A Study of the Inspection Support Tool Development Using the Neural Network

Much of the quality inspection work in the manufacturing industry is automated. However, there are many products for which inspection work cannot be automated. For example, the tip of a rotating tool (Diamond bar) is attached with diamond particles. The same thing does not exist as one diamond bar. In this study, we will develop a tool to replace manual inspection. As a basic research, we constructed a model using neural network and convolutional neural network. The model was evaluated under various conditions. As a result, the accuracy was about 75% for the neural network model and about 85% for the convolutional neural network model.

I. Introduction

Numerous quality inspection works are mechanized and automated by the quality control section of the manufacturing industry. Conversely, the conditions in which inspection by hand cannot be eliminated owing to products that cannot be automated is neglected. For example, any product with the same shape as another does not exist as same because minute diamond particles are stuck by electrodeposition on the point of the rotating tool used for dental treatment (diamond bar). Therefore, it is inspected with a watch using a microscope in the current state. We have examined the basic consideration of the inspection support tool making for which writers still use mechanical learning.