Grau, GerdBrishty, Fahmida Pervin2022-12-142022-12-142020-11-042022-12-14http://hdl.handle.net/10315/40585Inkjet printing is an active domain of additive manufacturing and printed electronics due to its promising features, starting from low-cost, scalability, non-contact printing, and microscale on-demand pattern customization. Up until now, mainstream research has been making headway in the development of ink material and printing process optimization through traditional methods, with almost no work concentrated on machine learning and vision-based drop behavior prediction, pattern generation, and enhancement. In this work, we first carry out a systematic piezoelectric drop on demand inkjet drop generation and characterization study to structure our dataset, which is later used to develop a drop formulation prediction module for diverse materials. Machine learning enables us to predict the drop speed and radius for particular material and printer electrical signal configuration. We verify our prediction results with untested graphene oxide ink. Thereafter, we study automated pattern generation and evaluation algorithms for inkjet printing via computer vision schema for several shapes, scales and finalize the best sequencing method in terms of comparative pattern quality, along with the underlying causes. In a nutshell, we develop and validate an automated vision methodology to optimize any given two-dimensional patterns. We show that traditional raster printing is inferior to other promising methods such as contour printing, segmented matrix printing, depending on the shape and dimension of the designed pattern. Our proposed vision-based printing algorithm eliminates manual printing configuration workload and is intelligent enough to decide on which segment of the pattern should be printed in which order and sequence. Besides, process defect monitoring and tracking has shown promising results equivalent to manual short circuit, open circuit, and sheet resistance testing for deciding over pattern acceptance or rejection with reduced device testing time. Drop behavior forecast, automatic pattern optimization, and defect quantization compared with the designed image allow dynamic adaptation of any materials properties with regards to any substrate and sophisticated design as established here with varying material properties; complex design features such as corners, edges, and miniature scale can be achieved.Author owns copyright, except where explicitly noted. Please contact the author directly with licensing requests.Computer engineeringElectrical engineeringElectronic Thesis or Dissertation2022-12-14Machine learningRandom forestDecision treeGradient boostingEnsembleVoting stackingRMSEJettingContributionBiasFeature engineeringImage processingPredictionClassificationInkjet printingRegressionComputer visionPrinted electronicsSymmetric printingContourRasterVectorDefectSegmentationPrecisionRecall.