May 19, 2024

How Generative AI Can Be Used In Electronics Manufacturing

CEO of DarwinAI. Passionate about transforming manufacturing by instilling trust in AI.


President Biden’s invocation of the Defense Production Act to accelerate the production of printed circuit boards (PCBs) produced some excitement across industries. The move highlighted the critical role PCBs play in national defense, energy and healthcare and was seen as a corollary of sorts to the CHIPS Act enacted last year. In the words of President Biden, “action to expand the domestic production capability for printed circuit boards and advanced packaging is necessary to avert an industrial resource or critical technology item shortfall that would severely impair national defense capability.”

In parallel, a second advancement spawned even greater enthusiasm: the emergence of generative artificial intelligence (AI), the technology exemplified by ChatGPT. The intersection of these developments—the urgent need for PCBs coupled with advanced AI technology—has created a unique opportunity for the domestic manufacturing of PCBs and electronics in general.

As its name suggests, generative AI involves the generation of novel and useful assets by an AI system. ChatGPT, the now well-known tool from OpenAI, utilizes the technology to generate content that human beings find convincing while pharmaceutical companies have leveraged generative models to design novel molecules for potential new drugs. Insilico Medicine, for example, has developed a platform called molecular activity predictor (MAP) that uses generative AI techniques to devise new molecules with potential therapeutic applications.

A natural question is how generative AI can be used for something as different as PCB manufacturing. In this context, the transformative impact of generative AI centers on data.

Generative AI And Synthetic Data

One key obstacle in applying AI to PCB inspection is the dependence on large, labeled datasets for training AI systems. Collating these datasets can be resource-intensive, especially because images of component defects—what’s sometimes termed “negative data”—can be difficult to obtain. What’s more, not all manufacturers possess the required hardware imaging capabilities to collect data in a format that’s suitable for AI applications.

Generative AI can address this shortcoming. By leveraging the technology, AI practitioners have been able to create synthetic data for PCB boards and defects with impressive detail. This data can then be used to train AI, enabling it to detect various PCB defects through visual quality inspection (VQI) systems, including surface mount technology (SMT) and through-hole technology (THT) anomalies. Faster inspection times give way to increased throughput and the more rapid production of PCBs—a key outcome envisioned by President Biden’s order.

A second potential benefit of VQI systems with advanced AI is the time required to configure new products (i.e., setting up new circuit board configurations that require analysis). As a VQI system has been trained on a vast dataset of labeled PCB images—portions of which have been synthetically produced by generative AI—the configuration process can be markedly simpler.

To operate, a VQI system uses a reference image of a “golden board,” which is a PCB without defects. Unlike automated optical inspection (AOI), configuring the system doesn’t require rules-based or code-based programming. Instead, a quality engineer enters the model number and the physical dimensions of the board, places a golden board in an inspection chamber, and the system captures images and automatically detects the board components. The quality engineer can then review and approve the board profile, at which point the system is ready to perform production-level inspections.

VQI systems can also adapt to manufacturing realities, such as board refinements and component swaps—typically without incurring downtime or complications. For example, due to material and component shortages, a manufacturer will sometimes stipulate that alternative components may be used on a given product (e.g., an alternate resistor). To this end, an alternate component feature allows for the “union” of multiple golden boards, by which the AI can automatically determine if a new component is an acceptable alternative to the standard.

On the manufacturing floor, the VQI system works alongside human operators to achieve levels of performance that neither could attain alone. Such systems have been successfully implemented in several leading original equipment manufacturer (OEM) and electronics manufacturing services (EMS) organizations, yielding rapid returns on investment due to increased throughput, improved product quality, reduced labor requirements and decreased scrap.

Creativity For The National Good

Although generative AI has captured the public imagination with impressive feats such as simulating human-like conversations and designing new pharmaceutical molecules, it has equal importance on more functional matters related to the national good, such as contributing to stronger and more resilient domestic capabilities for electronics manufacturing.

As Einstein once said, “Creativity is intelligence having fun.” By tapping into the creative potential of generative AI, we can help assuage our national security and infrastructure concerns while paving the way for innovative solutions across a range of industries. The application of generative AI to electronics manufacturing represents an initial step in this continuum, illustrating the boundless possibilities when human ingenuity is joined with cutting-edge technology.

Follow me on Twitter or LinkedIn. Check out my website. Sheldon Fernandez

Leave a Reply

Your email address will not be published. Required fields are marked *