Optimization on Adhesive Stamp Mass-Transfer of Micro-LEDs With Support Vector Machine Model

In this work, the process of adhesive stamp mass-transfer of micro light-emitting diode (micro-LED) is optimized by a Support Vector Machine (SVM) model. The pick-up experiments have been performed repeatedly for hundreds of times from which the separation speed and the force between the stamp and the donor substrate are extracted as signal features. The SVM model with a Gaussian kernel function is designed to classify pick-up results into success and failure. In addition, the optimal cost parameter <inline-formula> <tex-math notation="LaTeX">$C$ </tex-math></inline-formula> as well as the Gaussian kernel function parameter gamma <inline-formula> <tex-math notation="LaTeX">$(\gamma)$ </tex-math></inline-formula> has been optimized, leading to the improvement of the classification by Particle Swarm Optimization (PSO) algorithm. Finally, an 85% classification accuracy is achieved based on the SVM model, implying that more sophisticated definition of signal features is demanded in future work.