Predict Mean Diameter Research Articles

GRAINet: mapping grain size distributions in river beds from UAV images with convolutional neural networks

Abstract. Grain size analysis is the key to understand the sediment dynamics of river systems. We propose GRAINet, a data-driven approach to analyze grain size distributions of entire gravel bars based on georeferenced UAV images. A convolutional neural network is trained to regress grain size distributions as well as the characteristic mean diameter from raw images. GRAINet allows for the holistic analysis of entire gravel bars, resulting in (i) high-resolution estimates and maps of the spatial grain size distribution at large scale and (ii) robust grading curves for entire gravel bars. To collect an extensive training dataset of 1491 samples, we introduce digital line sampling as a new annotation strategy. Our evaluation on 25 gravel bars along six different rivers in Switzerland yields high accuracy: the resulting maps of mean diameters have a mean absolute error (MAE) of 1.1 cm, with no bias. Robust grading curves for entire gravel bars can be extracted if representative training data are available. At the gravel bar level the MAE of the predicted mean diameter is even reduced to 0.3 cm, for bars with mean diameters ranging from 1.3 to 29.3 cm. Extensive experiments were carried out to study the quality of the digital line samples, the generalization capability of GRAINet to new locations, the model performance with respect to human labeling noise, the limitations of the current model, and the potential of GRAINet to analyze images with low resolutions.

Modeling and analysis of SiO2 deposition during high-purity fused silica glass synthesis by SiCl4 chemical vapor deposition

Large-size and high-purity fused silica glass fabricated by chemical vapor deposition (CVD) has been widely applied to many advanced optical systems. The deposition process is critical during fused silica glass synthesis. In this study, SiO2 deposition at the substrate is modelled and analyzed. First, numerical simulations are performed with the computational fluid dynamics method to predict the flame feature. The gas temperature and velocity agree well with the experimental data. Then, the predicted distributions of gas temperature and concentrations are used as the input parameters to a two-dimensional soot dynamic model. The particle characteristics and particle deposition at the substrate are analyzed. It is found that the particle number density first increases and then decreases in the flame along the axial and radial directions. The predicted mean diameter of the particle is smaller than 50 nm near the substrate surface. The mechanism for SiO2 deposition is different along the substrate surface. In the central region of substrate, the particle deposition is dominant, while in the exterior region the surface chemical deposition is more important.

A rapid urban site index for assessing the quality of street tree planting sites

Urban trees experience site-induced stress and this leads to reduced growth and health. A site assessment tool would be useful for urban forest managers to better match species tolerances and site qualities, and to assess the efficacy of soil management actions. Toward this goal, a rapid urban site index (RUSI) model was created and tested for its ability to predict urban tree performance. The RUSI model is field-based assessment tool that scores 15 parameters in approximately five minutes. This research was conducted in eight cities throughout the Midwest and Northeast USA to test the efficacy of the RUSI model. The RUSI model accurately predicted urban tree health and growth metrics (P<0.0001; R2 0.18–0.40). While the RUSI model did not accurately predict mean diameter growth, it was significantly correlated with recent diameter growth. Certain parameters in the RUSI model, such as estimated rooting area, soil structure and aggregate stability appeared to be more important than other parameters, such as growing degree days. Minimal improvements in the RUSI model were achieved by adding soil laboratory analyses. Field assessments in the RUSI model were significantly correlated with similar laboratory analyses. Other users may be able to use the RUSI model to assess urban tree planting sites (<5min per site and no laboratory analyses fee), but training will be required to accurately utilize the model. Future work on the RUSI model will include developing training modules and testing across a wider geographic area with more urban tree species and urban sites.

Predicting distributional change of forest cover and volume in future climate of South Korea

This study was performed to estimate the forest cover and volume change using the potential forest cover map and National Forest Inventory (NFI) data of South Korea. The regression models were developed to predict mean diameter at breast height (DBH), tree height (h) and number of trees (Nha) for Pinus densifloraroman Pinus koraiensisroman Pinus rigidaroman Larix kaempferiroman Castanea crenata and Quercus spp. stand using NFI data. The second step was to prepare potential forest cover maps after 50 and 100 years using the Hydrological and Thermal Analogy Groups (HyTAGs), and then we compared the produced map with the present forest cover map. For the area where forest cover is changed after 50 years, therefore, the volume could be calculated using regression models and DBH and tree height estimated by “age class 1” in this study. On the other hand, the volume of unchanged forest area could be predicted with DBH k+5 and Nhak+5 adding age class 5 (50 years) to the present age class k on the forest cover map. The forest volume after 100 years was also calculated with the same process of after-50 years. As a result, it was predicted that the forest cover would be changed rapidly into Quercus spp. forests for the coming 100 years, accompanying the diminution of coniferous forest. The forest volume would dramatically decrease by 358,719,160 m3 for the coming 50 years because about 80% of coniferous forests are changed into young forests of Quercus spp. by climate change. The forest volume after 100 years would increase to 315,810,920 m3 due to the growth of young forest during 50 years. However, it should be noted that the change of forest cover and volume was estimated without the consideration of mortality, thinning, and tree planting in this study.

Numerical study on bouncing and separation collision between two droplets considering the collision-induced breakup

The main purpose of the present study is to perform numerical study on bouncing and separation collision between two droplets considering the collision-induced breakup. In this study, the collision model proposed in our previous study is used for simulation of collision-induced breakup, and we modify this model to consider the effect of liquid property on the behavior of droplet-droplet collision. This collision model is based on the conservation laws for mass, momentum, and energy between before and after collision and provides several formulae for post-collision characteristics of colliding droplets and satellite droplets. Improving the accuracy of the model, in this study, appreciate criterion for bouncing collision is added and dissipation energy during collision process is newly modeled. To validate the new model, numerical calculations are performed and their results are compared with experimental data published earlier for binary collisions of water, propanol, and tetradecane droplets. It is found from the results that the new model shows good agreement with experimental data for the number of satellite droplets. It can be also shown that the predicted mean diameter by the new model decrease with increasing the Weber number because of the collisioninduced breakup, whereas the O’Rourke model fails to predict the size reduction via the binary droplet collision.

Analysis of turbulence in a micro-channel emulsifier

Turbulent flow of water in an 1 mm long and 0.4 mm high gap of an emulsifier is investigated experimentally using micro-PIV technique and compared with numerical predictions. The purpose of the investigations is to develop a procedure for well-controlled generation of mono-disperse suspension of micro-droplets. The micro-flow measurements are based on epi-fluorescence illumination and high-speed imaging. The experimental data are compared with the numerical results obtained using direct simulation and turbulent flow model. From the turbulent modelling ( k − ε ) it was found that, the maximum of turbulent energy dissipation rate takes place mainly in the micro-channel gap. However, measured and directly simulated (DNS) velocity fields indicate that flow turbulization is delayed and develops shortly behind the gap. Experimental and numerical results are used to predict droplet dimension of two different oil emulsions. Very low amount of oil-phase fraction in investigated emulsions justifies us to use flow characteristics evaluated for pure water to predict mean diameter of oil droplets. These predictions are validated using experimental data on statistics of droplets size.

Modeling of droplet collision-induced breakup process

The present article proposes a new droplet collision model considering droplet collision-induced breakup process with the formation of satellite droplets. The new model consists of several equations to investigate the post-collision characteristics of colliding droplets and satellite droplets. These equations are derived from the conservations of droplet mass, momentum, and energy between before and after collision, and make it possible to predict the number of satellite droplets, and the droplet size and velocity in the analytical way. To validate the new collision model, numerical calculations are performed and their results are compared with experimental data published earlier for binary collision of water droplets. It is found from the results that the new model shows good agreement with experimental data for the number of satellite droplets. It can be also shown that the predicted mean diameter by the new model decrease with increasing the Weber number because of the collision-induced breakup, whereas the O’Rourke model fails to predict the size reduction via the binary droplet collision.