Intel Collaborates with LG Innotek to Implement an AI-powered Smart Factory

Estimated reading time: 3 minutes

Intel Core and Intel Xeon processors along with Intel Arc GPUs enable a comprehensive inspection process inside LG Innotek’s factory in Gumi, Korea.

In the world of precision component manufacturing – think of the tiny parts inside modern cellphones, automotive car displays and other smart devices – being able to replicate the same part hundreds of thousands of times with perfect accuracy and zero defects has been a moonshot goal.

Like many electronic component makers, LG Innotek, a South Korea-based global leader in materials and component manufacturing, aims to reach that goal with the help of innovative technology.

In LG Innotek’s factory in Gumi, Korea, it will do that with an orchestra of Intel technologies: Intel® Core™ processors, Intel® Xeon® processors and Intel® Arc™ built-in graphics processing units (GPUs). It’s all brought together with the OpenVINO™ software tool kit.

While LG Innotek has long used both rule-based inspection and deep learning-based systems at specific terminals on its production lines to ensure higher product quality, LG aims to build a fully automated system without any performance degradations through the broad use of AI across every stage of the manufacturing process.

Intel Delivers an AI-Backed Inspection Solution using Intel CPU/GPU

In 2024, LG Innotek began discussions with Intel for help with an AI-powered automated inspection solution based on Intel’s Core and Xeon processors and its Arc discrete GPUs.

This is how the solution works:

• Data from the production line streams into connected PCs powered by Intel Core CPUs, whose built-in graphics provide a cost-effective solution in analyzing defect data.
• Heavier workloads – like running multiple algorithms on high-resolution images – is managed by the Intel Arc discrete GPU.
• Over time, accumulated datasets are sent to pre-training servers powered by Intel Xeon.

In future collaborations, the joint team plans to include servers with Intel® Gaudi® AI accelerators to manage pre-training workloads. This orchestra of Intel-based tech – anchored by the CPU and accelerated by Intel built-in and discrete GPUs – enabled a massive reduction in the cost of building AI inspection system.

Boosts in Cost Efficiencies Enable Ability to Scale

LG Innotek’s decision to use an Arc-based discrete GPU in its primary manufacturing facility increased its cost efficiency versus equivalent hardware performance from other vendors.

That cost savings is opening the door to further economies of scale.

LG Innotek first applied Intel AI vision inspection solutions to its mobile camera module production line last year.

This year, the company plans to expand the application to its major domestic production centers, such as its Gumi 4 factory, which produces FC-BGA (flip-chip ball grid array), and overseas production lines in stages.

By building a smart factory based on Intel AI vision inspection solutions, the company plans to accelerate cost competitiveness.

OpenVINO Speeds GPU Integration and AI Development

LG Innotek engineers were initially apprehensive about embracing built-in GPUs, as the company’s technology environment was already set up to use specific discrete graphics cards in its deep-learning environment. Teams feared the challenge of rewriting and remapping code to suit built-in GPUs.

That fear was dispelled when LG Innotek saw how easy it was to use OpenVINO.

Since OpenVINO launched in 2018, Intel has enabled developers worldwide to dramatically accelerate AI-based development. It’s an open source AI tool kit that makes it easier for developers to write once and deploy AI models anywhere.

Maximizing Bandwidth for Retraining

LG Innotek engineers are also considering Intel Xeon CPUs – packed with AI accelerators perfect for AI-based workloads – to retrain deep-learning models as processes change or raw materials are altered during mass production.

Xeon CPUs can maximize parallel computation speed and support Intel® Advanced Matrix Extensions, a specialized built-in accelerator that improves the performance of deep-learning training and inference.

LG Innotek plans to actively utilize Xeon CPUs to handle its computational workloads, in a clear shift away from its traditional setup that relied on both Xeon CPUs and discrete third-party GPUs working in tandem. By relying on CPUs alone to crunch AI-based fine-tuning tasks, the company expects to minimize costs.