Multi-domain Features Fusion Adaptive Neural Network Tool Wear Recognition Model

Hanyang Wang, Ming Luo, Fengshou Gu

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review


The condition of cutting tools is affecting the product quality, production cost and profit. Monitoring the condition correctly and accurately is very import in machining industry. In this paper, a tool wear recognition model based on adaptive neural networks with multi-domain feature fusion is presented. First, the vibration signals obtained from the sensors mounted on the working area is processed to generate the time-domain and frequency-domain features, which form a multi-dimension space. Then the core features are identified according to the distance criteria. Finally, LSTM neural network is used to determine the tool wear condition during the machining process by processing the core features. The model is verified by the data collected from industry practical experiments. The results shows that our model can successfully increase the precision of tool wear classification and has certain generalization ability under different working conditions compared with the single eigenvalue prediction method.

Original languageEnglish
Title of host publicationProceedings of TEPEN 2022
Subtitle of host publicationEfficiency and Performance Engineering Network
EditorsHao Zhang, Yongjian Ji, Tongtong Liu, Xiuquan Sun, Andrew David Ball
PublisherSpringer, Cham
Number of pages15
ISBN (Electronic)9783031261930
ISBN (Print)9783031261923, 9783031261954
Publication statusPublished - 4 Mar 2023
EventInternational Conference of The Efficiency and Performance Engineering Network 2022 - Baotou, China
Duration: 18 Aug 202221 Aug 2022

Publication series

NameMechanisms and Machine Science
Volume129 MMS
ISSN (Print)2211-0984
ISSN (Electronic)2211-0992


ConferenceInternational Conference of The Efficiency and Performance Engineering Network 2022
Abbreviated titleTEPEN 2022
Internet address


Dive into the research topics of 'Multi-domain Features Fusion Adaptive Neural Network Tool Wear Recognition Model'. Together they form a unique fingerprint.

Cite this