TY  - JOUR
T1  - Tool Wear Monitoring by Using Extended Kalman Filter with Functional Update
AU - , . Hameed Hussain AU - , S. Purushothaman 
JO  - Journal of Engineering and Applied Sciences
VL  - 1
IS  - 4
SP  - 530
EP  - 537
PY  - 2006
DA  - 2001/08/19
SN  - 1816-949x
DO  - jeasci.2006.530.537
UR  - https://makhillpublications.co/view-article.php?doi=jeasci.2006.530.537
KW  - Steepest-Descent Method (SDM)
KW  -Extended Kalman Filter (EKF)
KW  -Functional Update Method (FUM)
AB  - Artificial Neural Network (ANN) is applied for pattern recognition of the tool wear in lathe.  Conventional back-propagation algorithm which uses Steepest-Descent Method (SDM), is applied to train the ANN.  One of adaptive algorithms which is the Extended Kalman Filter (EKF) algorithm to train the ANN, especially for XOR problem.  The ANN has been trained using EKF and EKF with functional update method.  To show the supremacy of EKF over SDM, the results of EKF algorithm is found to be faster than the convergence speed of SDM.  The performance of EKF algorithm is almost the same as the performance of SDM algorithm.  Experiments were conducted on a lathe to collect cutting force data and tool flank wear land width for various machining conditions.  The network was trained offline.  Fresh patterns were tested by using the weights and thresholds obtained during training.  The classification performance of EKF algorithm for the test patterns is above 75%.
ER  -