Prediction of remaining life of industrial cutting tools with deep learning assisted image processing techniques

In the industrial production process, raw materials are transformed into new products by being subjected to different processes on large benches. During this transformation, cutting tools are used to process the raw material on the benches. Cutting tools that have completed their lifespan damage the raw material they process. In addition, cutting tools that are changed prematurely lead to inefficient use. In both cases, producers experience loss of cost, labor and time. Therefore, estimating remaining life of cutting tool is important to avoid these losses. Cutting tool life varies depending on factors such as material type, structure, thickness, and this causes uncertainties in the remaining life estimation. Although there are studies on the residual life estimation in the literature, there is a lack of empirical investigation on the estimation of the remaining life based on the damage caused by the cutting tool to the raw material. Within the scope of this study, hole photos taken from the hole drilling machine of a company operating in the aerospace sector were used. Photos been labeled into five classes according to hole quality. This problem was transformed into a classification problem and tried to be solved by using convolutional neural networks, which is one of the deep learning methods. In the study, an 18-layer model developed from scratch was compared with models using transfer learning and VGG16, ResNet50 architectures, and a hybrid model using convolutional neural network and support vector machines. It has been determined that the 18-layer architecture gives better results compared to other architectures. The results of this study showed that the remaining life of the cutting tools can be estimated from the damage to the raw material. Experiments have proven that deep learning methods can be used as an alternative for remaining cutting tool life estimation.