficacy and be established as novel MAO-B inhibitors in the furture.

Multi-objective optimal scheduling of electricity consumption in smart building based on resident classification

By the end of 2021, the urbanization rate of China had reached 64.72%. Increasing urbanization rates have led to an 80 percent share of carbon emissions in urban areas. Therefore, a load optimal scheduling model of smart buildings is proposed in this paper, which takes into account the cost of carbon emissions. The model aims to minimize the electric cost for building residents and maximize the utilization of distributed energy. The research focuses on the load optimal scheduling for two types of residents, i.e., the working residents and the retired ones, based on their electricity usage patterns and habits. To address the challenges of uneven load distribution, nonlinear behavior, and uncertainties introduced by electric vehicles and photovoltaic energy, the multi-objective beluga whale optimization algorithm with hybrid reverse learning competitive strategies (RCMBWO) is introduced. It utilizes the energy storage and discharge capabilities of electric vehicles to mitigate the uncertainties of photovoltaic energy generation. The simulation results show that load optimal scheduling for both types of residents can lead to significant cost savings for a smart building with 100 households, approximately 29,554 RMB per year. Furthermore, the optimized results can guide the determination of the appropriate size for the photovoltaic energy storage system, reducing energy waste resulting from insufficient energy consumption capability of smart buildings. The research on targeted load optimal scheduling for classified residents presents a viable solution for enhancing the cost-effectiveness and environmental benefits of smart buildings.

Indoor thermal environment assessment of a historic building for a thermal and energy retrofit scenario using a CFD model

Research on retrofit scenarios for the energy efficiency of historical buildings is growing. Despite their potential to improve the energy consumption of a building, retrofit scenarios for indoor thermal environments and the thermal comfort of historical buildings has not been sufficiently studied. Therefore, in this study, indoor airflow, heat distribution, and thermal comfort levels of indoor residents were predicted using computational fluid dynamics (CFD) modeling of historic buildings based on measured values. Although the heating was turned on in the meeting room in winter, heat loss caused by low thermal performance, airtightness, and inflow of cold air caused local thermal discomfort. Individuals seated near the windows had a temperature difference of up to 2.6 °C between the head and ankle levels. Heat loss was identified from the exterior walls and windows, with air infiltration significantly affecting indoor thermal comfort. The airtightness of the window frames of historic buildings was low, thus necessitating related improvement in heat and energy retrofit scenarios. This study propose a suitable approach to resolve this problem.

Analysis of carbon emission characteristics and establishment of prediction models for residential and office buildings in China

Intensifying environmental issues make carbon reduction crucial, especially in the construction industry. Embodied carbon emissions (ECE) in buildings characteristically involve concentrated, intensive emissions over a short period. Implementing energy-saving and carbon-reducing technologies has raised ECE rates. Therefore, analyzing carbon emission (CE) characteristics during the production and construction phases is necessary. Currently, essential data on ECE is lacking, the calculation methods are complex, and ECE's characteristics and influential factors in residential and office buildings remain unclear. Moreover, there is a lack of prediction models for ECE in the construction phase that apply to current construction methods and CE factors. The research on prediction models for ECE has primarily focused on the production phase but has paid little attention to the contribution of non-structural materials to embodied carbon (EC). In this study, reinforced concrete residential and office buildings in hot summer and cold winter regions (located in the southeast part of mainland China, with a subtropical monsoon climate) were considered typical types, and the CE characteristics of two main phases, the production and construction phases, were analyzed. Based on critical features, such as the number of floors and the basement, prediction models for ECE in residential and office buildings were established at the design stage. The models were validated with mean absolute percentage error (MAPE) values below 4%, and coefficient of variation (root mean square error) (CV [RMSE]) values less than 0.02. The prediction models enabled a rapid estimation of ECE, encouraging designers to consider EC in the early design stages.

Integrating HEC-RAS, GIS, and LISREL for assessing and enhancing urban building resilience against flood threats: Comprehensive model and analysis

Floods pose significant threats to urban areas, resulting in substantial human and financial losses annually. The vulnerability of key urban centers to these risks diminishes their efficiency, leading to public dissatisfaction and service deficiencies. Recognizing and enhancing the resilience of essential buildings becomes crucial in mitigating these challenges. This study employs a comprehensive approach to achieve a resilience model for critical buildings facing floods. The research process involves the identification of city assets based on leveling criteria, utilizing GIS technology for spatial mapping. HECRAS software aids in river flow modeling, identifying areas lacking flood-carrying capacity. By overlaying vulnerable gravity centers with flood-prone regions, building resilience components are computed through structural factor analysis and LISREL modeling. The study identifies ten key criteria. Further analysis includes resilience modeling using TOPSIS and AHP methods. The positive ideal and negative ideal solutions are determined, resulting in the grading of building resilience. Notably, the balance redundancy index with cascading potential effects attains the highest positive ideal value at 0.257, while the resistance to a level of stress index achieves the lowest negative ideal value at 0.02. This comprehensive approach and modeling contribute to the understanding and enhancement of urban building resilience in the face of flood threats.

U-Net architecture for high-resolution thermal image segmentation: Enhancing CFD models in smart HVAC design

This paper presents an enhanced U-Net architecture for accurately segmenting high-resolution thermal images to facilitate the design of intelligent Heating, Ventilation, and Air Conditioning (HVAC) systems using Computational Fluid Dynamics (CFD) models. Indoor thermal comfort for occupants will improve under the same energy load by imposing comfort demands. Segmented data is fed into CFD simulations using a U-Net variant for thermal imagery. Integrating the segmented thermal data into the CFD simulations increases the model’s accuracy by 15 % for segmentation and 20% for prediction when compared to a traditional U-Net architecture. Experimental findings demonstrate the potential to achieve energy savings of up to 25 % in the simulated scenario by using the trained and validated model on a dataset of 1,500 high-resolution thermal images of various building styles. Compared to existing methods, the proposed method saves 30% in total simulation time for complex building models. Deep learning enables building the next generation of HVAC systems by pushing the boundaries of energy-efficient and environmentally friendly buildings. This study shows that advanced computer vision can be leveraged to create more innovative and sustainable built environments, leading to drastic improvements in energy efficiency.

Facile and efficient superhydrophobic film for radiative daytime cooling

Abstract Facile and efficient radiative daytime cooling technology is still a huge challenge. In this paper, we proposed an eco-friendly, facile, efficient superhydrophobic radiative daytime cooling film composed of polydimethylsiloxane, nano TiO2, and nano SiO2. This radiative daytime cooling film could have about 15°C cooling of Al sheet, 9°C cooling for the building model on a sunny summer day, and 8°C cooling for the building model on a cloudy summer day, respectively. This method is promising to be used in building cooling fields.

Numerical reconstruction of atmospheric boundary layer seasonal turbulent wind field over a complex forest terrain

The wind characteristics of the atmospheric boundary layer (ABL) over complex forest terrains are inherently intricate, influenced by the interplay of rugged topography, seasonal climatic fluctuations, and periodic vegetation dynamics. These effects are especially evident in the near-ground wind field, which exhibits substantial seasonal variability. Conventional wind characterization methods, as outlined in current standards, often fail to accurately capture these seasonal variations, thereby complicating the reconstruction of the near-ground ABL turbulent wind field in complex forest terrains. Accordingly, we employ the narrow band synthetic random flow generation (NSRFG) technique within large eddy simulation (LES) to generate inflow turbulence representing the growing and baldness seasons in a complex forest terrain by adjusting parameter equations for seasonal adaptation and introducing new empirical equations for the turbulent spectrum. Subsequently, we verified the seasonal turbulent flow's statistical characteristics and flow structure to assess its feasibility and validity, ultimately establishing a ‘seasonal numerical wind field’ model. Finally, the seasonally modified LES-NSRFG method was applied to the numerical simulation of turbulent flow around the Commonwealth Advisory Aeronautical Research Council (CAARC) standard high-rise building model. A comprehensive comparison of wind effects was conducted for the CAARC model under varying incoming flow conditions. The results indicate that seasonal winds in a complex forest terrain significantly affect the building's vortex wake, increasing the irregularity and complexity of the structural wind pressure and base moment coefficients. Thus, the seasonal wind effect must be considered when designing wind-resistant engineering structures in forest regions moving forward.

Building stock energy modeling to assess annual progress in stock energy efficiency and carbon emission reduction of commercial buildings

Short-term progress assessments of carbon dioxide emission reductions are essential for nations to meet their medium/long-term targets. Bottom-up building stock energy modeling (BSEM) methods can contribute to this issue but have not been applied because of the need to represent short-term building stock dynamics in building systems and energy conservation measures (ECMs). To fill this gap, this study developed a BSEM framework that integrates top-down building stock decomposition, including building systems and ECMs, and bottom-up physics-based energy demand quantification using reference building models representing building stock segments. The framework was verified through a case study of a Japanese commercial building stock. The results indicate that the developed model effectively captures the short-term dynamics and contributions of the technologies. Although there are significant errors in estimating some building subsectors and end uses, the model predicts the changes in the aggregated energy consumption with acceptable accuracy. The Japanese 2030 emission reduction target cannot be achieved with the current technology deployment trends; however, the shortfall can be addressed by applying additional measures. Owing to its applicability to diverse building stocks, the framework promotes the use of BSEM for policy assessment and evidence-based policymaking for climate change mitigation.

A boundary-aware point clustering approach in Euclidean and embedding spaces for roof plane segmentation

Roof plane segmentation from airborne light detection and ranging (LiDAR) point clouds is an important technology for three-dimensional (3D) building model reconstruction. One of the key issues of plane segmentation is how to design powerful features that can exactly distinguish adjacent planar patches. The quality of point feature directly determines the accuracy of roof plane segmentation. Most of existing approaches use handcrafted features, such as point-to-plane distance, normal vector, etc., to extract roof planes. However, the abilities of these features are relatively low, especially in boundary areas. To solve this problem, we propose a boundary-aware point clustering approach in Euclidean and embedding spaces constructed by a multi-task deep network for roof plane segmentation. We design a three-branch multi-task network to predict semantic labels, point offsets and extract deep embedding features. In the first branch, we classify the input data as non-roof, boundary and plane points. In the second branch, we predict point offsets for shifting each point towards its respective instance center. In the third branch, we constrain that points of the same plane instance should have the similar embeddings. We aim to ensure that points of the same plane instance are close as much as possible in both Euclidean and embedding spaces. However, although deep network has strong feature representative ability, it is still hard to accurately distinguish points near the plane instance boundary. Therefore, we first robustly group plane points into many clusters in Euclidean and embedding spaces to find candidate planes. Then, we assign the rest boundary points to their closest clusters to generate the final complete roof planes. In this way, we can effectively reduce the influence of unreliable boundary points. In addition, to train the network and evaluate the performance of our approach, we prepare a synthetic dataset and two real datasets. The experiments conducted on synthetic and real datasets show that the proposed approach significantly outperforms the existing state-of-the-art approaches in both qualitative evaluation and quantitative metrics. To facilitate future research, we will make datasets and source code of our approach publicly available at https://github.com/Li-Li-Whu/DeepRoofPlane.

Model Building for Prediction of Valuation of Growth Stocks

It has been a great challenge for investors to evaluate growth stocks because these types of stocks are especially hard to forecast because of their inconsistencies and uncertainty. This paper establishes not only regression models, which include a regular Multivariable Linear Regression (MVLR) and a Fixed Effects (FE) Regression Model but also the application of a Panel Vector Autoregression (PVAR) Model on a dataset aiming to predict the Price-to-Sales Ratio (P/S) with several variables. Excel, Python, and Rstudio environments are used in the process of coding and model building. More specifically, Excel is used to conduct the MVLR model, Python is used for the FE model, and Rstudio is used for the PVAR model. By comparing different models of each type, the most optimized model was determined using a training set. After applying predictions with all types of models using another testing set, the best model is identified to be the best approach to the problem. Ultimately, all three models are compared. PVAR model and FE model are proved to be the more reliable ones out of the three models, because the R-squared for MVLR is 0.1859, lower than that of the FE model, 0.524, and the Mean Absolute Error (MAE) of for PVAR Model and FE model are respectfully 53.2139 and 55.1529. This project serves as a good starting point for future analysis and prediction of financial metrics like P/S ratio and is able to have great significance as it discovers some key points for the prediction of P/S value like its connection with time and entity despite the fact that the models themselves aren’t the best fit mainly caused by the limitations of data and variable access.

Effect of prevalent irregular configuration of infills on the seismic collapse of Indian RC buildings

Rapid urbanization and land scarcity obligated the urban population for multi-use of Reinforced Concrete (RC) buildings to meet both residential and commercial purposes together. Un-Reinforced Masonry (URM) infills are often placed irregularly in plan and or elevation of the buildings in some floor(s), particularly in the ground floor to suit the need of occupational requirements accompanied by the inherent irregularities due to openings for doors and windows to meet functional requirements for residence. Seismic performance of infilled frame highly depends upon degree of infill-frame interaction, and it is of utmost importance to classify infilled frame buildings in various Model Building Types (MBTs) based on the type of irregular placement of infills in addition to the parameters affecting seismic performance. Based on field pilot surveys carried out in Indian cities, URM infilled buildings have been classified into 14 different MBTs. Seismic performance of the prevalent categories of MBTs have been evaluated through Incremental Dynamic Analysis (IDA). It has been observed that irregular placement of infills increases the seismic collapse risk significantly, and reduces collapse capacity to approximately 50% as compared to ideal uniformly infilled building. Global collapse occurred due to failure of structural members in the vicinity of irregularity created due to infills.

Predictive Model Building for Aggregation Kinetics Based on Molecular Dynamics Simulations of an Antibody Fragment.

Computational methods including machine learning and molecular dynamics simulations have strong potential to characterize, understand, and ultimately predict the properties of proteins relevant to their stability and function as therapeutics. Such methods would streamline the development pathway by minimizing the current experimental testing required for many protein variants and formulations. The molecular understanding of thermostability and aggregation propensity has advanced significantly along with predictive algorithms based on the sequence-level or structural-level information on a protein. However, these approaches focus largely on a comparison of protein sequence variations to correlate the properties of proteins to their stability, solubility, and aggregation propensity. For therapeutic protein development, it is of equal importance to take into account the impact of the formulation conditions to elucidate and predict the stability of the antibody drugs. At the macroscopic level, changing temperature, pH, ionic strength, and the addition of excipients can significantly alter the kinetics of protein aggregation. The mechanisms controlling aggregation kinetics have been traced back to a combination of molecular features, including conformational stability, partial unfolding to aggregation-prone states, and the colloidal stability governed by surface charges and hydrophobicity. However, very little has been done to evaluate these features in the context of protein dynamics in different formulations. In this work, we have combined a range of molecular features calculated from the Fab A33 protein sequence and molecular dynamics simulations. Using the power of advanced, yet interpretable, statistical tools, it has been possible to uncover greater insights into the mechanisms behind protein stability, validating previous findings, and also develop models that can predict the aggregation kinetics within a range of 49 different solution conditions.

Predicting the risk of colorectal cancer among diabetes patients using a random survival forest-guided approach.

Colorectal cancer (CRC) is the third most frequently diagnosed cancer worldwide. Diabetes and CRC share many overlapping lifestyle risk factors such as obesity, heavy alcohol use, and diet. This study aims to develop a risk scoring system for CRC prediction among diabetes patients using routine medical records. A retrospective cohort study was conducted using electronic health records of Hong Kong. Patients who received diabetes care in public general outpatient clinics between 2010 and 2019 and had no cancer history were identified, and followed up until December 2019. The outcome was diagnosis of CRC during follow-up. For model building, predictors were first selected using random survival forest, and weights were subsequently assigned to selected predictors using Cox regression. Of the 386,325 patients identified, 4,199 patients developed CRC during a median follow-up of 6.2 years. The overall incidence rate of CRC was 1.93 per 1000 person-years. In the final scoring system, age, waist-to-hip ratio, and serum creatinine were included as predictors. The C-index on test set was 0.651 (95%CI: 0.631-0.669). Elevated serum creatinine (≥127 µmol/L) could be a potential important predictor of increased CRC risk. While obesity is a well-known risk factor for CRC, renal dysfunction could be potentially linked to an elevated risk of CRC among diabetes patients. Further studies are warranted to explore whether renal function could be a potential parameter to guide screening recommendation for diabetes patients.

Analysis of Public Interest in Smartfren SIM Cards using the K-Nearest Neighbors Method

The use of Smartfren SIM cards is increasing along with the public's need for fast and stable internet services. However, a deep understanding of public interest in the SIM card is necessary to optimize marketing strategies and increase sales. Proper analysis can help companies identify potential target markets and develop effective marketing strategies. We chose the K-Nearest Neighbors method to analyze public interest in using Smartfren SIM cards. This study aims to develop and evaluate the K-Nearest Neighbors model in predicting public interest in using Smartfren SIM cards. This study uses a dataset containing information about Smartfren SIM card users. We divide the data into two sets: a training set for model building and a test set for evaluating model performance. We apply the K-Nearest Neighbors method to classify the data into two categories: interested and not interested. We evaluate the model performance using accuracy, precision, recall, and F1-score metrics. We present the evaluation results as a confusion matrix. The developed K-Nearest Neighbors model showed excellent performance with an accuracy of 94.29%, a precision of 94.20%, a recall of 100%, and an F1-score of 97.01%. These results indicate that the K-Nearest Neighbors model is effective in predicting people's interest in Smartfren SIM cards. The high recall value indicates that the model is able to identify all interested individuals without missing any, while the high precision value indicates that the model rarely makes false positive prediction errors. This study concludes that the K-Nearest Neighbors method is very effective for use in analyzing people's interest in using Smartfren SIM cards. We can rely on the developed model's strong performance for real-world applications in marketing strategies.

Analysis Of Public Interest In Smartfren SIM Cards Using The K-Nearest Neighbors Method

The use of Smartfren SIM cards is increasing along with the public's need for fast and stable internet services. However, a deep understanding of public interest in the SIM card is necessary to optimize marketing strategies and increase sales. Proper analysis can help companies identify potential target markets and develop effective marketing strategies. We chose the K-Nearest Neighbors method to analyze public interest in using Smartfren SIM cards. This study aims to develop and evaluate the K-Nearest Neighbors model in predicting public interest in using Smartfren SIM cards. This study uses a dataset containing information about Smartfren SIM card users. We divide the data into two sets: a training set for model building and a test set for evaluating model performance. We apply the K-Nearest Neighbors method to classify the data into two categories: interested and not interested. We evaluate the model performance using accuracy, precision, recall, and F1-score metrics. We present the evaluation results as a confusion matrix. The developed K-Nearest Neighbors model showed excellent performance with an accuracy of 94.29%, a precision of 94.20%, a recall of 100%, and an F1-score of 97.01%. These results indicate that the K-Nearest Neighbors model is effective in predicting people's interest in Smartfren SIM cards. The high recall value indicates that the model is able to identify all interested individuals without missing any, while the high precision value indicates that the model rarely makes false positive prediction errors. This study concludes that the K-Nearest Neighbors method is very effective for use in analyzing people's interest in using Smartfren SIM cards. We can rely on the developed model's strong performance for real-world applications in marketing strategies.

Метод опорних моделей синтезу інтелектуальних систем ідентифікації нелінійних динамічних об’єктів

The paper is devoted to resolving the contradiction between the accuracy of modeling nonlinear dynamics and the speed of model construction under conditions of limited computing resources. The aim of the work is to reduce the time for building models of nonlinear dynamic objects with continuous characteristics in the form of neural networks while ensuring a given modeling accuracy. This goal is achieved by developing a new method for synthesizing intelligent systems based on the superposition of pre-trained (support) models in the form of neural networks that reflect the basic properties of the object. The scientific novelty of the work lies in the development of a method for identifying nonlinear dynamic objects in the form of neural networks with time delays based on a set of pre-trained neural network models that reflect the basic properties of the subject area. In contrast to the traditional approach to pre-training, the developed method allows building models of lower complexity and with shorter training time while ensuring the required accuracy. To determine the initial parameters of the model, expressions based on the superposition of reference models in the form of neural networks are proposed. The practical utility of the work lies in the development of an algorithm for the method of reference models for training neural networks with time delays, which can significantly reduce the training time of neural networks without losing model accuracy

Overcoming Dataset Bias: A Deep Learning Approach to Improve Cross-Cultural Image Classification

The difficulties and strategies concerning cross-ethnic image categorization are discussed throughout this report based on deep learning techniques for preventing model bias towards specific ethnic groups. In this part of the project, we continued our discussion of the theory and practice of bias in training datasets and how it impacts the model in the case of emotion recognition and facial analysis using simulations. Three real-time scenarios were explored: public security applications, health status evaluations, education performance, and awareness, all of which must include culturally sensitive data to maximize performance. Some of the issues we defined include the data bias problem, the problem that different cultures exhibit different levels of emotions, and technical concerns such as the limitations of machine learning algorithms, among others that we have described in the paper. Some solutions we provided include data augmentation, real-time learning, and ethical concerns such as using machine learning in making decisions. In doing so, we propose solutions that include building models that are not only precise but also culturally sensitive to avoid prejudice in applications, ranging from credit scoring to complaints handling.

Perencanaan Ulang Bangunan Gedung Dormitory Kawasan Sains dan Teknologi (KST) Nuklir Yogyakarta Berbasis Building Information Modelling (BIM) 5D

Building Information Modeling (BIM) is a method of processing data during the life cycle of a construction project that uses three-dimensional, real-time, and dynamic building modeling software, thereby creating a complete model related to project data from the planning to the construction phase of the project. In the re-planning of the Yogyakarta Nuclear Science and Technology (KST) Dormitory Building based on Building Information Modeling (BIM) 5D, a special moment resisting frame system (SRPMK) structure was used. The planned building consists of 7 floors, including the lower structure consisting of a 50x50 square pile foundation and five types of pile cap, the upper structure consisting of columns, beams, and floor plates. The quality of concrete in the structure used is fc' 40 Mpa and 35 Mpa with reinforcing steel quality fy 420B. In this planning, 2D and 3D modeling was carried out. Structural analysis was carried out using the SAP2000 application, which was then continued with 2D 3D modeling and Quantity take-off results using Revit and Scheduling using Ms. Projects. The Draft Budget (RAB) results for this planning are Rp. 37,691,368,000.00. The planned duration of work is 26 weeks. The results of the scheduling and RAB are then integrated using the Naviswork software so that the simulation results of the work to be planned can be seen.