Applicative Framework for End-to-End AOI Implementation
Two main criticalities were encountered :
1 . Integration of the new system with the DMC camera already implemented . The coexistence of DMC reading , and porosity check has been guaranteed tuning the timing of the Input / Output coordinator in order not to hinder the DMC reading , that is performed as soon as a new component is detected : the Input / Output coordinator , instead , is set to wait few seconds from the presence sensor signal reception before triggering the AOI light source ignition , in order to not affect the DMC camera proper work . As already stated , the proposed system allows full control and adaptability .
2 . Making the defect visible . After heat treatment , cast iron appears very dark . As a result , the capability of a proper illumination system to highlight the defect plays a significant role . The chosen illuminator provides circular radial lighting , so that the hidden surface appears bright in color while missing material / cavities stand out as black , since light is not reflected .
Figure 3-5 : Example of OK ( left ) and NOK ( right ) parts analyzed through the proposed AOI system spotting the presence of a porous cavity on the defective part .
The part of the system customized specifically for the described application is the AI imageprocessing module . Again , the algorithm is composed by two different steps :
• Component isolation . The surface to be inspected is limited to the outer part of the pump . Given the circular shape of the pump , isolation is achieved through the Hough Circle Transform algorithm . As mentioned , the algorithm finds any circumference with a radius included in a pre-defined range , which is set in advance as a hyperparameter . Then , it returns the coordinates in terms of radius , x-center and y-center .
14 March 2022