Robust multidirectional smoothing of isoclinic parameter in digital photoelasticity

Abstract

In digital photoelasticity, one gets full-field information about the difference in principal stresses and their orientations by postprocessing the intensity data. Isoclinic data obtained using phase shifting techniques have inherent noise. The higher the external load, the higher the noise. Further, noise removal becomes complex if the isoclinic data has π-jumps, isotropic points, etc. Initially, the origin of the noise in isoclinic data obtained using phase shifting techniques is discussed. An explanation for the presence of excessive noise in a circular polariscope-based algorithm is provided. Three methods for noise removal of isoclinic data calculated using a plane polariscope are proposed using an outlier smoothing algorithm. The first two use quality and standard deviation measures to identify the noise-free pixel from each individual scan. The last method involves progressive multidirectional smoothing. The effectiveness of the proposed smoothing schemes is demonstrated using the benchmark problem of a circular disc under diametral compression. Multidirectional progressive smoothing is found to be effective in removing the noise at lower as well as at higher loads. Finally, this scheme is used to smooth isoclinic data in two other problems, one that has a π-jump and the other that has both an isotropic point and a π-jump.