Abstract
Machining condition monitoring is becoming an increasingly important issue in automation. Despite many successful results, this area is still not mature enough to be used for complete unmanned automation. One primary reason is the lack of complete physical models relating variables of interest in machining. In this study, two critical machining conditions, tool wear and surface finish, are extracted during machining from three components of force signals by using neural networks. It is generally known that cutting forces are related to the state of tool wear. The presence of crater wear under production conditions make it difficult to estimate tool wear from force data. The present study adopted three layer back propagation neural network for the monitoring of system conditions. The networks are first trained using the back propagation algorithm with a known set of measured data at the training stage. Off-line measurements were taken for tool wear and surface finish at pre-determined intervals. A hierarchical network architecture was chosen to represent the physical relation of variables and reduce network sizes. After training is completed, the networks are exposed to external stimuli, i.e., cutting forces, in order to extract process conditions. Recognition of patterns can be feasible in real time. Application of the method with experimental data is also presented.
Original language | English (US) |
---|---|
Pages | 85-100 |
Number of pages | 16 |
State | Published - 1990 |
Event | Winter Annual Meeting of the American Society of Mechanical Engineers - Dallas, TX, USA Duration: Nov 25 1990 → Nov 30 1990 |
Other
Other | Winter Annual Meeting of the American Society of Mechanical Engineers |
---|---|
City | Dallas, TX, USA |
Period | 11/25/90 → 11/30/90 |
All Science Journal Classification (ASJC) codes
- Industrial and Manufacturing Engineering
- Mechanical Engineering