Machine learning for projection multi-photon 3D printing

Projection multi-photon lithography, like all additive manufacturing techniques, requires optimization of process parameters to achieve geometrically accurate results. Determining these optimal parameters is often time-consuming. Machine learning can be used to avoid the need for experimentation by predicting optimal process parameters. A data collection scheme is presented where image analysis on optical microscope images is used to measure the dimensions of individual 2D layers printed with the projection multi-photon printing process for a range of process parameters. The dimensional accuracy of these 2D shapes is then used to train a Gaussian process regression model for forward prediction.