الفهرس 
CHAPTER (1): INTRODUCTIONOnly 14 pages are availabe for public view
 |  | from | 152 |  |  |
| | | from | 152 | | |
Abstract
Burnishing process is one of the most important surface finishing processes.
Despite being known for more than 30 years, it has recently increased its industrial use in several applications such as manufacture of injection molds, tools for metal forming, as well as, various components in the automotive, railroad and aerospace industries; so its research is essential. It has been increasingly applied in manufacturing due to its several advantages that are lacked in the other finishing processes. Though, many aspects of this technology need to be deeply understood through study, predict and analyzing the influence of specific parameters to improve the understanding of the process and the optimization of the parameters involved in it.
In this thesis, in addition to Taguchi method and Response Surface Methodology
(RSM), Artificial Neural Network (ANN) technique is used to model and simulate
the roller burnishing process. Significant roller burnishing characteristics (such as
rotational speed, number of passes, depth of penetration, and feed rate) are
considered, as well as process output responses (surface roughness, surface
microhardness, surface out of roundness, and diameter change). The Design of
Experiments approach is utilized to efficiently determine values of roller burnishing parameters used during experiments for both Al-2011 and Al-7075.
Tests were conducted on a CNC Lathe machine. An accurate Single roller
burnishing tool was used. Five values of each parameter were employed in the
investigation. Burnishing speed in the range from 200 to 1000 rpm. Five different
passes in the range from 1 to 5 were employed. Experiments were carried out by applying five different depths of penetration from 25 to 145 μm. In addition, feed rates in the range from 25 to 125 μm/rev were used. One of the main objectives of this investigation is to present a development of single and multi-optimizations for obtaining the best combinations of roller burnishing parameters that lead to good results of different responses through three
methodologies (Taguchi method, RSM, and ANN). The analyses shows that the Multi-Layer Back Propagation Artificial Neural
Network using Levenberg-Marquardt (LM) is the most accurate methodology for obtaining the best combinations of roller burnishing parameters that lead to good
results of different responses. Although ANN model provided better prediction
accuracy than Taguchi method and RSM, both methods showed relatively high prediction accuracy and can be used for the same purpose. In addition, Grey relational analysis (GRA) in the Taguchi approach for multi-response problems optimization is a very valuable tool for predicting different surface characteristics in
the burnishing process. Furthermore, RSM enables the quadratic model to be fitted
for a variety of responses. It was found that the best network for Al-2011 is 4-30-1 for average surface roughness (Ra), 4-5-1 for surface microhardness (HV), 4-40-1 for surface out of roundness (OR), and 4-30-1 for change in diameter (∆D). The best network for Al7075 is 4-50-1 for average surface roughness (Ra), 4-50-1 for surface microhardness (HV), 4-30-1 for surface out of roundness (OR), and 4-40-1 for change in diameter (∆D). Burnishing speed interacts with the number of passes, depth of penetration, and burnishing feed to govern the final surface roughness of an aluminum alloy 7075 workpiece. Surface finish is improved by combining high speed with high number of passes. Surface roughness improves slightly when burnishing speed is increased at low depth of penetration. The provided results will assist production and design engineers in allocating and selecting roller burnishing settings in order to produce surfaces with high performance characteristics.