A Method for Straw Crushing Tool Wear State Recognition Based on HFS and Ensemble Learning Model

• Published in: International journal of precision engineering and manufacturing Vol. 26; no. 8; pp. 1907 - 1920
• Main Authors: Zhou, Long, Xie, Lirong, Bian, Yifan, Lin, Zhikang, Shi, Minglei
• Format: Journal Article
• Language: English
• Published: Seoul Korean Society for Precision Engineering 01.08.2025; Springer Nature B.V
• Subjects: Accuracy; Crushing; Engineering; Ensemble learning; Industrial and Production Engineering; Machine learning; Materials Science; Newton-Raphson method; Noise reduction; Redundancy; Regular Paper; Support vector machines; Tool wear; INRBO-XGBoost; HFS; Multi-domain feature; Signal cancellation; Straw crushing tool
• ISSN: 2234-7593; 2005-4602
• DOI: 10.1007/s12541-025-01232-7

Aiming at the problem of low recognition accuracy due to high feature dimensionality, redundancy and inconspicuous key features in the process of straw crushing tool wear state recognition, the study designed a hybrid feature selection (HFS) with maximum mutual information coefficient (MIC)combined with support vector machine recursive feature elimination(SVM_REF) and improved newton–raphson-based optimizer (INRBO) to optimize the extreme gradient boosting (XGBoost) method for tool wear state recognition. improved newton–raphson-based optimizer (INRBO) to optimize extreme gradient boosting (XGBoost) for tool wear state identification. Initially, the vibration acceleration signal undergoes detrending and noise reduction, followed by the extraction of its time-domain, frequency-domain, and multi-scale arrangement entropy to compile a multi-domain feature set. Subsequently, redundant features are eliminated through a HFS approach, and INRBO-XGBoost is deployed for tool wear state identification. Ultimately, comparative experiments are executed on the proposed methodology. The results of these experiments demonstrate that the proposed method attains a recognition Accuracy of 99.12%, with Precision, Recall, and F-score metrics all exceeding 0.99.This highlights the method’s high reliability and practical applicability in the recognition of wear states for straw crushing tools.