Robust model-based fault diagnosis for dynamic systems
Robust model-based fault diagnosis for dynamic systems
Application of RBF and SOFM neural networks on vibration fault diagnosis for aero-engines
ISNN'06 Proceedings of the Third international conference on Advances in Neural Networks - Volume Part III
| Hi-index | 0.00 |
As the semiconductor industry moves toward ULSI era, stringent and robust fault detection technique becomes an essential requirement. Most of the semiconductor processes have nonlinear dynamics and exhibit inevitable steady drift in nature, traditional statistical process control (SPC) must be used together with the time series model in order to detect any nonrandom departure from the desired target. However, it is difficult to create the time series model and sometimes found not tolerant to non-stationary variation. To overcome this difficulty, a fault detection technique using radial basis function (RBF) neural network was developed to learn the characteristics of process variations. It is adaptive and robust for processes subject to tolerable drift and for varied process setpoints. Therefore, equipment malfunctions and/or faults can be detected and the false alarms can be avoided.