Simplify Smart Factory Transformation with AI Vision for Enhanced Yield and Predictive Maintenance

The Vision of a Smarter Factory

Predictive maintenance has long been the aspiration of factory management. By implementing predictive maintenance and anomaly detection systems capable of halting equipment when issues are identified, product quality, yield, and productivity can be significantly improved, creating a positive feedback loop for continuous optimization. NXP Semiconductors’ smart factory epitomizes this pursuit of manufacturing excellence. Focused on Vision AI and predictive maintenance, it aims to achieve full automation and operational intelligence.

NXP^® partnered with ADLINK, deploying SP2-IMX8 ARM-Based Panel PC and EMU-210 programmable IIoT Gateway, both powered by the NXP i.MX 8M Plus applications processor and integrated with the EdgeGo platform as edge device management and NXP’s eIQ^® AI training software for Automated Optical Inspection (AOI), enabling real-time AI computing at the production lines. This marks the initial step in the transformation of manufacturing operations, with the ultimate objective of achieving comprehensive predictive maintenance capabilities. By fostering a sustainable cycle of continuous improvement, AI-driven systems pave the way for long-term growth and scalability.

1. Utilizing Vision AI to detect production anomalies and improve product yield.
2. Reducing maintenance costs through predictive data analysis.
3. Building a knowledge base to streamline process development, reducing costs and time to market.

To transform manufacturing, set clear goals, invest in the right technology, adopt AI tools, faster collaboration, ensure security, and implement solutions step by step, optimizing as you grow.

Challenges

In semiconductor manufacturing, integrated circuit (IC) processes from wafer testing, wafer back-grinding, mounting, die saw, die bonding, oven, wire bonding, molding, marking, ball placement, singulation to final inspection, pack, and shipping are critical steps where precision is paramount. Customers expect nothing less than zero-defect products, making rigorous quality control and flawless execution essential to meet these high standards and ensure customer satisfaction.
 

To achieve this, the challenges are reaching a False Positive Rate (FPR) and Error Rate at zero, fast response, low power consumption, and ensuring the confidentiality of process inspection technology. Propri-etary techniques must remain secure and protected from unauthorized disclosure during the IC packaging.
 

By adopting edge computing, the factory eliminates the concerns and gains several advantages over the traditional Client-Server Model: real-time performance with low-latency processing at the production line, improved system stability by reducing network dependence, modu-lar scalability for optimizing detection models, and enhanced data security through localized processing.

NXP has full confidence in our chips, with a 15-year guaranteed product lifecycle, and ADLINK’s panel PC and gateway designs align ideally with the stringent requirements of industrial automation. With their extensive expertise in system integration, this collaboration is truly a win-win. We have confidence in ADLINK.

Benefits

The deployment of edge computing and industrial display solutions integrates 10” mechanical customized panel PCs equipped with a 4K MIPI/USB camera for AOI, enabling machine learning (ML) intelligence directly at the edge. Secured data is then transmitted through a gateway to the cloud for advanced analysis, resulting in measurable improvements across key manufacturing stations:

1. AI-Powered Vision Inspection:

• Wafer Testing: Increased probing efficiency by 5x, with 100% detection accuracy and a reduced FPR of 0.001%.
• Die Bond Anomaly Detection: Real-time inspection with 99.99% product yield.
• Wire Bonding: Detection of particle anomalies during bonding, with secure cloud transmission.
• Final Product Packing: Eliminated human errors, such as labeling and placement.

2. Centralized Edge Management with EdgeGo software:

• Unified device control and standardization of operations via EdgeGo software.

3. Data Security with Edge Computing:

• Mitigates concerns over transmitting sensitive data to the cloud by keeping it securely on-site.

This fulfills more than unman inspection but utilizes in-depth data from the edge to make AI-driven predictive mainte-nance possible that reduces downtime and optimized equipment life spans, fostering a virtuous cycle of improved productivity and reduced resource waste.

AI Inference Speed & Spec Comparison

System Description

Highlights