Events

As one of the world’s largest microfluidic device producers, HP is exploring more into AI tools for accelerating the commercialization of novel technology. There are many factors going into the accuracy of small printer heads that let ink flow onto a page of paper. This project is intended to accelerate microfluidic design in HP’s EMAG R&D lab by building applications on top of a functioning deep neural network to improve this accuracy. This work will focus on a web application that reads and writes neural network I/O using an interactive generative design for microfluidic images. There is the ability to pass parameterized inputs to the neural network, generate new designs, and view neural network predictions in order to see how editing these designs affects fluid viscosity, fluid velocity, max overshoot rates, and more.  

In a high-level sense, we were tasked with creating a webapp for the HP R&D EMAG team in order to help with their prediction process. Our webapp lets the user input fluid parameters, create fresh .GEO files (layer-based), and run them through our PyTorch neural network to gain valuable statistics about the geometry. The goal of this project is to build an interactive interface to output the parameters for Thermal Inkjet printhead performance. The simulation operates based on a deep neural network built on a large historic fluidic database. With the large breadth of data available to us via the neural network, we can make accurate predictions based on the given geometry input. These predicted outcomes will help engineers adjust their design for the Thermal Inkjet printhead to meet the expected performance. This project will serve as a tool for the HP engineers so that they can expedite the processes between geometry generation and prediction. This will save huge amounts of time and money in experimenting with different designs, as well as automate a lot of routine tasks.

All in all, this project will help serve the HP R&D teams with their development behind the Thermal Inkjet printer heads. Our web application gives a full workflow from start to finish and presents the data in a well-organized, attractive design. Which this tool isn’t available to the outside world (NDA and IP’s signed), it will serve a great purpose in advancing the technology behind HP’s printers.