Modeling Mechanical Component Classification Using Support Vector Machine with A Radial Basis Function Kernel

Abstract

The process of identification and classification of products in the era of modern manufacturing industries has become a crucial pillar in enhancing efficiency, productivity, and product quality. In this research, the modeling of manufacturing product classification, such as mechanical components consisting of four classes: bolts, washer, nuts, and locating pin, was conducted. The proposed model in this study is the Support Vector Machine (SVM) with Radial Basis Function (RBF). The dataset used consists of digital images of mechanical components, with each component having 400 samples, resulting in a total of 1600 samples. The dataset is divided into training and testing data, with 300 samples for each component in the training set, and 100 samples removed from the training set for external testing as model validation. The best model parameters were determined using grid search by varying the parameter values of C and γ (gamma). The model was evaluated using K=3 fold cross-validation, and external testing utilized a confusion matrix to calculate Accuracy, Precision, Recall, and F1-Score values. The research results indicate that the SVM model with the RBF kernel, using the combination of C=10 and γ=scale, achieves the best performance in classifying the four mechanical components. This is evident from the training results of the model, which were able to obtain evaluation metrics such as Accuracy of 94.17%, Precision of 0.94, Recall of 0.94, and F1-Score of 0.94. Meanwhile, the validation results with new data not present in the training dataset show that the model can achieve evaluation metrics with an Accuracy of 93%, Precision of 0.93, Recall of 0.93, and F1-Score of 0.93. These results are consistent with the training performance, indicating that the SVM model with the RBF kernel excels in classifying digital images of mechanical components, such as bolts, nuts, washer, and locating pin.