IMAGE-BASED DEFECT ANALYSIS FOR 3D-PRINTED ITEM SURFACE USING MACHINE LEARNING METHODS

Subject of Research. Defect detection is an actual and challenging task in the field of additive technologies. This process enables the user to save time and reduce the consumption of material. Also, it prevents the further possible faults and defects. On that account, an automation of surface control process in 3D-printing is an essential step not only for large manufacturing companies but also for private users. The paper proposes an image-based method for quality control of a 3D-printed product by applying machine learning algorithms. Method. 3D-printed item images were taken and processed on an experimental setup composed of a camera and a single board microcomputer. The paper presents a defect detecting method based on development of image preprocessing algorithms and further machine learning by applying support vector machine method. Main Results. The presented method enables the user to find and identify “over-extrusion” and “under-extrusion” defects with high precision on the surface of the manufactured items. Practical Relevance. The developed method is intended to provide practical benefits for the private users of 3D-printing devices and companies manufacturing or applying these devices. There are the following advantages of the method application: the 3D-printing parameters are easy to be set, the reports about the product and its features are saved, the solutions to any problem occurred during printing are simple and fast. The developed method of visual quality control of the 3D-printed item surface can be significantly helpful to the expansion of automation possibilities for fast prototyping processes and take 3D-printing process to a new level.