@misc{alegre_tool_2008,
title = {Tool wear classification using shape signatures from digital images and neural networks},
author = {Enrique Alegre and Rocío Aláiz-Rodríguez and Joaquín Barreiro and J Ruiz},
url = {https://d1wqtxts1xzle7.cloudfront.net/44449799/TOOL_WEAR_CLASSIFICATION_USING_SHAPE_SIG20160405-3188-z6pcbc-libre.pdf?1459894371=&response-content-disposition=inline%3B+filename%3DTool_Wear_Classification_Using_Shape_Sig.pdf&Expires=1739807418&Signature=EvzaKBJe-t9uYSU0VHYfIsHLonQIni35AX6MmZfYX-OGZpEp8t2BJx27qJd32s9rE2Cc9pvkE7~9h~snb0YBv54nDj8YR4vULhHEIdJeYDH3ggm0seHqVpOg0pkE5KRCJByFN6SpvJdwTX6zzuuuXMbbQ0kTnZ0a~pzRBxr4yuCiHStUIcT7gpfiHcYVtBgznT0XiC~AM1vif5Wed6giDaCYv3YT7bs2Hn8h3DYoE0VJGSpQwIe8rhe9GBeQE0gJmfRsc4hacjCdf2ROSX-TJqo4CEPlVk97SWMlmOoEGa3ZfDGfaIMozuJUFUptedp7W8U-0-bsAIQJQihJHdch7g__&Key-Pair-Id=APKAJLOHF5GGSLRBV4ZA},
year = {2008},
date = {2008-01-01},
publisher = {na},
abstract = {This paper presents a computer vision-based method with a neural network classifier to estimate wear in metal cutting inserts and determine replacement timing. A supervised classification approach distinguishes between low and excessive wear to ensure timely tool replacement. A dataset of 1,383 wear flank images was processed into binary images, and the wear perimeter was described using a normalized shape signature resized to 40 and 100 values. Classification using k-NN and MLP achieved error rates of 5.5% and 5.1%, respectively.},
keywords = {Contour Signature, Neural Network Classification, Tool Life, Tool wear},
pubstate = {published},
tppubtype = {misc}
}
This paper presents a computer vision-based method with a neural network classifier to estimate wear in metal cutting inserts and determine replacement timing. A supervised classification approach distinguishes between low and excessive wear to ensure timely tool replacement. A dataset of 1,383 wear flank images was processed into binary images, and the wear perimeter was described using a normalized shape signature resized to 40 and 100 values. Classification using k-NN and MLP achieved error rates of 5.5% and 5.1%, respectively.