Kernel Density-Based Pattern Classification in Blind Fasteners Installation

Abstract

In this work we introduce a kernel density-based pattern classification approach for the automatic identification of behavioral patterns from monitoring data related to blind fasteners installation. High density regions are estimated from feature space to establish behavioral patterns, automatically removing outliers and noisy instances in an iterative process. First the kernel density estimator is applied on the fastener features representing the quality of the installation. Then the behavioral patterns are identified from resulting high density regions, also considering the proximity between instances. Patterns are computed as the average of related monitoring torque-rotation diagrams. New fastening installations can be thus automatically classified in an online fashion. In order to show the validity of the approach, experiments have been conducted on real fastening data. Experimental results show an accurate pattern identification and classification approach, obtaining a global accuracy over 78% and improving current detection capabilities and existing evaluation systems.