Affiliation:
1. Mechanical and Production Engineering Department, National University of Singapore, Singapore 119260
Abstract
Tool failure and chatter are two major problems during machining. To detect and distinguish the occurrences of these two abnormal conditions, a novel parallel multi-ART2 neural network has been developed. An advantage of this network is more reliable identification of a variety of complex patterns. This is due to the sharing of multi-input feature information by its multiple ART2 subnetworks which allow for finer vigilance thresholds. Using the maximum frequency-band coherence function of two acceleration signals and the relative weighted frequency-band power ratio of an acoustic emission signal as input feature information, the network has been found to identify various tool failure and chatter states in turning operations with a total of 96.4% success rate over a wide range of cutting conditions, compared to that of 80.4% obtainable with the single-ART2 neural network.
Subject
Industrial and Manufacturing Engineering,Computer Science Applications,Mechanical Engineering,Control and Systems Engineering
Reference32 articles.
1. Bendat, J. S., and Piersol, A. G., 1980, Engineering Applications of Correlation and Spectral Analysis, John Wiley & Sons, Inc.
2. Burke
L. I.
, 1993, “Unsupervised Neural Network Approach to Tool Wear Identification,” IIE Transactions, Vol. 25, No. 1, pp. 16–25.
3. Carpenter
G. A.
, and GrossbergS., 1987, “ART2: Self-organization of Stable Category Recognition Codes for Analog Input Patterns,” Applied Optics, Vol. 26, No. 23, pp. 4919–4930.
4. Chryssolouris
G.
, DomroeseM., and BeaulieuP., 1992, “Sensor Synthesis for Control of Manufacturing Process,” ASME JOURNAL OF ENGINEERING FOR INDUSTRY, Vol. 114, pp. 158–174.
5. Chung, E. S., Chiou, Y. S., and Liang, S. Y., 1993, “Tool Wear and Chatter Detection in Turning via Time-Series Modeling and Frequency Band Averaging,” Proc. of the 1993 ASME Winter Annual Meeting, New Orleans, LA, pp. 351–358.
Cited by
24 articles.
订阅此论文施引文献
订阅此论文施引文献，注册后可以免费订阅5篇论文的施引文献，订阅后可以查看论文全部施引文献