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

Geopolitical tensions and pandemic-related disruptions have revealed deep vulnerabilities within the supply chain for manufacturers as well as the businesses and governments that rely on them. Pcb And Pcba

In particular, the well-documented global shortage of semiconductors, printed circuit boards (PCBs) and other essential electronic components have limited the production of everything from automobiles to medical devices and critical infrastructure.

To mitigate risk, manufacturers are exploring ways to increase the efficiencies by which such components are produced. In parallel, recent legislation such as the CHIPS Act has endeavored to boost domestic semiconductor research and manufacturing.

Although most of the commentary is focused on semiconductors, the CHIPS Act also applies to PCBs and complements industry standards such as IPC-1791 to combat poor quality and counterfeit components.

Technology and humans work within a complex process to manufacture PCBs, which are ubiquitous and embedded in everything from microwaves to aircraft. For a myriad of reasons (including different board sizes, component diversity and complicated geometries), the inspection of PCBs is an intense and laborious task that necessitates specialized skills. For example, coating a board to shield it from environmental fluctuations—"conformal coating" in PCB parlance—can give rise to an assortment of defects that are painstakingly difficult for a human to identify.

Even with proficient personnel, manual inspection can contribute to the high cost of poor quality (COPQ) due to rework, high scrap and the additional expense of uncovering failures later in production. Moreover, as experienced inspectors retire, few qualified workers will be available to replace them. Data from the Labor Bureau reveals there were 846,000 job openings in the U.S. manufacturing sector in August 2022, and Deloitte forecasts the number will rise to more than 2 million by 2030.

Good intentions notwithstanding, no amount of funding, incentives, tax rebates or recruitment videos can close this gap in the short to medium term. Likewise, the ambitious production increases envisioned by the CHIPS Act are far from fruition, as new facilities will take years to realize meaningful outputs.

To mitigate these challenges, manufacturers need to make the most of disruptive and transformational technologies.

The latest developments in artificial intelligence have given rise to highly automated, human-in-the-loop visual quality inspection (VQI) systems that can outperform automated optical inspection (AOI) incumbents.

Such systems work as follows.

1. A human operator configures the platform by means of a "Golden Board" that serves as the ground truth for all inspections. This takes minutes, in contrast with the painstaking programming required by AOIs, which can take days or weeks.

2. One or more cameras capture multiple images of a given PCB.

3. An AI engine examines the images and identifies defects in an order of magnitude faster than human operators (e.g., 20 seconds).

4. The AI presents its findings to an operator via an intuitive user interface. In addition, explainable AI (XAI) can highlight the reason why a particular anomaly was classified as a defect.

5. The operator validates the AI's decision or overrides it, with each outcome nudging the AI to higher levels of inspection accuracy.

The result is a more accurate, highly automated and nondestructive inspection process that can amplify the impact of a small number of human experts while lessening the cognitive burden placed upon them. Moreover, the image and accompanying data for each PCB can be archived to perform powerful analytics.

One key obstacle in implementing modern AI systems is the dependence on training data as well as the necessary overhead and effort to collate the large, labeled datasets that machine learning systems require. This is especially true in manufacturing contexts, where images of component defects are scarce and difficult to obtain.

In recent years, however, proprietary techniques—including those of my own organization—have made it possible to train VQI systems using vastly reduced amounts of data. Thereafter, the feedback loop kicks in, continually training the system and refining its capabilities. Accuracy rapidly catches up to that of a highly skilled inspector before surpassing the limits of human expertise.

The AI, however, isn't the only actor that becomes smarter, as the system equips the manufacturer with a constant stream of data that can be fed into analytics tools to assist with root cause analysis efforts to implement process and design optimizations.

Although the benefits of VQI systems are straightforward, deploying them to production may be less so, given the relative adolescence of AI in manufacturing. To this end, it is important to consider how organizations can begin the transition to this powerful technology.

First, technology leaders should obtain organizational buy-in by focusing on quick wins. Executing "low-hanging fruit" pilots with concrete ROI can illustrate the economic benefits of such solutions before championing them more expansively.

Second, they should get ahead of the weighty enterprise requirements (IT, security, governance) that will inevitably materialize when such systems are moved to production, with special consideration for the unique aspects of AI-based solutions. For example, the non-deterministic aspects of machine learning models often necessitate new and updated infrastructure technologies for versioning and stress testing.

Finally, technology leaders should weigh the pros and cons of the "build vs. buy" approach. While some manufacturers might have the expertise to source an AI engine and do the rest themselves, others might benefit from an off-the-shelf solution that can be bolted—literally and figuratively—to an existing production line. While deployment rapidity will be paramount for some organizations, others will favor fostering in-house teams with a long-term view to maintaining certain competitive advantages.

After years of promise, AI is transforming production workflows in concrete ways and delivering meaningful ROI. Automated VQI can not only help manufacturers mitigate palpable risks in the short term but should also help manufacturers be strongly positioned to leverage transformational technologies in the future.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

I am a staff writer at Forbes covering retail. I have been at Forbes since 2013, first on the markets and investing team and then on the billionaires team. In the course of my reporting, I have interviewed the father of Indian gambling, the first female billionaire to enter the space race and the immigrant founder of one of the nation’s most secretive financial upstarts. My work has also appeared in Money Magazine and CNNMoney.com. Tips or story ideas? Email me at ldebter@forbes.com.

I’m a digital media executive recognized for building engaging content experiences that delight and inspire audiences. As Chief Product Officer of Forbes, I lead a team of world-class product managers, designers, e-commerce leaders, and other experts focused on building the products that shape the Forbes brand across the web, mobile, social, and emerging platforms.

I am an assistant editor based in Columbia, Missouri covering consumer technology. I graduated from the Missouri School of Journalism with a master's degree in magazine journalism and before that, got my bachelor's degree in investigative journalism. Before Forbes, I was a reporter and writer at Missouri Business Alert and Vox Magazine in Columbia, Missouri and the Ewing Marion Kauffman Foundation in Kansas City. You can reach me at rshrivastava@forbes.com

I am an assistant editor based in Columbia, Missouri covering consumer technology. I graduated from the Missouri School of Journalism with a master's degree in magazine journalism and before that, got my bachelor's degree in investigative journalism. Before Forbes, I was a reporter and writer at Missouri Business Alert and Vox Magazine in Columbia, Missouri and the Ewing Marion Kauffman Foundation in Kansas City. You can reach me at rshrivastava@forbes.com

EVP and CIO of Werner Enterprises. Read Daragh Mahon's full executive profile here.

Rigid Pcb Timothy Liu is the CTO and cofounder of Hillstone Networks. Read Tim Liu's full executive profile here.