Units
1. Introduction and Limits of AI in Electrical and Electronics Engineering 2. Circuit Analysis and Design Support 3. Embedded Systems and Microcontroller Code 4. Signal Processing and Data Analysis 5. Control Systems Design Support 6. Power Systems and Energy Analysis 7. PCB Design and Hardware Documentation 8. Automation, PLC Logic and Sensor/IoT Data 9. Standard Research, Simulation and Accountability with Python
Unit 1 / 9

Introduction and Limits of AI in Electrical and Electronics Engineering

Gains:

  • Ability to distinguish where AI is accelerating and where it is risky in circuit, power and embedded system work
  • Ability to explain why verification and engineer approval is mandatory in electrical works involving life and property safety.
  • Ability to position AI output as a draft input to be audited, rather than as a replacement for engineering judgment

When you watch an electrical and electronics engineer's day, the picture that emerges is similar in most teams: reading datasheets, filling out spreadsheets, writing and debugging code, drawing schematics, searching for standards, preparing test reports, e-mail and meetings. In other words, the time devoted to the actual "engineering" part, that is, the judgment of designing a system safely and accurately, is crushed under repetitive work. This is where artificial intelligence comes into play. AI doesn't do the design for you; It prepares you for design, produces drafts, speeds up calculations and puts processed information in front of you for decision-making. In this unit, we will clarify at which stages of the electrical-electronic workflow AI adds real value, which decisions should definitely remain with the engineer, and the safety discipline you must adhere to when doing this. The goal is clear: to position AI not as an “automatic designer” but as a disciplined assistant whose output is verified every time.

In Which Businesses Is AI Accelerator, In Which Businesses Is It Risky?

Electrical and electronic works are located on a two-pronged spectrum in terms of their results. At one end are revocable, low-risk office jobs; On the other end, there are irreversible decisions that directly affect the safety of life and property. The value of AI varies depending on where you stand on this spectrum.

business type

AI contribution

Engineer's role

Datasheet/standard summary

Remove relevant item from long document

Verify value from official source

Code draft (MCU, Python)

Skeleton and function generation

Register, timing, test verification

Account preliminary work

Formula setting, first number

Unit, rank and standard control

Documentation

Report/BOM/test procedure draft

Technical accuracy and terminology

protection coordination

suggestion, scenario

Final selection and approval

commissioning

Checklist draft

Measurement and signature in the field

The rule is simple: the risk of an AI output equals the damage that output will cause if it makes a mistake. Misspelling an axis label is harmless; Miscalculating a breaker current setting can result in a fire. So the first question to ask before using the output is: "What happens if this is wrong and who will notice?"

Attention: AI produces fluent and confident text. Fluency is no guarantee of accuracy. A language model can "concoct" (hallucinate) believable formulas, item numbers, and datasheet values ​​even when it has no real data. In electricity, this does not remain on paper; turns into malfunction, damage or accident in the field.

Decisions That Should Be Left to the Engineer

Some decisions should never be fully automated. These pose technical, legal and ethical risks:

  • Safety-critical choices: Breaker/fuse rating, protection coordination, grounding and insulation class.
  • Commissioning and energizing: The decision to apply voltage to a board, motor, or board's power line for the first time.
  • Standard conformity declaration: The statement "This design complies with the relevant standard" requires the engineer's signature.
  • Values ​​in contact with humans: Residual current protection threshold, touch voltage, EMC/EMF limits.
Warning: Even if an AI-generated calculation results in something like "50 A breaker is sufficient", it is unacceptable to apply it without an engineer verifying it with short-circuit current, selectivity and cable strength. Each safety-critical output must be independently verified and approved by the authorized engineer before implementation.

Verification Discipline: Three-Layer Control

To use AI output with the eyes of the editor and auditor rather than blindly, apply a three-layer control:

  1. Unit and rank (sanity check): Is the unit of the result correct? Is the size physically reasonable? If there is 500 kW loss in a 100 kVA transformer, there is a fault somewhere.
  2. Independent reproduction: Reproduce the calculation manually, with a calculator, or with a short Python/SPICE run. If two different methods give the same result, trust increases.
  3. Standard and source verification: Every item number, datasheet value and limit given by AI must be confirmed verbatim from the official source.

Verification prompt (makes it easier to check the output): "Clearly list ALL assumptions, formulas, and units you used in calculating the following. Show each intermediate step on a separate line. Mark the standard value or datasheet number you used with a 'source required' tag; do not make up any numbers you do not know for sure."

Weak Prompt / Strong Prompt

WEAK:"Tell me the cable cross-section suitable for this motor."(Result: a single number without context; it is unclear which current, length, mounting method and voltage drop limit are assumed.)STRONG:"Preliminarily calculate the copper cable cross-section for the following conditions and show the steps:- Motor: 15 kW, 400 V, three-phase, cosφ 0.85, efficiency 0.90- Cable length: 60 m, conduit installation, ambient 35 °C- Allowable voltage drop: 3%Calculate current, voltage drop and thermal current carrying capacity SEPARATELY. Write down the unit at each step. Note the need to verify the result with the relevant standard table; do not make a definitive choice statement.

The difference is that the strong prompt gives both the actual input and the boundary (make no declaration, show steps) to the model. This makes the output auditable.

Mini Case

Embedded systems engineer Deniz prints the ADC read code for a temperature measurement card to the AI. AI establishes the formula assuming the reference voltage for 12-bit ADC is 5 V. Deniz opens the datasheet of the card; the reference is actually 3.3 V. AI's assumption of 5 V would have each measurement read with about 50% bias. The sea fixes the reference, uploads the code to the card, and compares the reading to the actual thermometer at a known temperature. AI gave speed; The sea added credibility by verifying the assumption and confirming it with measurement. If AI were blindly trusted, the error would be carried to the field.

Common Mistakes

  • Accepting AI's assumptions (reference voltage, temperature, conductor type) without questioning.
  • Applying safety-critical values ​​(breaker, protection, grounding) without verifying.
  • Not noticing made-up datasheet numbers and standard item numbers.
  • Assuming the code "works" without measuring it in hardware.
  • Bypassing volume control; Overlooking the V, A, W, VA, VAR confusion.
  • Substituting AI output for engineering approval.
Tip: Ask each AI output these three questions: (1) What assumptions did it make that conclusion, and are the assumptions correct? (2) Have I reproduced this in an independent way? (3) Is this security-critical, and if so, have I received authoritative verification and approval? If all three are clear, proceed.

In summary

  • AI adds speed and scale in the preparation, calculation, code and document stages of electrical and electronic works; but he is not the decision maker.
  • The risk of an outcome is equal to the damage it will cause if it makes a mistake; Safety-critical work requires independent verification and engineer approval.
  • Fluid text is not correct text; Every assumption, number and standard value must be verified.
  • Three layers of control (unit/rank, independent reproduction, source verification) make the output reliable.
  • The powerful prompt gives the actual input and the boundary together to the model.
  • Position AI as “a fast assistant whose every output is monitored.”

Application task

Have the AI ​​solve a calculation you recently made (cable cross-section, resistor selection, power loss, etc.) from scratch. Then extract each assumption and number in the output into a list; decide "verified / resource required / faulty" for each. Reproduce the calculation yourself, either manually or with a short Python run, and compare the two results. If different, find out why. This exercise builds the habit of using AI with auditor discipline.

Next