Non-reproducible alignment and fitting algorithm effects on Laser Radar measurement

摘要

In recent years, manufacturing companies have adopted a Zero Defect Manufacturing (ZDM) approach to reduce product defects and to improve Right-First-Time (RFT) capability with minimum waste of resources [1]–[3]. In order to achieve ZDM, data collection alone is not sufficient; data mining methods are critical to evaluate the inherent variation of manufacturing processes. During New Product Introduction (NPI), Statistical Process Control (SPC) and similar tools are used to identify and eliminate defects from occurring/reoccurring. In the automotive industry inspection is typically performed using Coordinate Measuring Machines (CMMs) [4], [5] that provide high accuracy and repeatability, but are housed in dedicated off-line facilities that require a controlled environment [6], [8]. This off-line process is time consuming and only a limited number of samples can be measured. There is a rising trend to move away from off-line sample measurement to in-line data collection in order to predict defects before they happen or identify trends in the production process [9], [10].