Search results for: Mindlin’s 2nd gradient model

Authors: Zachary Blanks, Solomon Sonya

Abstract:

Poaching presents a serious threat to endangered animal species, environment conservations, and human life. Additionally, some poaching activity has even been linked to supplying funds to support terrorist networks elsewhere around the world. Consequently, agencies dedicated to protecting wildlife habitats have a near intractable task of adequately patrolling an entire area (spanning several thousand kilometers) given limited resources, funds, and personnel at their disposal. Thus, agencies need predictive tools that are both high-performing and easily implementable by the user to help in learning how the significant features (e.g. animal population densities, topography, behavior patterns of the criminals within the area, etc) interact with each other in hopes of abating poaching. This research develops a classification model using machine learning algorithms to aid in forecasting future attacks that is both easy to train and performs well when compared to other models. In this research, we demonstrate how data imputation methods (specifically predictive mean matching, gradient boosting, and random forest multiple imputation) can be applied to analyze data and create significant predictions across a varied data set. Specifically, we apply these methods to improve the accuracy of adopted prediction models (Logistic Regression, Support Vector Machine, etc). Finally, we assess the performance of the model and the accuracy of our data imputation methods by learning on a real-world data set constituting four years of imputed data and testing on one year of non-imputed data. This paper provides three main contributions. First, we extend work done by the Teamcore and CREATE (Center for Risk and Economic Analysis of Terrorism Events) research group at the University of Southern California (USC) working in conjunction with the Department of Homeland Security to apply game theory and machine learning algorithms to develop more efficient ways of reducing poaching. This research introduces ensemble methods (Random Forests and Stochastic Gradient Boosting) and applies it to real-world poaching data gathered from the Ugandan rain forest park rangers. Next, we consider the effect of data imputation on both the performance of various algorithms and the general accuracy of the method itself when applied to a dependent variable where a large number of observations are missing. Third, we provide an alternate approach to predict the probability of observing poaching both by season and by month. The results from this research are very promising. We conclude that by using Stochastic Gradient Boosting to predict observations for non-commercial poaching by season, we are able to produce statistically equivalent results while being orders of magnitude faster in computation time and complexity. Additionally, when predicting potential poaching incidents by individual month vice entire seasons, boosting techniques produce a mean area under the curve increase of approximately 3% relative to previous prediction schedules by entire seasons.

Keywords: ensemble methods, imputation, machine learning, random forests, statistical analysis, stochastic gradient boosting, wildlife protection

Procedia PDF Downloads 263

Authors: Fadhl Qutaish

Abstract:

This paper examines the pharyngealization spread in one of the Yemeni Arabic dialects, namely, Ibbi Arabic (IA). It investigates how pharyngealized sounds spread their acoustic features onto the neighboring vowels and change their default features. This feature has been investigated quietly well in MSA but still has to be deeply studied in the different dialect of Arabic which will bring about a clearer picture of the similarities and the differences among these dialects and help in mapping them based on the way this feature is utilized. Though the studies are numerous, no one of them has illustrated how far in the multi-syllabic word the spread can be and whether it takes a steady or gradient manner. This study tries to fill this gap and give a satisfactory explanation of the pharyngealization spread in Ibbi Dialect. This study is the first step towards a larger investigation of the different dialects of Yemeni Arabic in the future. The data recorded are represented in minimal pairs in which the trigger (pharyngealized or the non-pharyngealized sound) is in the initial or final position of monosyllabic and multisyllabic words. A group of 24 words were divided into four groups and repeated three times by three subjects which will yield 216 tokens that are tested and analyzed. The subjects are three male speakers aged between 28 and 31 with no history of neurological, speaking or hearing problems. All of them are bilingual speakers of Arabic and English and native speakers of Ibbi-Dialect. Recordings were done in a sound-proof room and praat software was used for the analysis and coding of the trajectories of F1 and F2 for the low vowel /a/ to see the effect of pharyngealization on the formant trajectory within the same syllable and in other syllables of the same word by comparing the F1 and F2 formants to the non-pharyngealized environment. The results show that pharyngealization spread is gradient (progressively and regressively). The spread is reflected in the gradual raising of F1 as we move closer towards the trigger and the gradual lowering of F2 as well. The results of the F1 mean values in tri-syllabic words when the trigger is word initially show that there is a raise of 37.9 HZ in the first syllable, 26.8HZ in the second syllable and 14.2HZ in the third syllable. F2 mean values undergo a lowering of 239 HZ in the first syllable, 211.7 HZ in the second syllable and 176.5 in the third syllable. This gradual decrease in the difference of F2 values in the non-pharyngealized and pharyngealized context illustrates that the spread is gradient. A similar result was found when the trigger is word-final which proves that the spread is gradient (progressively and regressively.

Keywords: pharyngealization, Yemeni Arabic, Ibbi dialect, pharyngealization spread

Procedia PDF Downloads 193

Authors: Mohamad Molavi Nojumi, Xiaodong Wang

Abstract:

In this research fracture behaviour of linear elastic isotropic functionally graded materials (FGMs) are investigated using modified finite element method (FEM). FGMs are advantageous because they enhance the bonding strength of two incompatible materials, and reduce the residual stress and thermal stress. Ceramic/metals are a main type of FGMs. Ceramic materials are brittle. So, there is high possibility of crack existence during fabrication or in-service loading. In addition, damage analysis is necessary for a safe and efficient design. FEM is a strong numerical tool for analyzing complicated problems. Thus, FEM is used to investigate the fracture behaviour of FGMs. Here an accurate 9-node biquadratic quadrilateral graded element is proposed in which the influence of the variation of material properties is considered at the element level. The stiffness matrix of graded elements is obtained using the principle of minimum potential energy. The implementation of graded elements prevents the forced sudden jump of material properties in traditional finite elements for modelling FGMs. Numerical results are verified with existing solutions. Different numerical simulations are carried out to model stationary crack problems in nonhomogeneous plates. In these simulations, material variation is supposed to happen in directions perpendicular and parallel to the crack line. Two special linear and exponential functions have been utilized to model the material gradient as they are mostly discussed in literature. Also, various sizes of the crack length are considered. A major difference in the fracture behaviour of FGMs and homogeneous materials is related to the break of material symmetry. For example, when the material gradation direction is normal to the crack line, even under applying the mode I loading there exists coupled modes I and II of fracture which originates from the induced shear in the model. Therefore, the necessity of the proper modelling of the material variation should be considered in capturing the fracture behaviour of FGMs specially, when the material gradient index is high. Fracture properties such as mode I and mode II stress intensity factors (SIFs), energy release rates, and field variables near the crack tip are investigated and compared with results obtained using conventional homogeneous elements. It is revealed that graded elements provide higher accuracy with less effort in comparison with conventional homogeneous elements.

Keywords: finite element, fracture mechanics, functionally graded materials, graded element

Procedia PDF Downloads 149

Authors: Ritesh Watharkar, Sourabh Chakraborty, Brijesh Srivastava

Abstract:

Reduction of water from the fruits and vegetables using different drying techniques is widely employed to prolong the shelf life of these food commodities. Heat transfer occurs inside the sample by conduction and mass transfer takes place by diffusion in accordance with temperature and moisture concentration gradient respectively during drying. This study was undertaken to study and model the thin layer drying behavior of Bhimkol pulp. The drying was conducted in a tray drier at 500c temperature with 5, 10 and 15 % concentrations of added maltodextrin. The drying experiments were performed at 5mm thickness of the thin layer and the constant air velocity of 0.5 m/s.Drying data were fitted to different thin layer drying models found in the literature. Comparison of fitted models was based on highest R2(0.9917), lowest RMSE (0.03201), and lowest SSE (0.01537) revealed Middle equation as the best-fitted model for thin layer drying with 10% concentration of maltodextrin. The effective diffusivity was estimated based on the solution of Fick’s law of diffusion which is found in the range of 3.0396 x10-09 to 5.0661 x 10-09. There was a reduction in drying time with the addition of maltodextrin as compare to the raw pulp.

Keywords: Bhimkol, diffusivity, maltodextrine, Midilli model

Procedia PDF Downloads 182

Authors: Tejeet Singh, Harmanjit Singh

Abstract:

Creep behavior of thick-walled functionally graded cylinder consisting of AlSiC and subjected to internal pressure and high temperature has been analyzed. The functional relationship between strain rate with stress can be described by the well-known threshold stress based creep law with a stress exponent of five. The effect of imposing non-linear particle gradient on the distribution of creep stresses in the thick-walled functionally graded composite cylinder has been investigated. The study revealed that for the assumed non-linear particle distribution, the radial stress decreases throughout the cylinder, whereas the tangential, axial and effective stresses have averaging effect. The strain rates in the functionally graded composite cylinder could be reduced to significant extent by employing non-linear gradient in the distribution of reinforcement.

Keywords: functionally graded material, pressure, steady state creep, thick-cylinder

Procedia PDF Downloads 450

Authors: Alanoud Moraya Aldalan, Abdulaziz Almaleh

Abstract:

A crucial component of maintaining a customer-oriented business as in the telecom industry is understanding the reasons and factors that lead to customer churn. Competition between telecom companies has greatly increased in recent years. It has become more important to understand customers’ needs in this strong market of telecom industries, especially for those who are looking to turn over their service providers. So, predictive churn is now a mandatory requirement for retaining those customers. Machine learning can be utilized to accomplish this. Churn Prediction has become a very important topic in terms of machine learning classification in the telecommunications industry. Understanding the factors of customer churn and how they behave is very important to building an effective churn prediction model. This paper aims to predict churn and identify factors of customers’ churn based on their past service usage history. Aiming at this objective, the study makes use of feature selection, normalization, and feature engineering. Then, this study compared the performance of four different machine learning algorithms on the Orange dataset: Logistic Regression, Random Forest, Decision Tree, and Gradient Boosting. Evaluation of the performance was conducted by using the F1 score and ROC-AUC. Comparing the results of this study with existing models has proven to produce better results. The results showed the Gradients Boosting with feature selection technique outperformed in this study by achieving a 99% F1-score and 99% AUC, and all other experiments achieved good results as well.

Keywords: machine learning, gradient boosting, logistic regression, churn, random forest, decision tree, ROC, AUC, F1-score

Procedia PDF Downloads 106

Authors: Xiao-Dong Wang, Wei-Mon Yan

Abstract:

The flow field design in the bipolar plates affects the performance of the proton exchange membrane (PEM) fuel cell. This work adopted a combined optimization procedure, including a simplified conjugate-gradient method and a completely three-dimensional, two-phase, non-isothermal fuel cell model, to look for optimal flow field design for a single serpentine fuel cell of size 9×9 mm with five channels. For the direct solution, the two-fluid method was adopted to incorporate the heat effects using energy equations for entire cells. The model assumes that the system is steady; the inlet reactants are ideal gases; the flow is laminar; and the porous layers such as the diffusion layer, catalyst layer and PEM are isotropic. The model includes continuity, momentum and species equations for gaseous species, liquid water transport equations in the channels, gas diffusion layers, and catalyst layers, water transport equation in the membrane, electron and proton transport equations. The Bulter-Volumer equation was used to describe electrochemical reactions in the catalyst layers. The cell output power density Pcell is maximized subjected to an optimal set of channel heights, H1-H5, and channel widths, W2-W5. The basic case with all channel heights and widths set at 1 mm yields a Pcell=7260 Wm-2. The optimal design displays a tapered characteristic for channels 1, 3 and 4, and a diverging characteristic in height for channels 2 and 5, producing a Pcell=8894 Wm-2, about 22.5% increment. The reduced channel heights of channels 2-4 significantly increase the sub-rib convection and widths for effectively removing liquid water and oxygen transport in gas diffusion layer. The final diverging channel minimizes the leakage of fuel to outlet via sub-rib convection from channel 4 to channel 5. Near-optimal design without huge loss in cell performance but is easily manufactured is tested. The use of a straight, final channel of 0.1 mm height has led to 7.37% power loss, while the design with all channel widths to be 1 mm with optimal channel heights obtained above yields only 1.68% loss of current density. The presence of a final, diverging channel has greater impact on cell performance than the fine adjustment of channel width at the simulation conditions set herein studied.

Keywords: optimization, flow field design, simplified conjugate-gradient method, serpentine flow field, sub-rib convection

Procedia PDF Downloads 272

Authors: Huda Algharib, Amal Algharib, Hanan Algharib, Ali Mohammad Alqudah

Abstract:

Image registration is an important topic for many imaging systems and computer vision applications. The standard image registration techniques such as Mutual information/ Normalized mutual information -based methods have a limited performance because they do not consider the spatial information or the relationships between the neighbouring pixels or voxels. In addition, the amount of image noise may significantly affect the registration accuracy. Therefore, this paper proposes an efficient method that explicitly considers the relationships between the adjacent pixels, where the gradient information of the reference and scene images is extracted first, and then the cosine similarity of the extracted gradient information is computed and used to improve the accuracy of the standard normalized mutual information measure. Our experimental results on different data types (i.e. CT, MRI and thermal images) show that the proposed method outperforms a number of image registration techniques in terms of the accuracy.

Keywords: image registration, mutual information, image gradients, image transformations

Procedia PDF Downloads 216

Authors: Entisar A. Elgmati

Abstract:

Recurrent infant diarrhea is studied using daily data collected in Salvador, Brazil over one year and three months. A logistic regression model is fitted instead of Aalen's additive model using the same covariates that were used in the analysis with the additive model. The model gives reasonably similar results to that using additive regression model. In addition, the problem with the estimated conditional probabilities not being constrained between zero and one in additive model is solved here. Also martingale residuals that have been used to judge the goodness of fit for the additive model are shown to be useful for judging the goodness of fit of the logistic model.

Keywords: additive model, cumulative probabilities, infant diarrhoea, recurrent event

Procedia PDF Downloads 605

Authors: Abhigyan Hedau, Priya Shelke, Riddhi Mirajkar, Shreyash Chaple, Mrunali Gadekar, Himanshu Akula

Abstract:

The heart is the coordinating centre of the major endocrine glandular structure of the body, which produces hormones that profoundly affect the operations of the body, and diagnosing cardiovascular disease is a difficult but critical task. By extracting knowledge and information about the disease from patient data, data mining is a more practical technique to help doctors detect disorders. We use a variety of machine learning methods here, including logistic regression and support vector classifiers (SVC), K-nearest neighbours Classifiers (KNN), Decision Tree Classifiers, Random Forest classifiers and Gradient Boosting classifiers. These algorithms are applied to patient data containing 13 different factors to build a system that predicts heart disease in less time with more accuracy.

Keywords: logistic regression, support vector classifier, k-nearest neighbour, decision tree, random forest and gradient boosting

Procedia PDF Downloads 19

Authors: Nosheen Zareen Khan, Abdul Majeed Siddiqui, Muhammad Afzal Rana

Abstract:

The steady flow of a second order fluid through constricted tube with slip velocity at wall is modeled and analyzed theoretically. The governing equations are simplified by implying no slip in radial direction. Based on Karman Pohlhausen procedure polynomial solution for axial velocity profile is presented. An expressions for pressure gradient, shear stress, separation and reattachment points and radial velocity are also calculated. The effect of slip and no slip velocity on velocity, shear stress, pressure gradient are discussed and depicted graphically. It is noted that when Reynolds number increases velocity of the fluid decreases in both slip and no slip conditions. It is also found that the wall shear stress, separation and reattachment points are strongly effected by Reynolds number.

Keywords: approximate solution, constricted tube, non-Newtonian fluids, Reynolds number

Procedia PDF Downloads 366

Authors: Maatoug Hassine, Mourad Hrizi

Abstract:

In this paper, we consider a geometric inverse source problem for the heat equation with Dirichlet and Neumann boundary data. We will reconstruct the exact form of the unknown source term from additional boundary conditions. Our motivation is to detect the location, the size and the shape of source support. We present a one-shot algorithm based on the Kohn-Vogelius formulation and the topological gradient method. The geometric inverse source problem is formulated as a topology optimization one. A topological sensitivity analysis is derived from a source function. Then, we present a non-iterative numerical method for the geometric reconstruction of the source term with unknown support using a level curve of the topological gradient. Finally, we give several examples to show the viability of our presented method.

Keywords: geometric inverse source problem, heat equation, topological optimization, topological sensitivity, Kohn-Vogelius formulation

Procedia PDF Downloads 271

Authors: Hosein Falahaty, Hitoshi Gotoh, Abbas Khayyer

Abstract:

Performance of a Hamiltonian based particle method in simulation of nonlinear structural dynamics is subjected to investigation in terms of stability and accuracy. The governing equation of motion is derived based on Hamilton's principle of least action, while the deformation gradient is obtained according to Weighted Least Square method. The hyper-elasticity models of Saint Venant-Kirchhoff and a compressible version similar to Mooney- Rivlin are engaged for the calculation of second Piola-Kirchhoff stress tensor, respectively. Stability along with accuracy of numerical model is verified by reproducing critical stress fields in static and dynamic responses. As the results, although performance of Hamiltonian based model is evaluated as being acceptable in dealing with intense extensional stress fields, however kinds of instabilities reveal in the case of violent collision which can be most likely attributed to zero energy singular modes.

Keywords: Hamilton's principle of least action, particle-based method, hyper-elasticity, analysis of stability

Procedia PDF Downloads 313

Authors: Armansyah, I. P. Almanar, M. Saiful Bahari Shaari, M. Shamil Jaffarullah

Abstract:

Finite element approach have been used via three-dimensional models by using Altair Hyper Work, a commercially available software, to describe heat gradients along the welding zones (axially and coronaly) in Friction Stir Welding (FSW). Transient thermal finite element analyses are performed in AA 6061-T6 Aluminum Alloy to obtain temperature distribution in the welded aluminum plates during welding operation. Heat input from tool shoulder and tool pin are considered in the model. A moving heat source with a heat distribution simulating the heat generated by frictions between tool shoulder and work piece is used in the analysis. The developed model was then used to show the effect of various input parameters such as total rate of welding speed and rotational speed on temperature distribution in the work piece.

Keywords: Frictions Stir Welding (FSW), temperature distribution, Finite Element Method (FEM), altair hyperwork

Procedia PDF Downloads 502

Authors: Katawut Kaewbanjong

Abstract:

We analyzed a volume of data and found significant user satisfaction in software project factors. A statistical significance analysis (logistic regression) and collinearity analysis determined the significance factors from a group of 71 pre-defined factors from 191 software projects in ISBSG Release 12. The eight prediction models used for testing the prediction potential of these factors were Neural network, k-NN, Naïve Bayes, Random forest, Decision tree, Gradient boosted tree, linear regression and logistic regression prediction model. Fifteen pre-defined factors were truly significant in predicting user satisfaction, and they provided 82.71% prediction accuracy when used with a neural network prediction model. These factors were client-server, personnel changes, total defects delivered, project inactive time, industry sector, application type, development type, how methodology was acquired, development techniques, decision making process, intended market, size estimate approach, size estimate method, cost recording method, and effort estimate method. These findings may benefit software development managers considerably.

Keywords: prediction model, statistical analysis, software project, user satisfaction factor

Procedia PDF Downloads 89

Authors: Martin Cermak, Tomas Karasek, Jaroslav Rojicek

Abstract:

The key role in phenomenological modelling of cyclic plasticity is good understanding of stress-strain behaviour of given material. There are many models describing behaviour of materials using numerous parameters and constants. Combination of individual parameters in those material models significantly determines whether observed and predicted results are in compliance. Parameter identification techniques such as random gradient, genetic algorithm, and sensitivity analysis are used for identification of parameters using numerical modelling and simulation. In this paper genetic algorithm and sensitivity analysis are used to study effect of 4 parameters of modified AbdelKarim-Ohno cyclic plasticity model. Results predicted by Finite Element (FE) simulation are compared with experimental data from biaxial ratcheting test with semi-elliptical loading path.

Keywords: genetic algorithm, sensitivity analysis, inverse approach, finite element method, cyclic plasticity, ratcheting

Procedia PDF Downloads 419

Authors: Kang Lin Peng

Abstract:

Space tourism emerged in 2001 and became famous in 2021, following the development of space technology. The space market is twisted because of the excess demand. Space tourism is currently rare and extremely expensive, with biased luxury product pricing, which is the seller’s market that consumers can not bargain with. Spaceship companies such as Virgin Galactic, Blue Origin, and Space X have been charged space tourism prices from 200 thousand to 55 million depending on various heights in space. There should be a reasonable price based on a fair basis. This study aims to derive a spacetime pricing model, which is different from the general pricing model on the earth’s surface. We apply general relativity theory to deduct the mathematical formula for the space tourism pricing model, which covers the traditional time-independent model. In the future, the price of space travel will be different from current flight travel when space travel is measured in lightyear units. The pricing of general commodities mainly considers the general equilibrium of supply and demand. The pricing model considers risks and returns with the dependent time variable as acceptable when commodities are on the earth’s surface, called flat spacetime. Current economic theories based on the independent time scale in the flat spacetime do not consider the curvature of spacetime. Current flight services flying the height of 6, 12, and 19 kilometers are charging with a pricing model that measures time coordinate independently. However, the emergence of space tourism is flying heights above 100 to 550 kilometers that have enlarged the spacetime curvature, which means tourists will escape from a zero curvature on the earth’s surface to the large curvature of space. Different spacetime spans should be considered in the pricing model of space travel to echo general relativity theory. Intuitively, this spacetime commodity needs to consider changing the spacetime curvature from the earth to space. We can assume the value of each spacetime curvature unit corresponding to the gradient change of each Ricci or energy-momentum tensor. Then we know how much to spend by integrating the spacetime from the earth to space. The concept is adding a price p component corresponding to the general relativity theory. The space travel pricing model degenerates into a time-independent model, which becomes a model of traditional commodity pricing. The contribution is that the deriving of the space tourism pricing model will be a breakthrough in philosophical and practical issues for space travel. The results of the space tourism pricing model extend the traditional time-independent flat spacetime mode. The pricing model embedded spacetime as the general relativity theory can better reflect the rationality and accuracy of space travel on the universal scale. The universal scale from independent-time scale to spacetime scale will bring a brand-new pricing concept for space traveling commodities. Fair and efficient spacetime economics will also bring to humans’ travel when we can travel in lightyear units in the future.

Keywords: space tourism, spacetime pricing model, general relativity theory, spacetime curvature

Procedia PDF Downloads 87

Authors: Eric Pla Erra, Mariana Jimenez Martinez

Abstract:

While aggregated loads at a community level tend to be easier to predict, individual household load forecasting present more challenges with higher volatility and uncertainty. Furthermore, the drastic changes that our behavior patterns have suffered due to the COVID-19 pandemic have modified our daily electrical consumption curves and, therefore, further complicated the forecasting methods used to predict short-term electric load. Load forecasting is vital for the smooth and optimized planning and operation of our electric grids, but it also plays a crucial role for individual domestic consumers that rely on a HEMS (Home Energy Management Systems) to optimize their energy usage through self-generation, storage, or smart appliances management. An accurate forecasting leads to higher energy savings and overall energy efficiency of the household when paired with a proper HEMS. In order to study how COVID-19 has affected the accuracy of forecasting methods, an evaluation of the performance of a state-of-the-art LGBM (Light Gradient Boosting Model) will be conducted during the transition between pre-pandemic and lockdowns periods, considering day-ahead electric load forecasting. LGBM improves the capabilities of standard Decision Tree models in both speed and reduction of memory consumption, but it still offers a high accuracy. Even though LGBM has complex non-linear modelling capabilities, it has proven to be a competitive method under challenging forecasting scenarios such as short series, heterogeneous series, or data patterns with minimal prior knowledge. An adaptation of the LGBM model – called “resilient LGBM” – will be also tested, incorporating a concept drift detection technique for time series analysis, with the purpose to evaluate its capabilities to improve the model’s accuracy during extreme events such as COVID-19 lockdowns. The results for the LGBM and resilient LGBM will be compared using standard RMSE (Root Mean Squared Error) as the main performance metric. The models’ performance will be evaluated over a set of real households’ hourly electricity consumption data measured before and during the COVID-19 pandemic. All households are located in the city of Barcelona, Spain, and present different consumption profiles. This study is carried out under the ComMit-20 project, financed by AGAUR (Agència de Gestiód’AjutsUniversitaris), which aims to determine the short and long-term impacts of the COVID-19 pandemic on building energy consumption, incrementing the resilience of electrical systems through the use of tools such as HEMS and artificial intelligence.

Keywords: concept drift, forecasting, home energy management system (HEMS), light gradient boosting model (LGBM)

Procedia PDF Downloads 80

Authors: Esther Chiramal, Kelvin Soh Boon Kai

Abstract:

Explainable AI is a strong strategy implemented to understand complex black-box model predictions in a human-interpretable language. It provides the evidence required to execute the use of trustworthy and reliable AI systems. On the other hand, however, it also opens the door to locating possible vulnerabilities in an AI model. Traditional adversarial text attack uses word substitution, data augmentation techniques, and gradient-based attacks on powerful pre-trained Bidirectional Encoder Representations from Transformers (BERT) variants to generate adversarial sentences. These attacks are generally white-box in nature and not practical as they can be easily detected by humans e.g., Changing the word from “Poor” to “Rich”. We proposed a simple yet effective Grey-box cum Black-box approach that does not require the knowledge of the model while using a set of surrogate Transformer/BERT models to perform the attack using Explainable AI techniques. As Transformers are the current state-of-the-art models for almost all Natural Language Processing (NLP) tasks, an attack generated from BERT1 is transferable to BERT2. This transferability is made possible due to the attention mechanism in the transformer that allows the model to capture long-range dependencies in a sequence. Using the power of BERT generalisation via attention, we attempt to exploit how transformers learn by attacking a few surrogate transformer variants which are all based on a different architecture. We demonstrate that this approach is highly effective to generate semantically good sentences by changing as little as one word that is not detectable by humans while still fooling other BERT models.

Keywords: BERT, explainable AI, Grey-box text attack, transformer

Procedia PDF Downloads 112

Authors: Jurriaan Gillissen

Abstract:

This work is concerned with stabilizing the numerical integration of the Navier-Stokes equation (NSE), backwards in time. Applications involve the detection of sources of, e.g., sound, heat, and pollutants. Stable reverse numerical integration of parabolic differential equations is also relevant for image de-blurring. While the literature addresses the reverse integration problem of the advection-diffusion equation, the problem of numerical reverse integration of the NSE has, to our knowledge, not yet been addressed. Owing to the presence of viscosity, the NSE is irreversible, i.e., when going backwards in time, the fluid behaves, as if it had a negative viscosity. As an effect, perturbations from the perfect solution, due to round off errors or discretization errors, grow exponentially in time, and reverse integration of the NSE is inherently unstable, regardless of using an implicit time integration scheme. Consequently, some sort of filtering is required, in order to achieve a stable, numerical, reversed integration. The challenge is to find a filter with a minimal adverse affect on the accuracy of the reversed integration. In the present work, we explore an adjoint gradient method (AGM) to achieve this goal, and we apply this technique to two-dimensional (2D), decaying turbulence. The AGM solves for the initial velocity field u0 at t = 0, that, when integrated forward in time, produces a final velocity field u1 at t = 1, that is as close as is feasibly possible to some specified target field v1. The initial field u0 defines a minimum of a cost-functional J, that measures the distance between u1 and v1. In the minimization procedure, the u0 is updated iteratively along the gradient of J w.r.t. u0, where the gradient is obtained by transporting J backwards in time from t = 1 to t = 0, using the adjoint NSE. The AGM thus effectively replaces the backward integration by multiple forward and backward adjoint integrations. Since the viscosity is negative in the adjoint NSE, each step of the AGM is numerically stable. Nevertheless, when applied to turbulence, the AGM develops instabilities, which limit the backward integration to small times. This is due to the exponential divergence of phase space trajectories in turbulent flow, which produces a multitude of local minima in J, when the integration time is large. As an effect, the AGM may select unphysical, noisy initial conditions. In order to improve this situation, we propose two remedies. First, we replace the integration by a sequence of smaller integrations, i.e., we divide the integration time into segments, where in each segment the target field v1 is taken as the initial field u0 from the previous segment. Second, we add an additional term (regularizer) to J, which is proportional to a high-order Laplacian of u0, and which dampens the gradients of u0. We show that suitable values for the segment size and for the regularizer, allow a stable reverse integration of 2D decaying turbulence, with accurate results for more then O(10) turbulent, integral time scales.

Keywords: time reversed integration, parabolic differential equations, adjoint gradient method, two dimensional turbulence

Procedia PDF Downloads 199

Authors: V. Radulescu, S. Dumitru

Abstract:

Introduction: Determination of the temperature field inside a fluid in motion has many practical issues, especially in the case of turbulent flow. The phenomenon is greater when the solid walls have a different temperature than the fluid. The turbulent heat and mass transfer have an essential role in case of the thermal pollution, as it was the recorded during the damage of the Thermoelectric Power-plant Oradea (closed even today). Basic Methods: Solving the theoretical turbulent thermal pollution represents a particularly difficult problem. By using the semi-empirical theories or by simplifying the made assumptions, based on the experimental measurements may be assured the elaboration of the mathematical model for further numerical simulations. The three zones of flow are analyzed separately: the vicinity of the solid wall, the turbulent transition zone, and the turbulent core. For each area are determined the distribution law of temperature. It is determined the dependence of between the Stanton and Prandtl numbers with correction factors, based on measurements experimental. Major Findings/Results: The limitation of the laminar thermal substrate was determined based on the theory of Landau and Levice, using the assumption that the longitudinal component of the velocity pulsation and the pulsation’s frequency varies proportionally with the distance to the wall. For the calculation of the average temperature, the formula is used a similar solution as for the velocity, by an analogous mediation. On these assumptions, the numerical modeling was performed with a gradient of temperature for the turbulent flow in pipes (intact or damaged, with cracks) having 4 different diameters, between 200-500 mm, as there were in the Thermoelectric Power-plant Oradea. Conclusions: It was made a superposition between the molecular viscosity and the turbulent one, followed by addition between the molecular and the turbulent transfer coefficients, necessary to elaborate the theoretical and the numerical modeling. The concept of laminar boundary layer has a different thickness when it is compared the flow with heat transfer and that one without a temperature gradient. The obtained results are within the margin of error of 5%, between the semi-empirical classical theories and the developed model, based on the experimental data. Finally, it is obtained a general correlation between the Stanton number and the Prandtl number, for a specific flow (with associated Reynolds number).

Keywords: experimental measurements, numerical correlations, thermal pollution through pipelines, turbulent thermal flow

Procedia PDF Downloads 133

Authors: Muneeb Ullah, Daishihan, Xiadong Young

Abstract:

Deep learning plays a crucial role in identifying COVID-19 and preventing its spread. To improve the accuracy of COVID-19 diagnoses, it is important to have access to a sufficient number of training images of CXRs (chest X-rays) depicting the disease. However, there is currently a shortage of such images. To address this issue, this paper introduces COVID-19 GAN, a model that uses generative adversarial networks (GANs) to generate realistic CXR images of COVID-19, which can be used to train identification models. Initially, a generator model is created that uses digressive channels to generate images of CXR scans for COVID-19. To differentiate between real and fake disease images, an efficient discriminator is developed by combining the dense connectivity strategy and instance normalization. This approach makes use of their feature extraction capabilities on CXR hazy areas. Lastly, the deep regret gradient penalty technique is utilized to ensure stable training of the model. With the use of 4,062 grape leaf disease images, the Leaf GAN model successfully produces 8,124 COVID-19 CXR images. The COVID-19 GAN model produces COVID-19 CXR images that outperform DCGAN and WGAN in terms of the Fréchet inception distance. Experimental findings suggest that the COVID-19 GAN-generated CXR images possess noticeable haziness, offering a promising approach to address the limited training data available for COVID-19 model training. When the dataset was expanded, CNN-based classification models outperformed other models, yielding higher accuracy rates than those of the initial dataset and other augmentation techniques. Among these models, ImagNet exhibited the best recognition accuracy of 99.70% on the testing set. These findings suggest that the proposed augmentation method is a solution to address overfitting issues in disease identification and can enhance identification accuracy effectively.

Keywords: classification, deep learning, medical images, CXR, GAN.

Procedia PDF Downloads 52

Authors: Abdull Halim Abdull, Nasiman Sapari, Mohammad Haikal Asyraf Bin Anuar

Abstract:

Although soil thermal properties are required in many areas to improve oil recovery, they are seldom measured on a routine basis. Reasons for this are unclear, but may be related to a lack of suitable instrumentation and entropy theory. We integrate single probe thermal gradient for the radial conduction of a short-duration heat pulse away from a single electrode source, and compared it with the theory for an instantaneously heated line source. By measuring the temperature response at a short distance from the line source, and applying short-duration heat-pulse theory, we can extract all the entropy properties, the thermal diffusivity, heat capacity, and conductivity, from a single heat-pulse measurement. Results of initial experiments carried out on air-dry sand and clay materials indicate that this heat-pulse method yields soil thermal properties that compare well with thermal properties measured by single electrode.

Keywords: entropy, single probe thermal gradient, soil thermal, probe heat

Procedia PDF Downloads 414

Authors: Arijit Chakraborty, Anirban Roy

Abstract:

The Indian peninsula is strategically located in the Asian subcontinent with the Himalayas to the North and Oceans surrounding the other three directions with annual monsoons which takes care of water supply to the rivers. The total river water discharge into the Bay of Bengal and the Arabian Sea is 628 km³/year and 274 km³/year, respectively. Thus huge volumes of fresh water meet saline water, and this mixing of two streams of dissimilar salinity gives rise to tremendous mixing energies which can be harvested for various purposes like energy generation using pressure retarded osmosis or reverse electrodialysis. The present paper concentrates on analyzing the energy of mixing for the rivers in Goa. Goa has 10 rivers of various sizes all which meet the Arabian Sea. In the present work, the 8 rivers and their salinity (NaCl concentrations) have been analyzed along with their seasonal fluctuations. Next, a Gibbs free energy formulation has been implemented to analyze the energy of mixing of the selected rivers. The highest and lowest energies according to the seasonal fluctuations have been evaluated, and this provides two important insights into (i) amount of energy that can be harvested and (ii) decision on the location of such systems.

Keywords: Gibbs energy, mixing energy, salinity gradient energy, thermodynamics

Procedia PDF Downloads 183

Authors: Yash Bingi, Yiqiao Yin

Abstract:

Reduction of child mortality is an ongoing struggle and a commonly used factor in determining progress in the medical field. The under-5 mortality number is around 5 million around the world, with many of the deaths being preventable. In light of this issue, Cardiotocograms (CTGs) have emerged as a leading tool to determine fetal health. By using ultrasound pulses and reading the responses, CTGs help healthcare professionals assess the overall health of the fetus to determine the risk of child mortality. However, interpreting the results of the CTGs is time-consuming and inefficient, especially in underdeveloped areas where an expert obstetrician is hard to come by. Using a support vector machine (SVM) and oversampling, this paper proposed a model that classifies fetal health with an accuracy of 99.59%. To further explain the CTG measurements, an algorithm based on Randomized Input Sampling for Explanation ((RISE) of Black-box Models was created, called Feature Alteration for explanation of Black Box Models (FAB), and compared the findings to Shapley Additive Explanations (SHAP) and Local Interpretable Model Agnostic Explanations (LIME). This allows doctors and medical professionals to classify fetal health with high accuracy and determine which features were most influential in the process.

Keywords: machine learning, fetal health, gradient boosting, support vector machine, Shapley values, local interpretable model agnostic explanations

Procedia PDF Downloads 114

Authors: Namita Mittal, Himakshi Shekhawat, Ankit Vidyarthi

Abstract:

In medical research study, doctors and radiologists face lot of complexities in analysing the brain tumors in Magnetic Resonance (MR) images. Brain tumor detection is difficult due to amorphous tumor shape and overlapping of similar tissues in nearby region. So, radiologists require one such clinically viable solution which helps in automatic segmentation of tumor inside brain MR image. Initially, segmentation methods were used to detect tumor, by dividing the image into segments but causes loss of information. In this paper, a hybrid method is proposed which detect Region of Interest (ROI) on the basis of difference in intensity values and texture values of tumor region using nearby tissues with Gradient Vector Flow (GVF) technique in the identification of ROI. Proposed approach uses both intensity and texture values for identification of abnormal section of the brain MR images. Experimental results show that proposed method outperforms GVF method without any loss of information.

Keywords: brain tumor, GVF, intensity, MR images, segmentation, texture

Procedia PDF Downloads 403

Authors: Mehmet Izzettin Yilmazer, Emin Cadirli

Abstract:

Zn-7wt.%Al-2.96wt.%Cu eutectic alloy was directionally solidified upwards with different temperature gradients (from 6.70 K/mm to 10.67 K/mm) at a constant growth rate (16.4 Km/s) and also different growth rate (from 8.3 micron/s to 166 micron/s) at a constant temperature gradient (10.67 K/mm) using a Bridgman–type growth apparatus.The values of eutectic spacing were measured from longitudinal and transverse sections of the samples. The dependency of microstructures on the G and V were determined with linear regression analysis and experimental equations were found as λl=8.953xVexp-0.49, λt=5.942xVexp-0.42 and λl=0.008xGexp-1.23, λt=0.024xGexp-0.93. The measurements of microhardness of directionally solidified samples were obtained by using a microhardness test device. The dependence of microhardness HV on temperature gradient and growth rate were analyzed. The dependency of microhardness on the G and V were also determined with linear regression analysis as HVl=110.66xVexp0.02, HVt=111.94xVexp0.02 and HVl=69.66xGexp0.17, HVt=68.86xGexp0.18. The experimental results show that the microhardness of the directionally solidified Zn-Al-Cu alloy increases with increasing the growth rate. The results obtained in this work were compared with the previous similar experimental results.

Keywords: directional solidification, eutectic alloys, microstructure, microhardness

Procedia PDF Downloads 427

Authors: Zhen Cao, Yu Zhu, Junxue Fu

Abstract:

Immunoassay, i.e., antigen-antibody reaction, is crucial for disease diagnostics. To achieve the adequate signal of the antigen protein detection, a large amount of sample and long incubation time is needed. However, the amount of protein is usually small at the early stage, which makes it difficult to detect. Unlike cells and DNAs, no valid chemical method exists for protein amplification. Thus, an alternative way to improve the signal is through particle manipulation techniques to concentrate proteins, among which dielectrophoresis (DEP) is an effective one. DEP is a technique that concentrates particles to the designated region through a force created by the gradient in a non-uniform electric field. Since DEP force is proportional to the cube of particle size and square of electric field gradient, it is relatively easy to capture larger particles such as cells. For smaller ones like proteins, a super high gradient is then required. In this work, three-dimensional Ag/SiO2 nanorods arrays, fabricated by an easy physical vapor deposition technique called as oblique angle deposition, have been integrated with a DEP device and created the field gradient as high as of 2.6×10²⁴ V²/m³. The nanorods based DEP device is able to enrich bovine serum albumin (BSA) protein by 1800-fold and the rate has reached 180-fold/s when only applying 5 V electric potential. Based on the above nanorods integrated DEP platform, an immunoassay of mouse immunoglobulin G (IgG) proteins has been performed. Briefly, specific antibodies are immobilized onto nanorods, then IgG proteins are concentrated and captured, and finally, the signal from fluorescence-labelled antibodies are detected. The limit of detection (LoD) is measured as 275.3 fg/mL (~1.8 fM), which is a 20,000-fold enhancement compared with identical assays performed on blank glass plates. Further, prostate-specific antigen (PSA), which is a cancer biomarker for diagnosis of prostate cancer after radical prostatectomy, is also quantified with a LoD as low as 2.6 pg/mL. The time to signal saturation has been significantly reduced to one minute. In summary, together with an easy nanorod fabrication and integration method, this nanorods based DEP platform has demonstrated highly sensitive immunoassay performance and thus poses great potentials in applications for early point-of-care diagnostics.

Keywords: dielectrophoresis, immunoassay, oblique angle deposition, protein concentration

Procedia PDF Downloads 80

Authors: Shantosa Yudha Siswanto, Julianto Arief Ismail, Rachmat Harryanto

Abstract:

The research was conducted with the aim to know the effect of slope steepness on organic carbon and soil erodibility as erosion factor. This research was conducted from September to December 2011 in the Raharja and Cinanjung Village, Tanjungsari, Sumedang District, West Java, Indonesia. The study was carried out using physiographic free survey method, which is a survey based on land physiographic appearance. Soil sampling was carried out into transect on the similarity slope without calculating the point of observation range. Soil sampling was carried onto three classes of slope as follows: 8–15%, 15–25% and 25–40%. Each was consisted of three slope position i.e. top slope, middle slope and down slope and four samples of soil were taken from each of them, hence it resulted in 36 points of observation. The results of this study indicate that gradient of slope have some significant contribution in every sample. Middle slope with gradient 26-40% has the highest potential erosion occurrence. It has organic C content (0.84%) and the highest erodibility value (0.1092).

Keywords: slope steepness, erosion, erodibility, erosion factor

Procedia PDF Downloads 378

Authors: Z. Mortezaie, H. Hassanpour, S. Asadi Amiri

Abstract:

Captured images may suffer from Gaussian blur due to poor lens focus or camera motion. Unsharp masking is a simple and effective technique to boost the image contrast and to improve digital images suffering from Gaussian blur. The technique is based on sharpening object edges by appending the scaled high-frequency components of the image to the original. The quality of the enhanced image is highly dependent on the characteristics of both the high-frequency components and the scaling/gain factor. Since the quality of an image may not be the same throughout, we propose an adaptive unsharp masking method in this paper. In this method, the gain factor is computed, considering the gradient variations, for individual pixels of the image. Subjective and objective image quality assessments are used to compare the performance of the proposed method both with the classic and the recently developed unsharp masking methods. The experimental results show that the proposed method has a better performance in comparison to the other existing methods.

Keywords: unsharp masking, blur image, sub-region gradient, image enhancement

Procedia PDF Downloads 188