| Historically the hot-forging industry has suffered from high production costs due to a high defect rate. Current inspection methods are not capable of measuring hot parts which have just emerged from a forging process. Therefore inspection can only proceed after a 30-minute cooling period. As a result, hundreds of parts produced during that period have to be scrapped once a defect is found. As higher accuracy is demanded, the number of scrapped parts increases. A project is conducted here to implement a real-time inspection system by using vision devices and computer assistance. Research efforts are divided into two phases. In the first phase, the hardware layout of the system is designed, inspection and measurement methodologies are studied, accuracy of the vision system is optimized and fundamental software is developed. In this phase, inspection is driven manually by operators. A new measurement methodology is developed, and various edge-locating methods such as edge detection, thinning and approximation are studied and applied to help operators locate correct dimensions. Intelligent feedback is implemented to correct possible locating errors. The impact of camera-object alignment is also discussed thoroughly. The first phase of research is reported in this dissertation. The second phase, not included in this dissertation, will follow after completion of phase one research. In the second phase, thermal contraction will be studied and the system will be designed to conduct inspection automatically. |
