Publication Date

2016

Document Type

Thesis

Committee Members

Rory Roberts (Committee Member), Philippe Sucosky (Committee Member), Zifeng Yang (Advisor)

Degree Name

Master of Science in Mechanical Engineering (MSME)

Abstract

Particle Image Velocimetry (PIV) has been of relevant discussions lately as the equipment used to obtain temporally and spatially resolved flow fields have advanced rapidly. Despite these advancements, the accuracy of evaluating these images have yet to exceed expectations. Current techniques typically utilize one method, either correlation based (frequently) or optical flow (non-frequently), and both are vulnerable to specific conditions incorporated in the PIV images. Only through the combination of two methods, cross correlation and optical flow, can a technique benefit from the strengths of each method while concealing the flaws each individual method contains. The Hybrid Particle Image Velocimetry method utilizes the fairly unrestricted cross-correlation method, which can process images that contain particles with relatively large displacements, and the high resolution analysis of the Optical Flow method. Susceptible to large displacements, the Optical flow method is restricted to images with particularly small displacements. Combining the two methods requires the constraints set forth on the Optical flow method to be conserved. Meaning that the Cross-correlation results have to be manipulated into a form applicable for the Optical Flow method. Thus steps such as interpolation, shifting the image, and filtering the image are crucial for transitioning cross-correlation results to optical flow analysis. Validating the accuracy of the Hybrid method was conducted through standard PIV images that encompassed various parameters encountered in PIV. Each set of images were analyzed by the hybrid method and three other commonly-used correlation techniques in order to compare the hybrid method's accuracy with current methods. Results confirmed that the Hybrid method is consistently more accurate than the other methods, especially near the boundaries. Additionally, for cases dealing with large particles or small displacement, the Hybrid method attains more accurate results.

Page Count

65

Department or Program

Department of Mechanical and Materials Engineering

Year Degree Awarded

2016


Share

COinS