PCB makers trialing AI-based recognition modules to support AOI
Taiwan-based PCB makers are trialing AI-based recognition modules to hike accuracy for automatic optical inspection (AOI), according to industry sources.
AOI captures images of semi-finished products based on machine vision and determine whether they have defects based on machine learning. AOI is widely used in the PCB sector, taking up 64% of AOI use and followed by display panel industry with 15%.
To achieve high yield rates, PCB makers usually set strict parameters for AOI. AOI equipment is sensitive to ambient light, and can make wrong judgments under interference of slight external light. As a result, AOI has an over-screening rate of 70%, that is, 70% of AOI-determined defects are actually not defects. PCB makers have to conduct labor-based secondary inspection following AOI.
In order to solve the over-screening problem, the PCB manufacturing industry is trailing AI-based recognition modules to aid AOI. Specific to unknown defects, it is necessary to set parameters, equivalent to defining sample defects, for AOI to screen out defects exactly based on the parameters, whereas AI-based recognition modules can self-define the scope of defects based on AI-based learning and efficiently determine if there are defects, according to Smart Machinery Promotion Office. Trial use of AI-based recognition modules shows the over-screening rate can be lowered to 25%, it said.
Manufacturing industries in Taiwan are applying AI technology to predictive or preventive maintenance and quality inspection, IBM Taiwan noted. AOI is becoming the mainstream in development of quality inspection methods and its use has been extended from electronics and semiconductor industries to metal, plastics, rubber and food sectors, IBM Taiwan said, adding development of smart quality inspection will focus on 3D optical inspection and analysis of causes of defects.
IBM Taiwan and PCB maker Unimicron Technology have cooperated to develop AI-based decision-supporting mechanism for AOI. AI can model experience in labor-based inspection and make analysis and judgment based on algorithms to hike accuracy and reduce reliance on labor for quality inspection.
By continuing to use our site, we assume that you consent to receive cookies on Taiwan Smart Machinery website and accept our use of revised Terms of Service & Private Policy.