| dc.description.abstract | The dissertation has two aspects. The first aspect is to investigate the relationships between process parameters and responses of interest in two types of additive manufacturing processes: PolyJet and Binder Jetting. The second aspect is to compensate color deviations of printed samples from color specifications entered in the printer software using a data-driven approach.
In the first two chapters of the dissertation, the effects of PolyJet process parameters (such as surface finish type, print orientation, and layer thickness) on mechanical properties, surface roughness, and dimensional accuracy of printed samples are investigated by designed experiments. Analysis of variance (ANOVA) is used to determine whether these process parameters are statistically significant. Some process parameters are found to have significant main effects as well as interaction effects on the responses.
From chapter three to chapter five, color inaccuracy issue of the PolyJet process is addressed in three steps. First, the effects of surface finish type on the measured color of printed samples are investigated. Then, a machine learning algorithm, i.e., multilayer perceptron neural network, is utilized to model the relationship between color specification and measured color. The established model results in a high prediction accuracy. Lastly, a compensation methodology, which combines design of experiments, predictive modeling, and multi-response optimization, is proposed to find the optimal color specifications for the printer software to result in a color on the printed sample that has the smallest deviation from the target color. Case studies validate that the proposed methodology reduces color deviations.
In the last part of the dissertation, powder dispense rate of ultrasonic hopper dispensing system equipped on a Binder Jetting 3D printer is investigated. The effects of initial powder amount, ultrasonic intensity, ultrasonic frequency, and cumulative number of dispensing cycles on powder dispense rate are evaluated experimentally. It is found that powder dispense rate can be controlled by altering process parameters, but it decreases as the powder dispensing process continues, even with fixed process parameters. | |
