Nowadays, additive manufacturing which is also known as 3d printing, plays a major role in industries, healthcare, avionics and etc. The main advantages of this technology that make it popular are the ability rapid prototyping and producing final products with special geometries. Of course, the efficiency and accuracy of additive manufacturing machines has been always of great importance and are directly related to the product quality, manufacturing time, and also the amount of energy and materials consumed. Fused deposition modeling (FDM) is a low-cost conventional additive manufacturing method that is less reliable and thus more error prone in comparison to other additive manufacturing methods. Without using any monitoring methods to control the quality of products in this process, probable defects and faults in the manufacturing process are not detected up to the end of the process. As a result, the final product cannot be used and the time and cost are wasted. On the other hand, monitoring the process is not possible for the operator because of significant process time. Therefore, researchers have presented some automated methods so far for monitoring the process and ensuring the quality of products. Most of these try to detect defects and stop the process to decrease the waste of time and resources. There are also few methods that try to compensate defects and increase the production yield in this manufacturing method. In this thesis, an online monitoring system is designed for FDM machines. The first part of the system is based on the image processing techniques to monitor the quality of layers produced by the FDM process. Defects are identified by analyzing the texture of produced layers. Furthermore, proper process parameters are extracted to continue the process while keeping the desired degree of quality. To explain more, the extracted features are used as inputs to a machine learning algorithm to separate the flawless and defective layers. Then detected defects are classified into two different categories of over-extrusion and under-extrusion. Finally, a closed-loop controller uses this information to issue compensation commands and tuning the process parameters in the next layer to reduce the effect of defects. In this manner, the overall quality of the final product in addition to the process efficiency are improved. The comparison between the results of the proposed method and previous methods shows the superiority of the proposed method in terms of quantitative and qualitative criteria. As the second part, a real-time computer vision-based monitoring system is presented to identify any geometrical deviations in the printed object. The proposed method extracts a high-precision 3D model by aligning and merging different point clouds taken by two 3D cameras which covers the entire printing platform. Then this model is compared to the reference CAD model. If any defects (geometric deviation) are detected, the location of that defect is sent to the controller and the corresponding corrective commands are generated and sent to the 3D printer controller. All the steps are carried out automatically and in a real-time behavior. The eva‎luation results in this section also shows the significant improvements in the reliability of the FDM process compared to the previous methods.

امير خورسندي

رسول اميرفتاحي، نادر كريمي