Return to search

An artificial neural network approach to laser-based direct part marking of data matrix symbols

Certain applications have recently appeared in industry where a traditional
bar code printed on a label will not survive because the item to be tracked has to be
exposed to harsh environments. Laser direct-part marking is a manufacturing
process used to create permanent marks on a substrate that could help to alleviate
this problem. In this research, a 532 nm laser was utilized to create a direct-part
marked Data Matrix symbol onto carbon steel substrates with different carbon
content. The quality of the laser marked Data Matrix symbol was then evaluated
according to the criteria outlined in the ISO/IEC 16022 bar code technology
specification for Data Matrix.
Several experiments were conducted to explore the effects that different
parameters have on the quality of the laser direct-part marked symbols. First, an
experiment was conducted to investigate the effect of two different laser tool path
patterns. In later experiments, parameters such as type of carbon steel, percent of
laser tool path overlap, profile speed, average power and frequency were found to
have significant effects on the quality of laser direct-part marked Data Matrix
symbols. The analysis of the results indicated that contrast and print growth were
the critical standard performance measures that limited laser direct-part marked
Data Matrix symbols from achieving a higher final grade. No significant effects
were found with respect to other standard performance measures (i.e., encode, axial
uniformity, and unused error correction).
Next, the experimental data collected for contrast and print growth was
utilized as training, validation and testing data sets in the modeling of artificial
neural networks for the laser direct-part marking process. Two performance
measures (i.e., mean squared error and correlation coefficient) were employed to
assess the performance of the artificial neural network models. Single-output
artificial neural network models corresponding to a specific performance measure
were found to have good learning and predicting capabilities. The single-output
artificial neural network models were compared to equivalent multiple linear
regression models for validation purposes. The prediction capability of the single-output
artificial neural network models with respect to laser direct-part marking of
Data Matrix symbols on carbon steel substrates was superior to that of the multiple
linear regression models. / Graduation date: 2004

Identiferoai:union.ndltd.org:ORGSU/oai:ir.library.oregonstate.edu:1957/31157
Date08 March 2004
CreatorsJangsombatsiri, Witaya
ContributorsPorter, J. David
Source SetsOregon State University
Languageen_US
Detected LanguageEnglish
TypeThesis/Dissertation

Page generated in 0.002 seconds