In the context of optimizing the production of a fully connected "smart" 3d printers factory, machine learning methods like Genetic algorithms, Deep Neural Networks as well as more traditional algorithms like Job-shop were used in a simulation environment (Robotic Operating System).
Abstract — The genetic algorithm for factory orchestration is interesting due to its good results and flexibility. We present a genetic algorithm matching a 3D-printing factory setup and its surroundings so as to reduce the time needed and improve the quality of production. We programmed a factory setup to include multiple printers, conveyor belts and inspection stations. The makespan and the quality of the output are dependent on the complexity and the size of orders as well as inherent parameters of the printers. The genetic algorithm was tested with different weights in the fitness function, and different amounts of orders. The best results were achieved, when the weights of quality and makespan were nearly balanced.