Method for the Investigation of Mold Filling in the Fiber Injection Molding Process Based on Image Processing

Abstract

Fiber Injection Molding is an innovative process for manufacturing 3D fiber formed parts. Within the process fibers are injected in a special mold through a movable nozzle by an air stream. This process allows a resource efficient production of near net-shape long fiber-preforms without cutting excess. For the properties of the preforms the mold filling is decisive, but current state of the art lacks methods to monitor mold filling online. In this paper a system for monitoring the mold filling based on image processing methods is presented. Therefor a camera and back-lighting has been integrated into a fiber injection mold. The detected filling level and fiber distribution is passed to the PLC of the fiber injection molding machine, which allows the operator to monitor the current mold filling state by means of a visual display. The image processing approach consists of preprocessing, binarization and segmentation. For the preprocessing and binarization several methods including a k-means algorithm, the Otsu thresholding method and a convolutional artificial neural network have been implemented and evaluated. Additionally the illumination of the mold has been investigated and found to have a very large influence on the quality of the results of all investigated methods. The results of the binarization are evaluated on the basis of ground truth images, where an absolute difference between labeled and binarized images is formed and the number of misinterpreted pixels is counted. Among the investigated methods, the method based on the Otsu threshold has been found to be the most efficient with regard to the achievable performance as well as to the correct detection of the current filling. The investigated approach allows the acquisition of more data about the mold filling process to improve models.