AOI strategies for defects identification of micro-electro mechanical system elements. Issue 2 (2nd April 2016)

• Record Type: Journal Article
• Authors: Young, H. T.; Chiou, H. S.; Huang, W. R.
• Is Part Of: Advances in materials and processing technologies. Volume 2:Issue 2(2016)
• Page Start: 220
• Page End: 226
• Publication Date: 2016-04-02
• Subjects: Automated optical inspection (AOI) -- binarisation processing -- micro-electro mechanical system (MEMS) -- PatInspect -- Sobel -- threshold; Materials -- Periodicals; Manufacturing processes -- Periodicals; 620.1105
• DOI: 10.1080/2374068X.2016.1147790
• Languages: English
• ISSNs: 2374-068X

Abstract

Abstract: The structures of micro-electro mechanical system (MEMS) components are considerably fine, and their quality control (QC) is typically achieved using visual inspection with high-labour intensity. The manual detection approach is inefficient (low yield) and limited to the discrimination ability of the human eye in each individual, and thus, bears inconsistent quality problems. Furthermore, a 100% overall inspection requirement for critical parts can barely be achieved by human inspection. This paper proposes an automated optical inspection (AOI) to replace labour visual inspection. The proposed approach provides a high-recognition rate, and enhances the detection efficiency for QC practical industry purposes, which has made it an indispensable key technology in recent developments. Because of the high complexity of MEMS structures, the manufactured parts often consist of defects such as breaks, fillet, Ag paste, scratches, particles and chips. These microdefects on an MEMS device easily cause damage and even malfunction in the entire module, and therefore, developing effective detection techniques is crucial for emerging MEMS applications. Binarisation processing has previously been adopted, and frequently results in confusion caused by illegible image problems. This paper proposes a systematic approach by applying a colour charge-coupled device, in which the characteristics of light reflectance for different colours of light (at various wavelengths) are used, with the PatInspect method for image comparison and the Sobel gradient edge detection method for processing the structure boundary. The benchmark tests for the present AOI strategy revealed that the normal confusing defects were effectively detected as a result. Working cases are provided, showing that the detection efficiency, requiring only 8 s for each die in completing the identification process involved, is up to the industry standard.