(370r) Development of an Image Analysis Tool for High-Speed Imaging Data in a Dropwise Additive Manufacturing Process

Authors: 
Radcliffe, A. J., Purdue University
Reklaitis, G. V. R., Purdue University
The use of additive manufacturing (AM) technology has in recent decades gained considerable attention for its various applications in materials (e.g. ceramics, semiconductors) fabrication, medicine (e.g. bioprinting of tissue scaffolds), and pharmaceuticals, to list only a few [1]. Of the various AM techniques, the drop-on-demand printing method – adapted from inkjet printing – is particularly widely used as it is applicable to essentially any two-dimensional (2D) or three-dimensional (3D) printing process in which precise, repeatable dispensing of drops onto a substrate is required. In drop-on-demand (DoD) printing, drops are formed by the ejection of a fluid jet from a print head, driven by a thermal, piezoelectric or mechanical actuator, which subsequently breaks up into one or more drops, dependent on the ink rheology and process conditions [1, 2]. Such free surface flows and the related phenomena of jet breakup and drop formation are the subject of a considerable body of literature devoted to theoretical and experimental investigations, which, due to the small timescales at which these events occur, necessitate the use of high-speed imaging for observation [1, 2, 3].

From a systems perspective, this implies that in the development or adaptation a drop-on-demand manufacturing process, a high-speed image capture system is the only way to access the experimental process dynamics which provide information on the effects of ink rheology and process conditions. Such systems combine appropriate illumination and triggering with high speed cameras to achieve image capture rates of 103-106 (or higher) frames per second and development of such systems is a scientific endeavor in itself [3]. In this work we consider not the image capture system but the large data sets which are generated during experimental investigation of drop formation dynamics in DoD printing. Such data sets contain considerable information about the drop formation dynamics, which is useful to researchers working to elucidate physical phenomena related to free surface flows, and also to process engineers working toward development of new ink formulations or DoD processes.

The objective of this work is to develop an image analysis tool which is capable of extracting information of interest from such data sets in a systematic manner. Automation of the image analysis algorithm is simply a necessity in this context, given that the sizes of image data sets are potentially very large (total number of images, and also hundreds or thousands of GB) and thus would require extensive time to analyze. Investigation of drop formation dynamics in DoD printing using image analysis is certainly not new, and extant algorithms provide the time evolution of axial velocity, filament position and filament diameter [3,4]. However, of interest in this work is the development of a set of algorithms which provide more details, such as full axial velocity profile, number and volume of all drops that form from a single actuation event, the time and position of primary, secondary, and subsequent filament break-up events, and liquid filament retraction dynamics after break-off. Subsequent curve-fitting and analyses are to be performed in automated manner by detection of general conditions (e.g. start-up of flow, break-off of a liquid filament), so that the set of algorithms may be applied to an image data set without the need for human pre-processing. Additionally, the use of gradient-based edge detection methods for image binarization are discussed for their relevance to this work and its generalizability.

In this work an image analysis tool is presented as a set of algorithms which accept raw image data from a high-speed imaging system, used to record drop formation dynamics in a dropwise additive manufacturing system, and perform an automated analysis for all of the parameters and metrics briefly described above. Case studies drawn from real image data are used to illustrate the effectiveness of the algorithm in challenging scenarios; parallelization of the algorithm to improve processing speed is presented, and benchmarks are provided using a real data set of 500,000 images. The utility of this tool for systems engineering of a dropwise additive manufacturing process for pharmaceuticals is presented in terms of the insight into the effects of ink formulation and particle-driven phenomena on process performance.

References

  1. Basaran, O. A., Gao, H., & Bhat, P. P. (2013). Nonstandard inkjets. Annual Review of Fluid Mechanics, 45, 85-113.
  2. Eggers, J., & Villermaux, E. (2008). Physics of liquid jets. Reports on progress in physics, 71(3), 036601.
  3. Versluis, M. (2013). High-speed imaging in fluids. Experiments in fluids, 54(2), 1458.
  4. Hutchings, I., Martin, G., & Hoath, S. (2007). High speed imaging and analysis of jet and drop formation. Journal of Imaging Science and Technology, 51(5), 438-444.