Post-Processing for Three Class of Tool Wear Prognosis using Two Class ANN Classifier based on Vibration of CNC Milling

Anis Arendra, Sabarudin Akhmad, Herianto, Kukuh Winarso

2020

Abstract

This research propose a novel method of utilizing bi-levels tool wear classifiers to prognose three levels of tool wear through additional post-processing stages. The classifier uses a multi-layer perceptron (MLP), single hidden layer, trained using the resilient backpropagation method. The original classifier output range -1 to 1 and threshold 0.0 for the separator of two classes, has been able to achieve 100% classification accuracy of two CNC tool conditions, severe wear and normal one, based on vibration features in the time domain and order domain. This classifier was tried to classify three levels of tool wear: normal, moderate wear, severe wear, according to ISO 8688 standard. Output of existing MLP classifier is passed through a moving average filter with period 4 and using threshold of -0.8 and +0.8 for three level separation, normal tool, moderate wear, severe wear. The proposed method is proven to achieve 89.98% accuracy from 459 tests. Fail safe missclassification occurred from 153 test cases which were supposed to be moderate wear, 46 of them were incorrectly indicated as severe wear. For the severe wear test case and normal tool test case, no prediction errors were found. The 100% accuracy for both test case prediction

Download


Paper Citation


in Harvard Style

Arendra A., Akhmad S., Herianto. and Winarso K. (2020). Post-Processing for Three Class of Tool Wear Prognosis using Two Class ANN Classifier based on Vibration of CNC Milling. In Proceedings of the International Conference on Culture Heritage, Education, Sustainable Tourism, and Innovation Technologies - Volume 1: CESIT, ISBN 978-989-758-501-2, pages 269-276. DOI: 10.5220/0010307500003051


in Bibtex Style

@conference{cesit20,
author={Anis Arendra and Sabarudin Akhmad and Herianto and Kukuh Winarso},
title={Post-Processing for Three Class of Tool Wear Prognosis using Two Class ANN Classifier based on Vibration of CNC Milling},
booktitle={Proceedings of the International Conference on Culture Heritage, Education, Sustainable Tourism, and Innovation Technologies - Volume 1: CESIT,},
year={2020},
pages={269-276},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0010307500003051},
isbn={978-989-758-501-2},
}


in EndNote Style

TY - CONF

JO - Proceedings of the International Conference on Culture Heritage, Education, Sustainable Tourism, and Innovation Technologies - Volume 1: CESIT,
TI - Post-Processing for Three Class of Tool Wear Prognosis using Two Class ANN Classifier based on Vibration of CNC Milling
SN - 978-989-758-501-2
AU - Arendra A.
AU - Akhmad S.
AU - Herianto.
AU - Winarso K.
PY - 2020
SP - 269
EP - 276
DO - 10.5220/0010307500003051