A laser 2-focus velocimeter (L2F) has been applied for measurements of velocity and size of droplets in the dense region of diesel fuel spray under the atmospheric pressure condition. The fuel mass distribution in the spray obtained by the measurement was compared with the one simulated by using 3D-CFD code KIVA-3V. The radial velocity of spray droplets near the injector nozzle in the KIVA-3V code was changed to reproduce the measured fuel mass distribution. The combustion simulations of tetradecane (C14H30) and Et20, which consists of 80% tetradecane and 20% ethanol by mass, were conducted by using both original and modified KIVA-3V codes. It was found that the simulation results of heat release rate and ignition delay period remained almost unchanged even though the mass distribution was changed. However, the predicted amount of soot formation simulated by using original KIVA-3V code was different from that simulated by using modified KIVA-3V code in both of tetradecane and Et20 combustion simulations.
In construction industry, before shipping manufactured products, it is necessary to attach multiple stickers which are specified by each country. Since required stickers depend on the country, there are so many patterns for the stickers. Usually, we visually check whether the specified stickers are attached or not, which is inefficient and subject to human errors. In order to solve these problems, we may resort to automatically recognize the stickers. Recently, a deep learning model called SSD has achieved high recognition rates in object detection. However, since the recognition accuracy depends on the image data to be trained, it is not that easy to create training data for the stickers. The stickers on construction machines are all rectangles. Utilizing this property of stickers, we develop recognition systems. Our proposed methods achieve higher recognition rates than the conventional SSD.