Extraction of 3D Machined Surface Features and Applications.

Abstract

In the modern production, the measurement of surface functions becomes more and more important. Most previous work on surface functional characterization are focused on surface tribological properties (roughness domain) and cover only a small area of a large engineering surface. Therefore, characterizing large engineering surface comprehensively and rapidly presents significant challenges. This research is focused on extracting 3D surface features from waviness domain and using these features to predict surface function and detect machining errors. In this research, an improved Gaussian filter is first designed to accurately extract 3D surface waviness from a large surface height map measured by a large field view interferometer. This filter technique enhances the performance of the standard Gaussian filter when applied to a surface which has large form distortion and many sharp peaks/valleys/noise. Following this, a 3D surface waviness feature of the machined workpiece is defined and applied to assess severe tool wear. Secondly, a two-channel filter bank diagram is developed that applies a 2D wavelet to decompose a 3D surface into multiple-scale subsurfaces. 3D surface features extracted from multiple-scale subsurfaces are then used to predict surface functions and detect machining faults. In the proposed surface decomposition process, two important issues: the elimination of border distortion and the transformation between the wavelet scale and its physical dimension are addressed. Applications of 2D wavelet decomposition to 3D surfaces are demonstrated using several automotive case studies, including abrupt tool breakage detection, chatter detection, cylinder head mating/sealing surface leak path detection, and transmission clutch piston surface non-clean up detection. Finally, a novel and automated surface defect detection and classification system for flat machined surfaces is designed. The purpose of this work is to extract microscopic surface anomalies and assign each anomaly to a surface defect type commonly found on the automotive machined surfaces. A “breadboard” version surface defect inspection system using multiple directional illuminations is constructed. Related image processing algorithms are developed to detect and identify 5 types of 2D or 3D surface defects (pore, 2D blemish, residue dirt, scratch, and gouge).