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

E-quality is a process through which inspection of the process and quality of the part produced is done online resulting in the improvement of the process and reduction in the amount of time consumed for the overall process. Automated quality control involves using a methodology to classify the parts based on the dimensions of the features on a part. However, achieving 100% classification accuracy is not an easy task, especially in area of quality control where small differences in dimensions result in part fall into a different category. In this study, a novel approach for modifying the data before being used for training of Support vector Machines (SVM) is presented. A new methodology for classifying the parts into different categories is also presented and the classification accuracies of both the approaches are compared with that of the traditional SVM approach. SVM was used as benchmark keeping in view of its higher generalization ability especially when the data set is small and class overlap is non-existent, primary attribute of quality control data. The data extracted from Machine Vision Systems (MVS) in a robotic set up was used as case study to demonstrate the three procedures. Results show that the proposed new (sine) methodology yielded superior results compared to the rest in the current scenario with 100% classification accuracy. Moreover, it was found that with the proposed methodology, the classification accuracy can be improved up to the level of 8th decimal point by using more accurate \u27C\u27 value. In the current work, accuracy up to 4th decimal point was demonstrated. Any number of features on a part can be used without limit, for classification with the proposed new methodology, accuracy being unaffected