Search

Although machine learning models have been employed for the compressive strength (CS) of cement-based mortar containing metakaolin, it is difficult to understand how they work due to “black-box” nature. In order to explain the involved mechanism, Categorical Gradient Boosting (CatBoost) model with feature importance, feature interaction, partial dependence plot (PDP) and SHapley Additive exPlanations (SHAP) is proposed in this paper. A dataset consisting of 424 samples with six input variables is used to build the CatBoost model, which has optimal performance by tuning a set of seven hyper-parameters using sequential model-based optimization. Five quantitative measures (R2, MAE, RMSE, a10-, a20-index) are employed to evaluate the accuracy and the obtained results are superior to the previous study. It is from feature importance that the most significant input variable involving the CS is water-to-binder ratio, followed by age of specimen and cement grade. The strongest feature interaction is between water-to-binder ratio and metakaolin. A comprehensive parametric study is carried out via SHAP and PDP to investigate the effects of all input variables on the CS of cement-based mortar. WOS:000860590400005 Q1

Laminar fluid-particle flows in bend geometries are present in many industrial, pharmaceutical, and biomedical applications. Particle deposition has been studied extensively; however, little attention has been paid to the effect of particle sedimentation on particle position and deposition in different pipe geometry combinations. This study presented a comprehensive analysis of sedimentation effects on particle flow behaviour in 90{\deg} circular pipe bends of micron particles in laminar pipe flows. Pipe geometry combinations consisted of eight pipe diameters, nine bend radii, and 30 particle diameters in a range of 1 to 100 {\mu}m. The results demonstrated the locations of particles that sedimented to the bottom half of the straight pipe section, and the particle positions upstream from the pipe bend entrance, which was no longer in a fully developed profile. These new locations represent the effects of gravity, pulling the particles down. While obtaining these positions can be found through CFD analysis, we proposed an analytical solution to predict the particle trajectory from different release locations that would help to identify the initial particle distribution at locations upstream to the bend, to obviate the need for long upstream straight pipe sections in the CFD analysis., Comment: In-Press

No description supplied

No description supplied

This paper presents the process of implementing numerical simulation for a tri-copter in forward flight using the open source code OpenFOAM with the library of Virtual Blade Model (VBM). The tri-copter has 0.75-meter radius and 15-Newton weight. Its propulsion system consists of three two-blade propellers typed XOAR PJP-T-L 1245 with12-inch diameter and 4.5-inch pitch. In the first part, the aerodynamic performance of the tri-copter’s frame (not including the propellers) is investigated at the range of angle of attack from -12 degree to 12 degree by simulation in OpenFOAM. Secondly, these force coefficients are exploited in the analytical approach based on the Blade Element Theory (BET) to determine the rotor disk plane angle and the tri-copter’s frame angle for steady, level, forward flight at 15-m/s. Finally, after obtaining the relative angle of -2 degree between the rotor tip-path-plane and the tri-copter’s frame, the simulation for the tri-copter in forward flight including three rotor disks rotating at 4732-RPM will be carried out through the rotorDisksource library of the open source code OpenFOAM. A steady, incompressible solver with k-ω SST turbulence model is applied in the simple Foam algorithm. The aerodynamic forces and the dynamic fields surrounding the tri-copter will be discussed in detail, while the effect of tri-copter’s frame on rotor operation will also be considered. Moreover, the results of thrust from each propeller from simulation are compared to that of analytical BET method and that of propeller performance provided by the manufacturer. This comparison proves rational of using virtual disk to replace real propeller in the CFD simulation.

It is a big challenge to design mixture proportions of the high-performance concrete due to highly nonhomogeneous relationships and coherent among many components. Although machine learning (ML) algorithms have been employed effectively to solve this problem, they are black box models and do not show an explicit relation between the compressive strength and mixture proportions. In order to overcome this inherent weakness, this paper proposes general semi-empirical formulas involving nondimensionalization and optimisation techniques. The optimisation process employs the Nelder–Mead simplex algorithm and takes into account the behavior of uncertain variables, which may occur in experimental data. Successful compressive strength predictions of five datasets with high accuracy in compared to available ML models indicate that the proposed framework has the universal capacity, which can be used for various datasets. Furthermore, the explicit relation of semi-empirical formulas may be a useful tool for engineers and researchers in this area for the prediction purposes.

This work presents an approach to solve inverse problems in the application of water quality management in reservoir systems. One such application is contaminant cleanup, which is challenging because tasks such as inferring the contaminant location and its distribution require large computational efforts and data storage requirements. In addition, real systems contain uncertain parameters such as wind velocity; these uncertainties must be accounted for in the inference problem. The approach developed here uses the combination of a reduced-order model and a Bayesian inference formulation to rapidly determine contaminant locations given sparse measurements of contaminant concentration. The system is modelled by the coupled Navier-Stokes equations and convection-diffusion transport equations. The Galerkin finite element method provides an approximate numerical solution-the ’full model’, which cannot be solved in real-time. The proper orthogonal decomposition and Galerkin projection technique are applied to obtain a reduced-order model that approximates the full model. The Bayesian formulation of the inverse problem is solved using a Markov chain Monte Carlo method for a variety of source locations in the domain. Numerical results show that applying the reduced-order model to the source inversion problem yields a speed-up in computational time by a factor of approximately 32 with acceptable accuracy in comparison with the full model. Application of the inference strategy shows the potential effectiveness of this computational modeling approach for managing water quality.

We propose a robust semi-auto system of labeling faces for characters in video by combining face detection, tracking, and recognition. At the very first step, our system detects the faces automatically in each video frame. After that, the face detection responses will be linked together to form face sequences (raw tracklets) in each video shot by employing simple temporal-spatial constraints to associate. Then we apply a tracking algorithm to not only extend those raw tracklets bi-directionally to cover more appearance views of the object instead of focusing only the frontal view, but also to help fixing the “gaps” caused by missed detection. After being merged among the potential overlapped ones, in the next step, these extended non-overlapped face tracklets across the video are associated with each other by our proposed Heuristics clustering algorithm. In order to achieve high accuracy, we use both generative and discriminative appearance models of the faces and also the context information, which is the clothing color feature in our case. An extensive experiment is performed on approximately 3.5 hours of videos cut from two TV series “Friends”, “How I Met Your Mother” to show the robustness of our system.

This work presents an approach to solve stochastic optimal control problems in the application of flow quality management in reservoir systems. These applications are challenging because they require real-time decision-making in the presence of uncertainties such as wind velocity. These uncertainties must be accounted for as stochastic variables in the mathematical model. In addition, computational costs and storage requirements increase rapidly due to the stochastic nature of the simulations and optimisation formulations. To overcome these challenges, an approach is developed that uses the combination of a reduced-order model and an adjoint-based method to compute the optimal solution rapidly. The system is modelled by a system of stochastic partial differential equations. The finite element method together with collocation in the stochastic space provide an approximate numerical solution—the “full model”, which cannot be solved in real-time. The proper orthogonal decomposition and Galerkin projection technique are applied to obtain a reduced-order model that approximates the full model. The conjugate-gradient method with Armijo line-search is then employed to find the solution of the optimal control problem under the uncertainty of input parameters. Numerical results show that the stochastic control yields solutions that are above the bound of the set solutions of the deterministic control. Applying the reduced model to the stochastic optimal control problem yields a speed-up in computational time by a factor of about 80 with acceptable accuracy in comparison with the full model. Application of the optimal control strategy shows the potential effectiveness of this computational modeling approach for managing flow quality.Copyright © 2012 by ASME