Predicting Electrical Failures Before They Start
By Michelle Fiscus, Senior VP & Chief Communications Officer
Most people don’t think about the electrical systems quietly powering the modern world. When one of these systems fails, the impact can extend far beyond the equipment itself. It can halt production, increase costs, and disrupt critical operations. At the University of North Carolina at Charlotte, electrical engineering researcher Jim Gafford, Professor of Practice and Associate Director of Applied Research at the Energy Production and Infrastructure Center (EPIC), is focused on a problem most people will never see, but one that affects nearly everything they depend on: how to detect failure before it happens.
Gafford’s journey into this field started early. “I was introduced to power electronics as an undergraduate electrical engineering student, and I’ve remained inspired by the multidisciplinary nature of the field,” he said. “In engineering, you bring together analog circuits, semiconductor physics, electromagnetics, embedded systems, and thermodynamics. You can’t get bored with power electronics—there’s always something new to learn.” That curiosity followed him into graduate school, where he studied wide bandgap semiconductors, an emerging area at the time. Today, those devices are becoming commercially viable and are essential to systems enabling renewable energy, advanced manufacturing, and the growing needs of a more electrified economy. “The capabilities these advancements can deliver will drive entire economic systems,” Gafford said. “From inverter-based resources to high-voltage transmission, more reliable and efficient power electronics create a more resilient grid.”
As these systems scale, so do the stakes. What works in a lab doesn’t always translate to real-world conditions. “Laboratory capability of power electronics systems is trivial compared to operating at industrial scale,” Gafford explained. “Critical power systems must provide extremely high reliability, and achieving that requires new methods to weed out components with shorter lifespans.” When systems fail, the consequences are immediate and compounding. “At the most basic level, when these systems fail, productivity stops,” he said. “That creates a cascading impact on cost. Improving reliability increases output and reduces expenses.”
The challenge is that many of these failures don’t start with obvious breakdowns. They begin as faint, nearly invisible signals—subtle changes in complex electrical behavior that traditional testing methods can’t detect. Gafford’s research focuses on identifying those signals early. His team is developing a metrology and data acquisition system that captures high-precision measurements from devices as they operate. They are building detailed models that can predict when something is likely to fail. “We’re creating a system that generates characteristic models of devices,” he said. “By measuring faint signatures in real time, we can assess state of health and predict the likelihood of early-life failure.”
This approach marks a fundamental change from how most systems are currently tested. Traditional methods rely on binary pass-or-fail outcomes, often based on isolated measurements. Gafford’s work expands beyond that. “We’re moving past a simple pass/fail framework to understand how devices behave under combined stress conditions,” he said. “The reality is that failures often arise from multiple interacting factors that are not visible in traditional testing.”
Those factors can include issues like thermal cycling, where repeated heating and cooling gradually degrade materials, or internal defects such as substrate delamination and wire bond failures. These problems can exist long before they lead to a system failure and are notoriously hard to detect. “The precursors for these failure mechanisms require looking beyond first-order characteristics,” Gafford said. “Historically, that level of analysis has only been possible in controlled laboratory environments. Our goal is to extract that same level of insight in real operating conditions.”
To achieve this, his team is developing specialized instruments capable of capturing extremely precise voltage and current measurements, even in high-power environments. As more data is collected, the system becomes smarter, paving the way for advanced analysis and machine learning to identify patterns in complex datasets. “Machine learning could be a game changer,” Gafford said. “The ability to identify faint signatures in complex data is already proven. As we build larger datasets, we’ll uncover patterns that we can’t predict today.”
This is where NCInnovation plays a critical role—not just by funding the research, but by helping bring it outside the lab and into the real world. For Gafford and his team, the challenge is to prove the science and show that it works in complex manufacturing environments while delivering measurable value.
“NCInnovation provides resources, networks, support, and motivation to address commercialization potential for applied research,” Gafford said. “The funding speeds up development through both technology and pre-commercialization challenges.”
That support is shaping how the technology is being created. Instead of developing a system solely for academic validation, the team is designing prototypes with manufacturing integration in mind. They specifically target areas where the technology could fit into existing workflows, such as final functional testing or device burn-in stages.
“By aligning commercialization with technology development, the questions we’re asking inform both processes,” Gafford said. “That shortens the timeline for deployment and creates opportunities for greater commercial viability.”
The next phase of the work focuses on validating the value proposition with partners across the industry—from component manufacturers to system integrators—ensuring the technology solves real problems in real environments.
The long-term vision goes beyond testing. Gafford envisions a future where power electronic systems can monitor themselves in real time, continuously assessing their own health. “If we can reduce the size and cost of this technology, it could be integrated directly into devices,” he said. “That would allow critical systems to self-monitor and predict failures before they happen.”
That kind of capability could have far-reaching implications across industries—from advanced manufacturing to renewable energy to large-scale electrification efforts. It could reduce downtime, lower costs, and increase confidence in the systems that enable modern life.
North Carolina provides a unique environment to advance this work. With strengths in power generation, transmission, semiconductor manufacturing, and applied research, the state offers a full ecosystem for energy innovation. “Electric power generation, transmission and distribution, power conditioning products, and power semiconductor devices all have significant presences in North Carolina,” Gafford said. “The technology we’re developing impacts that entire spectrum.”
For Gafford, the ultimate goal is not just better technology, but more reliable systems at scale—systems people can depend on without ever having to think about them. “The reliability we expect from power systems doesn’t happen by accident,” he said. “It requires intentional design, careful engineering, and a deep understanding of how every component behaves within the larger system.”
If his work succeeds, the benefits will extend far beyond the lab—to manufacturers, utilities, and ultimately anyone who depends on reliable, affordable power. In a world increasingly driven by electricity, that kind of reliability isn’t just an engineering challenge. It’s a foundation for everything that comes next.