Processing of optical sensor data for tool monitoring with neural networks

The output of optical tool monitoring systems are in most cases contrasted images of tools showing the worn parts of the tools with a high resolution. However problems occur with fast and consistent evaluation of these images because multiple tool wear marks exist in various shapings. The idea of using neural networks to process optical sensor data seems to suggest itself because they are tolerant to errors and able to learn by teaching various frames. The design and optimization of a structural model based on a neural network for the evaluation of optical sensor data as an application of neural networks in manufacturing engineering is explained. Results of this application of neural networks during training phase as well as the ability to generalize are shown.<>