7 'Maker' Components You Should Never Install on the Plant Floor (and Their Industrial Equivalent)

The maker ecosystem is the best automation school there's ever been. We say that with full knowledge of the fact: Industrial Shields exists because Arduino, ESP32, and Raspberry Pi proved that open electronics could do serious work.

But there's a line that gets crossed far too casually: the one separating a prototype that works on the bench from an installation that has to run 24/7 for years. And some components simply cannot cross it. Not because they're "bad" — they do their job perfectly well for teaching and prototyping — but because they were designed for an environment (stable temperature, no vibration, no electrical noise, a human standing right there) that is the exact opposite of a plant floor.

These are the 7 components we've found most often inside failed control cabinets that land on our support desk. For each one: exactly what fails, why the failure doesn't show up on the test bench, and the industrial alternative. If you're running Arduino in an industrial environment and care about reliability, this list is your pre-installation audit.

#1 — The DHT11 / DHT22 Sensor

What fails. Everything, eventually. The DHT11 has an accuracy of ±2 °C and ±5% RH; the DHT22 improves on paper, but shares the same structural flaws: an exposed sensing element in an unprotected plastic housing, slow response, and a proprietary single-wire protocol — not differential, with zero noise immunity. In an environment with dust, condensation, or vapors, the humidity sensor's drift is a matter of months, and it doesn't warn you: it keeps outputting numbers, increasingly wrong ones.

Why the prototype fools you. On the bench it reads "fine" (nobody checks it against a reference standard), the cable is 20 cm long, and there are no VFDs nearby.

The industrial alternative. For temperature: Pt100/Pt1000 with a 4-20 mA transmitter, or probes with Modbus RTU output. For humidity: industrial transmitters with a replaceable probe and electronics separated from the measurement point. The 4-20 mA current loop exists precisely to carry a measurement across a noisy plant floor: the signal is current, unaffected by voltage drop, and a broken cable self-detects (0 mA = fault, not "zero degrees").

#2 — Dupont Jumpers and Breadboards

What fails. The contact. A Dupont jumper holds by friction, with no mechanical retention. Under the vibration of a running machine, every jumper is a potential intermittent contact — and an intermittent contact is the most expensive fault to diagnose there is, because it disappears the moment you show up with a multimeter and reappears the moment you leave. The breadboard adds its own problems: variable contact resistance and parasitic capacitance between rows.

The industrial alternative. Screw or spring terminals, cable with a crimped ferrule (never tin-coated wire straight into a screw terminal: solder cold-flows and the connection loosens on its own), torque per manufacturer spec, and a re-torque after the first weeks of real-world vibration. Every conductor needs two points of retention: electrical and mechanical.

#3 — Unbranded Relay Board Modules

What fails. Not the relay itself — the nonexistent engineering around it. On the blue marketplace board, the isolation between coil and contacts, the creepage and clearance distances, and the actual switching capacity haven't been verified by anyone. The "10 A 250 VAC" printed on the silkscreen is for pure resistive load; switch a real inductive load (contactor, solenoid valve, motor) without suppression, and arcing erodes the contacts until service life is measured in weeks. And when the cheap relay fails, it can fail welded shut (the load never turns off) or with compromised isolation.

The industrial alternative. Brand-name relays with real datasheets (Omron, Finder, Schneider…) mounted on DIN-rail sockets, with proper suppression (flyback diode on DC, RC snubber or varistor on AC) — or, better yet, integrated, protected relay outputs on the PLC itself, sized and documented for inductive loads.

#4 — Uncertified Power Supplies

What fails. By far the component that takes the most equipment down with it. The three-euro 24 V supply has no real surge or transient protection, its ripple is a lottery, its behavior under inrush current from connected loads is unpredictable, and its "CE" markings often never saw a lab. When it fails — and it will — it doesn't fail alone: it takes down whatever electronics it was powering, and in the worst case, lets mains voltage through to the 24 V side.

The industrial alternative. A DIN-rail supply from a recognized manufacturer (Mean Well, Phoenix Contact, Weidmüller…), with genuine compliance documentation, sized with margin for inrush rather than average consumption. And a power hierarchy: a separate supply (or at least a protected branch) for control electronics versus actuators that draw current spikes.

#5 — Consumer-Grade microSD Cards

What fails. Flash memory has a finite number of write cycles. A consumer card (TLC, with no industrial wear leveling) writing logs every second exhausts those cycles within months. And it doesn't degrade with a warning: one day the system won't boot, and the filesystem is corrupted. If your controller runs Linux and boots from the SD card, this isn't an inconvenience — it's a scheduled machine outage, built in by design.

The industrial alternative. Industrial pSLC/SLC cards with extended temperature range and serious wear leveling; a filesystem and mount configuration built to minimize writes (logs in RAM with periodic flush, read-only partitions); and the data that actually matters sent to a remote server or database, not left on the card. The local SD card is a cache, not an archive.

#6 — The Phone Charger as a Permanent Power Supply

This is the one we promised in the newsletter: it comes "factory installed" in thousands of prototypes running on plant floors right now — and it's a countdown clock.

What fails. The prototype was developed powered over USB, it worked, and nobody ever revisited the power supply: it got installed with the same phone charger it was programmed with. A consumer charger is designed to charge a phone a few hours a day in a 22 °C room — not to power control electronics 24/7 inside a cabinet at 45 °C. Its electrolytic capacitors, running hot without rest, dry out at an accelerated rate: ripple grows month after month until the CPU starts resetting "for no reason." It's a literal countdown: it doesn't fail in month one, it fails a year in, once nobody's watching the project anymore. Add a micro-USB connector held in by friction in a vibration-heavy environment, and you've got component #2's failure mode thrown in for free.

The industrial alternative. The same as #4 — a certified DIN-rail supply — plus equipment that accepts 12-24 Vdc from the plant with protection built into the power input (reverse-polarity, transients), wired through terminals, not a phone connector.

#7 — The Cheap or Counterfeit SSR

Last on the list, and the only one that isn't (just) a reliability problem: it's a safety problem.

What fails. Marketplace solid-state relays — including the mass-produced counterfeits of known brands that have circulated for years — hide a triac rated well below what's printed on the case, with insufficient heat dissipation. Switching the "rated" real-world load (a heating element, for example), the triac operates out of spec. And the typical failure mode of an overloaded triac is short circuit: the load stays permanently energized. A heating element that can't be switched off isn't a malfunction — it's a fire waiting to happen. On top of that, an SSR failing shorted provides no real separation, and the "isolation" claimed by a counterfeit has been verified by no one.

The industrial alternative. An SSR from a legitimate manufacturer, bought through an official distribution channel, sized with serious derating (rule of thumb: double the load current), a properly calculated heatsink, and adequate upstream protection (ultra-fast fuse if the SSR calls for one). And for critical heating loads: a cutoff element independent of the control loop — a safety contactor or a manual-reset thermostat — because no switching element, however good, has the right to be the only thing that can turn off a heater.

The Summary Table

The Common Pattern (and the Honest Conclusion)

Notice that none of the seven fails on day one. They all share the same profile: they work in the demo, and fail under temperature, vibration, electrical noise, and accumulated hours — exactly the four variables the test bench doesn't have. That's why the reliability of Arduino in an industrial environment isn't decided at the microcontroller (the ATmega, the ESP32, or the Raspberry Pi's Broadcom chip are all remarkably reliable) — it's decided by everything around it.

That is, quite literally, the reason Industrial Shields exists: take the open-source electronics you already know and put it in industrial form — optocoupled inputs, protected and properly sized outputs, 12-24 Vdc power with protection, real terminals, DIN rail, and genuine CE marking. Your prototype's code stays exactly as it is; the seven components on this list don't.

Download the checklist: "From Arduino prototype to industrial installation"
The full component-by-component audit — including these 7 — plus the wiring, power, and communications checks, in a printable PDF.
👉 Download the free checklist →

And if your prototype runs on Arduino, ESP32, or Raspberry Pi, the natural migration path is to the industrial PLC from the same family, with no code rewrite: https://www.industrialshields.com/shop