Offline development of active alignment based on empirical virtual environments

Active or sensor guided alignment presents a promising production approach for high-quality optical products and helps to overcome challenges in the tolerance chain. Application fields such as fast-axis collimation increasingly deploy automated productions solutions in order to improve and ensure stable quality even with relatively low production volume. The advantages are offset by the high demand towards the engineering as upfront cost, common to all automated production solutions with integrated evaluation of the system function. In this paper, we present our approach to overcome this problem. We propose the use of virtual environments, which are derived from empirical data generated with minimized effort in early phases of the product development. We will present options for building the necessary dataset and how our solution can derive an empirical simulation due to the application of artificial intelligence. Our solution is independent from ray tracing simulations and reflects defects, deviations and tolerances as observed in the actual product samples. This allows developing alignment algorithms offline, without the need for costly machine time and the risk of damage due to manual errors. We will present validation results, which demonstrates the capability to transfer active alignment algorithms from the virtual environment to actual automation equipment. Next to FAC alignment and the high-power laser industry our virtual environment solution is of special interest to application in the field of diffractive optical elements (DOE) and free-form optics, where low volume or changing designs demand adaptable automation solutions.