Fused Deposition Modelling (FDM) is one of the most popular 3D printing technologies because of its affordability and accessibility. FDM, however, often suffers from printing errors that result in wasted time, materials and energy. To address these challenges, this thesis introduces a novel fault detection system for FDM printers. This system is designed to identify a broad range of errors without interrupting the printing process. To achieve a real-time detection system, an innovative multi-camera setup is designed, integrating two side cameras and one nozzle camera. Our hypothesis is that a system including three cameras can provide a more comprehensive view and can ensure more error types to be detected. Error detection is achieved using Convolutional Neural Networks (CNNs). This is a type of machine learning that excels at image recognition and pattern detection, making it well-suited for identifying printing errors in real-time models. Two CNN models are developed to classify images into common 3D printing errors: one model for the nozzle and another for the side cameras. The models were trained and validated on diverse datasets containing various shapes, infills, and augmented data. The nozzle camera model achieved a high validation accuracy of 97.68% with a low loss of 0.07464. The side camera model achieved comparable performance with a validation accuracy of 97.61% and validation loss of 0.1196. These two well performing models were for the first time ever integrated into a unified fault detection system based on a logic-driven priority framework. From this research, we learned that integrating multiple viewpoints into a logic-driven priority framework significantly improved the robustness of error classification, as many more error types could be detected in-situ and real-time. As a result, the integrated system successfully detected 12 common printing errors. In summary, this work shows the feasibility of developing a robust multi-input fault detection system to improve 3D printing. It paves the way for further research and implementation for complex integrated error detection and correction mechanisms.