AI-Based Early Detection of Raw Material Defects
Applied to High-Value Semiconductor Substrates
Defect Analysis in 1 Minute... Time Reduced by 90%
LG Innotek announced on the 25th that it has developed and applied the industry's first 'Raw Material Incoming Inspection AI' that determines defects at the time of receipt to preemptively filter out defective raw materials.
LG Innotek introduced the 'Raw Material Incoming Inspection AI,' developed by integrating material information technology and artificial intelligence (AI) image processing technology, for the first time in the RF-SiP (Radio Frequency System in Package) process. Recently, it has also been expanded to FC-BGA (Flip Chip Ball Grid Array), and it is expected to contribute to further strengthening the quality competitiveness of LG Innotek's high-value-added semiconductor substrate products.
Previously, incoming raw materials before the process input were only inspected visually. However, the situation changed due to the high specifications of semiconductor substrate products. Even after improving all defect causes originating from the process, cases failing to pass the reliability evaluation threshold increased. This highlighted the quality of incoming raw materials used in the process as a decisive factor affecting reliability evaluation.
Key raw materials constituting semiconductor substrates (PPG, ABF, CCL, etc.) are received in a mixed form of glass fibers and inorganic compounds. Previously, even if voids (gaps between particles) or foreign substances occurred during the raw material mixing process, there was no problem in product performance implementation. However, as substrate product specifications advanced, such as reduced circuit spacing, defects began to occur depending on the size of voids or the amount of foreign substances. Consequently, it became practically impossible to identify which parts of the raw materials were defective using the existing visual inspection method, becoming a challenge in the industry. If a lot (Lot?a unit of raw materials with the same characteristics used in the production process) of raw material mixture is considered a lump of cookie dough, it is like being unable to visually confirm how much salt or sugar is concentrated on one side, how many air holes have formed, or how much foreign material is included.
LG Innotek found a solution to this industry challenge in AI. The 'Raw Material Incoming Inspection AI' developed by LG Innotek learned tens of thousands of data visualizing suitable and unsuitable material compositions for good products. Based on this, it analyzes the components and defective areas of semiconductor substrate raw materials with over 90% accuracy in just one minute and visualizes quality variations by raw material lot.
Through AI machine learning, it has become possible to visualize, quantify, and standardize material compositions optimized for good products. As a result, LG Innotek can fundamentally prevent defective raw materials from being input into the process. Based on the quality variation information visualized by AI, it is now possible to change material design and uniformly maintain the quality of raw material lots at a level suitable for good products before process input.
An LG Innotek official explained, "With the introduction of the 'Raw Material Incoming Inspection AI,' the time required for defect cause analysis has been reduced by up to 90% compared to before, and the additional costs incurred to resolve defect causes have also been significantly reduced."
LG Innotek plans to continuously advance the reading capabilities of the Raw Material Incoming Inspection AI through a 'digital partnership' that mutually shares raw material-related data with substrate field customers and partners. Alongside this, the company plans to expand the application of the 'Raw Material Incoming Inspection AI' to optical solution product groups capable of detecting raw material defects based on images, such as camera modules.
No Seung-won, CTO (Executive Vice President) of LG Innotek, said, "With the introduction of the 'Raw Material Incoming Inspection AI,' we have been able to complete LG Innotek's unique AI ecosystem that provides differentiated customer value by identifying various defect causes of products in advance. We will continue to pursue digital production innovation that enables the production of the highest quality products at minimal cost and shortest time."
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
