AI has crept into a lot of things that are now suddenly referred to as “electronic,” including your smart thermostat that supposedly “remembers” your temperature preferences, that factory plant manager who tells you how breakdowns are now avoided thanks to predictive maintenance, the electricity utility that boasts of having an automated “smart grid” that is capable of shifting energy consumption around at a moment’s notice, and your earbuds that have a surprising number of calculations to perform while you’re listening to your music.
It’s as if AI appeared out of nowhere and made the electrical parts of the world its own.
Here’s the reality. AI did not supplant electrical engineering. It hitched a ride. It takes the parameters that engineers are most interested in – voltage, current, temperature, vibration, timing – and finds the patterns that people can’t easily see when they’re tired, distracted, or looking at five screens at once. And then it suggests to a human or a machine what the best course of action might be. That’s it. No magic necessary.
This article is intended to provide a clear roadmap of the intersection of AI and electrical engineering – no hype, no math, just enough explanation so you don’t feel like everyone else got the memo and you didn’t.
And a gut-check test at the end is straightforward. If someone comes to you with an “AI-powered” something, without cocking an eyebrow can you at least ask what the three questions are?
What does it even measure? What does the AI attempt to detect or predict? And what consequence occurs when it thinks it has learned something?
If those questions are straightforward to answer, it’s probably a real technology. If not, you’re probably dealing with a buzzword.
What we actually are referring to as electrical engineering and AI
Take electrical engineering, for example. It sounds like a daunting field, but what does it boil down to? Electrical engineers manipulate electricity to make useful things. This can be something huge, like making sure a city stays lit, or something tiny, like making sure your phone doesn’t overheat in your pocket.
I tend to divide it into four basic categories, not just because textbooks do it that way, but it’s just easier for my brain to organize.
Power is a matter of production and distribution. The grid, solar panels, batteries, chargers. High current, high stakes.
Electronics refers to the hardware. Circuits, sensors, chips, boards. All the physical components that actually enable devices to be a thing.
Control is about controlling things. Motor speeds, temperature loops, safety interrupts. The thing that says “Whoa, slow down, you’re gonna melt the CPU.”
Communication refers to the method by which devices communicate with each other such as Wi-Fi, Bluetooth, industrial protocols, etc.
AI is not, as its name suggests, an intelligent being figuring out what to do. As used in virtually all actual commercial offerings, it is a software tool that has been trained on examples to perform a few narrowly defined tasks. It tells you that something looks OK or it doesn’t. It warns of an impending failure, or a potential demand surge. Sometimes it goes further and suggests what to do, or even does it, issuing an alarm or tweaking a control.
With this perspective, it’s not so surprising that there’s a lot of overlap. Electrical engineering develops systems that generate signals and data. AI is just a technology that analyzes that data to make decisions faster and more reliably, without requiring that someone has a good day. In my perspective, that’s not a takeover. It’s just the next technology we’ve developed to manage complexity, to keep ourselves from going insane.
Areas in which AI can actually be useful in electrical engineering
AI has value in electrical engineering when there’s a lot of noise. A lot of data. A lot of uncertainty. A lot of times when a tiny issue creeps up and nobody catches it until it becomes a big issue. Humans are great at handling uncertainty. We’re awful at simultaneously monitoring thousands of signals. AI fills that niche.
If you want to know where it’s actually helpful – not where it makes for a good slide deck – here are some of the sweet spots I keep stumbling upon.
Monitoring and anomaly detection
The power grid talks constantly. Volts and amps and Hz and heat and hum and harmonics. Never stops, and much of it looks perfectly dull… until it doesn’t. Old-fashioned alarms ring if something crosses a red line. AI rings alarms if it thinks it sees that something is getting ready to cross a red line that nobody drew.
Where this works:
- predicting power quality problems to avoid consumers experiencing flicker and dropouts
- early detection of unusual heat or vibration in motors and transformers
- identifying anomalous industrial behavior that is “within limits” but nonetheless seems suspicious
This is one of those scenarios where the AI is like an old operative who says “I don’t know what it means, but something doesn’t seem right here”.
Anticipating failures before they happen
The words “predictive maintenance” get thrown around a lot these days, perhaps too often. Yet when I think of it in the truly meaningful sense, I am looking for changes in the little things. A waveform is a little different. A temperature is a little higher. The amount of current a motor uses at a certain point under the same load is a little higher.
Other valid use cases might be:
- Detecting motor or bearing faults from current signatures
- Estimating the health of a transformer based on its temperature and loading history
- Prognostics of state of health and remaining useful life of the battery
This isn’t predicting the future. It’s just pattern matching based on lots of experience. I think this is where AI can save the most with the least controversy.
Energy efficiency and performance
There are so many trade-offs in electrical engineering – cost versus reliability, efficiency versus convenience, performance versus heat, etc. – and AI can help manage all of these trade-offs without requiring direct supervision.
Here are some of the typical wins:
- We are shaving peak demand charges by shifting loads in the background without any fuss
- Making HVAC systems smarter so buildings still feel right, but use less energy
- How to balance solar, batteries and grid electricity use, without tweaking every 7 days?
If it’s going well, there’s no celebration. Bills just go down. The system becomes less stressful. That’s usually an indication that it’s working.
Understanding and controlling large complex systems
With one charger, there’s no problem. With 100, you need a meeting. With 10,000, you’ve got a migraine. With that many moving parts, even the simplest choices become a nightmare.
AI is what ties it all together:
- controlling electric vehicle (EV) charging units so they don’t cause a circuit overload during the dinner hour.
- identify unusual smart meter usage trends
- and quickly identifying faults and restoration options for grid operators.
In the early days, AI isn’t so much about being brilliant as it is about being solid. It handles the drudge and logistics that humans don’t want to bother with.
Electronics inspection and quality control
Manufacturing is not very forgiving: Small flaws today could lead to massive problems down the road, and humans have limited patience to stare at boards all day. AI-enabled computer vision doesn’t get bored, and it doesn’t speed up to finish the last batch before lunch.
The best part:
- for inspection of solder joints, missing components or misregistration
- identifying surface cracks and other minor flaws in fabrication
- lowering returns by detecting problems at an earlier stage, where it’s less costly to correct them
This is one of the places where AI seems almost cheating, in a good way. It just keeps searching.
The bottom line
AI is most effective when it acts as what it really is – a pattern recognition and decision support tool layered on top of good instrumentation, engineering judgement, and basic control systems. The problems arise when AI attempts to supplant all that. When it supplements, with steady, reliable operation, it has a habit of quickly establishing confidence.
