Development of machinability data model for end milling using artificial neural networks
Machinability data is a crucial factor affecting manufacturing cost and quality. Two artificial neural network machinability data models have been developed for the recommendation of proper cutting speed and feed rate for the peripheral end milling process. The first model is for single tool of high...
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2009
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/51547/ http://psasir.upm.edu.my/id/eprint/51547/1/FK%202009%20115RR.pdf |
| Summary: | Machinability data is a crucial factor affecting manufacturing cost and quality. Two artificial neural network machinability data models have been developed for the recommendation of proper cutting speed and feed rate for the peripheral end milling process. The first model is for single tool of high speeds steel with inputs of material hardness, cutter diameter and ration of radial depth of cut to cutter radius. An identical model is developed with an additional input of cutter tool type has shown to be are able give appropriate recommendation of cutting speed and feed rate. The models were trained and tested with data from the most general and widely used
Machining Data Handbook by Metcut and Associates. Model A and B results in the best least MSE of 4.91 x 10-5 and 1.61 x 10-4 respectively, after being trained for 3 x 10-8 iterations. The development aspects of the models, the mapping ability of hyperbolic tangent functions in perspective of summation neurons used to develop
the neural network model are discussed. The minimum number of hidden neurons needed for mapping stepped pattern using hyperbolic tangent function was analysed. Two hidden layer networks are able to represent the nonlinearity of the machinability data to be modelled. The evaluation of the network is enhanced with the inclusion of standard deviation. |
|---|