A novel domain adversarial time-varying conditions intervened neural network for drill bit wear monitoring of the jumbo drill under variable working conditions. (28th February 2023)

• Record Type: Journal Article
• Authors: Lin, Lin; Guo, Hao; Guo, Feng; Lv, Yancheng; Liu, Jie; Tong, Changsheng
• Is Part Of: Measurement. Volume 208(2023)
• Publication Date: 2023-02-28
• Subjects: Wear measurement -- Domain adversarial mechanism -- Time-varying conditions intervention mechanism -- Variable working conditions -- Drill bit wear monitoring -- Image alignment; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2023.112474
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: A drill bit wear monitoring method of the jumbo drill based on DATCINN is proposed. ORB is adopted to align the images of drill bits to accurately measure the wear. Domain adversarial mechanism helps extract insensitive features. Time-varying conditions intervention improves the accuracy across operating modes. Results show high monitoring accuracy under variable working conditions. Abstract: The challenges in monitoring the drill bit wear of the jumbo drill lie in the difficulty of measuring the drill bit wear directly, and the coexistence of time-invariant conditions and time-varying conditions with complex and everchanging changes. The existing tool wear monitoring methods are limited to the time-invariant conditions with minor change, without resolving the situation of major change. And they do not consider the influence of the time-varying conditions on the wear. An accurate drill bit wear monitoring modelling method based on Domain Adversarial Time-varying Conditions Intervened Neural Network (DATCINN) is proposed in this paper. A novel measuring method based on ORB (Oriented FAST and Rotated BRIEF) is developed to accurately measure the drill bit wear using images, as the data sources. Domain adversarial mechanism helps extract features desensitized to time-invariant conditions to improve the model accuracy under variable working conditions. Time-varying conditions intervention mechanism uses the network correction term to deal with the influence of time-varying driving parameters on the drill bit wear to improve the model accuracy in different construction modes. The experimental results in 2 cases show that the accuracy and stability of the proposed model are better than the existing statistical and machine learning models. The performance is prominent in the task of monitoring drill bit wear under complex working conditions using model trained under single working condition. And DATCINN can be widely applied to wear monitoring of other non-precision cutters.