Publication | Closed Access
Vision Based Fault Detection of Automated Assembly Equipment
30
Citations
0
References
2011
Year
Unknown Venue
Fault DiagnosisRobotic SystemsEngineeringSmart ManufacturingReliability EngineeringImage AnalysisPattern RecognitionSystems EngineeringVision SystemMachine VisionMachine SystemsManufacturing IndustryComputer EngineeringMachine FaultsManufacturing SystemsComputer ScienceAutomated InspectionAutomatic Fault DetectionComputer VisionAssemblyAutomated ProcessingAutomationFault DetectionCamera Technology
Machine faults and breakdowns are a concern for the manufacturing industry, and automated assembly machines use many sensors to monitor health and report faults to a central controller for technician review. The project aims to evaluate the effectiveness of machine vision for detecting visually cued faults in automated assembly equipment, with ongoing work to develop a more robust system. Tests were performed on a laboratory‑scale conveyor that assembles a simple 3‑part component, using several conventional webcams and LabVIEW image processing. Preliminary results show the vision system can detect faults such as part jams and feeder jams, but its overall effectiveness is limited because it can only detect faults known beforehand.
Machine faults and breakdowns are a concern for the manufacturing industry. Automated assembly machines typically employ many different types of sensors to monitor machine health and feedback faults to a central controller for review by a technician or engineer. This paper describes progress with a project whose goal is to examine the effectiveness of using machine vision to detect ‘visually cued’ faults in automated assembly equipment. Tests were conducted on a laboratory scale conveyor apparatus that assembles a simple 3-part component. The machine vision system consisted of several conventional webcams and image processing in LabVIEW. Preliminary results demonstrated that the machine vision system could identify faults such as part jams and feeder jams; however the overall effectiveness was limited as this technique can only detect faults known prior to creating the vision system. Future work to create a more robust system is currently underway.