E-quality: Using dimensional index values for improving classification accuracy

E-quality is a holistic approach to gauge and ascertain product quality in real time with the use of advanced technologies. E-quality manifests a sensor-based, networked, fully automated quality control, through which a reduction of inspection time is attained. Even with the e-quality, a part classification still remains as one of the most challenging tasks because the classification is based on the minute differences among a multitude of dimensional attributes. Part classification entails many steps that complicate the accuracy of final outcome. Achieving 100% classification accuracy is not a trivial matter. In this context, this study focuses on a novel approach for improving part classification accuracy in tune with the notion of e-quality and concurrent engineering. Two approaches are proposed and compared with the traditional multiclass support vector machine classification method. One of the approaches is to modify the data before applying the support vector machine. The other is a completely new Sine methodology using dimensional index values for classifying parts into different categories. Support vector machine is employed due to its higher generalization ability, especially when the data set is small and the class overlap is nonexistent. The data extracted from a machine vision system in a networked robotic inspection cell is used to test the proposed approaches. Experimental results show that the new Sine methodology performs better than the others, displaying near 100% classification accuracy.