Committee neural network force prediction model in milling of fiber reinforced polymers
Increasingly, fiber reinforced polymers are being used in aerospace, naval, and automotive industries due to their high specific strength and high stiffness. Some of the damage problems encountered during machining these materials include: delamination, surface roughness and high rate of tool wear. Major factors that affect damage during machining in these materials are cutting forces, tool geometry, feed rate, and spindle speed. The first part of this study aims to develop an approximate mechanistic model to predict the cutting forces in the orthogonal cutting of unidirectional fiber reinforced polymers (FRP) when the fiber orientation varies from 00 to 1800. This work utilizes the mechanistic modeling approach for predicting cutting forces and simulating the milling process of fiber reinforced polymers with a straight cutting edge. Specific energy functions were developed by multiple regression analysis (MR) and committee neural network approximation (CN) of milling force data and a cutting model was developed based on these energies and the cutting geometry. Cutting force prediction models were constructed for principal and thrust cutting directions. The models are based on the specific cutting energy principle and account for a wide range of fiber orientations and chip thickness. Results from two forms of non-linear modeling methods, non-linear regression and committee neural networks, were compared. It was found that the committee neural networks provide better prediction capability of smoothing and capturing the inherent non-linearity in the data. The model predictions were found to be in good agreement with experimental results over entire range of fiber orientations from 0* to 180*. The second part of this study dealt with an improved mechanistic cutting force model for complex tool geometry by sectioning the helical cutting edge into a stacked series of straight edge cutter segments with angular offsets and calculating the forces for each segment, then adding the forces for all segments of the cutting edge. The scope of this work is to establish a three dimensional cutting force prediction model for complex cutting tool geometry using orthogonal machining database developed in first part. The cutting forces predicted have shown a good agreement with experimental results. The third part of this study dealt with building a generalized model to predict cutting forces for any combination of process parameters such as spindle speed (nt), feed rate (Vf), depth of cut (ae), rake angle (αi) and workpiece fiber layup direction ψ. Committee neural network is constructed using machining parameters – chip thickness (ac), fiber orientation angle (θ), spindle speed (nt) and feed rate (Vf) as input variables and average specific cutting energy values, (Kc and Kt) as output variables. Exhaustive experimentation is conducted to develop the model and to validate it. The training of the networks is performed with experimental machining data. Results showed that the model provides good results for unidirectional composites for all fiber orientation. The experimental results show a reasonably good fit to the predicted values, suggesting that the current approach is successful and well suitable for studying the machining of fiber reinforced polymers. Results also showed that the cutting forces are directly dependent on fiber orientation, chip thickness, rake angle, spindle speed, and feed rate.
Thesis (Ph.D.)--Wichita State University, College of Engineering, Dept. of Industrial and Manufacturing Engineering
Includes bibliographic references (leaves 149-154)