LG Innotek to use AI to clean up production

LG Innotek applied its AI-based inspection technology, developed by combining material information and AI image processing technologies, to the RF-SiP (Radio Frequency System-in-Package) process. Recently, the technology was also introduced for the FC-BGA (Flip Chip Ball Grid Array), and is expected to the competitiveness and quality of LG Innotek's semiconductor substrate products.

Previously, incoming raw materials underwent only a visual inspection before entering the production process. However, the continued advancement of semiconductor substrate technology changed this. Even after improving all in-process defect causes, failures in reliability evaluations continued to rise. This led the quality of incoming materials to gain attention as a decisive factor affecting reliability evaluations.

The core raw materials (that is Prepreg (PPG), Ajinomoto Build-up Film (ABF), and Copper-Clad Laminate (CCL)) that comprise semiconductor substrates arrive as a mixture of glass fibers, inorganic compounds, and other components. In the past, air voids – gaps between particles – or foreign particles generated during the material mixing process did not significantly impact product performance. However, as substrate specifications, such as circuit spacing, have become increasingly stringent, the presence of air voids and foreign particles, depending on their size, has started to cause defects.

As a result, it is virtually impossible to identify which part of the raw material is responsible for the defect using traditional visual inspection methods, which has become a significant challenge for the industry.

LG Innotek now says that it has found a way to overcome this industry challenge with AI. Its "AI-based Inspection System for Incoming Raw Materials" has been trained with tens of thousands of pieces of data on the composition of materials that are either suitable or unsuitable for a product. Based on this, it analyzes the components and defective areas of semiconductor substrate raw materials in only one minute, with an accuracy rate of over 90%, and visualises quality deviations in each lot of raw materials.

By using AI machine learning to visualise, quantify, and standardise material configurations optimised for quality, LG Innotek has been able to prevent defective raw materials from entering the production process. The company can change the material design based on the quality deviation information visualised by the AI system, allowing it to ensure that the quality of the raw materials lot is uniform at a suitable level before entering the process.

According to LG Innotek, the system has allowed the company to decrease the time required to analyse defects by up to 90%, and the cost of resolving the causes of defects has also been significantly reduced.

LG Innotek plans to enhance the AI system's detection capabilities by sharing raw materials-related data with customers and suppliers in the substrate sector through digital partnerships.