Parametric Optimization of PTV Based Temperature Measurement by Brownian Motion of Gold Nanoparticles

• Main Authors: Jun-Yang Lin, 林均洋
• Other Authors: 孫珍理
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/gjsxaj

碩士 === 國立臺灣大學 === 機械工程學研究所 === 105 === In this study, we explore the accuracy of using Brownian motion of gold nanoparticles to quantify the temperature of surrounding fluid. By employing the particle tracking velocimetry (PTV), the displacement of gold nanoparticles is measured and used to estimate the temperature through Einstein’s theory. In the image, the diameter of each particle is approximately 8 px. Influences of the tracking time, time interval, particle density and number of frames are investigated to obtain the optimal parameters, which minimize the total error. The experimental results show that particle density plays a minor role in the temperature estimation. Nonetheless, particle density lower than 10-8 ml-1 is recommended in order to avoid overlap of particles. When the root mean squared displacement (RMSD) is smaller than 1.5 pixels, pixel locking is more severe, which leads to higher systematic errors and random errors. However, random error can be reduced by increasing the number of frames at small RMSD. When RMSD falls between 1.5 and 2 pixels, temperature estimation has the lowest systematic error. Once RMSD exceeds 2 pixels, error is majorly influenced by number of frames. For a given tracking time, longer time interval is required to obtain larger RMSD, which results in fewer images. Hence, both systematic and random error raise with the increase of RMSD. For a given number of frames, on the other hands, systematic errors are nearly independent of RMSD if RMSD is larger than 1.3 pixel. Due to influence of pixel locking, systematic errors decreases with increasing RMSD if RMSD is smaller than 1.3 pixel. When tracking time is fixed, the optimal time interval is 0.06 s. When the number of frames is fixed, the optimal time interval is 0.08 s. In summary, the optimal RMSD falls between 1.5 and 2 pixels for a tracking time of 10s, so that number of frames are sufficient and the effect of pixel locking can be minimized. To reduced random errors, more than 166 frame should be taken with a RMSD larger than 1.5 pixel and smaller than 2 pixel.