The 8-Step Inspection Checklist: Catching Thermal and Electrical Failures Before They Cost You a Batch

The moment my “OK” was wrong

In 2022, I signed off on an 1100 HPLC pump that looked completely healthy. The system pressure was stable, the leak sensor was dry, and the run logs were clean. Two days later, a gradient batch failed when that same pump head went into thermal shutdown. The repair bill came to about $3,400. The discarded batch records cost more than the repair because we could no longer verify the data.

The pump didn’t fail without warning. It had been running warmer for weeks, but nobody was looking at temperature. That’s when I stopped trusting “looks fine” and started building the checklist below.

Use this list when you are responsible for equipment that runs unattended—an aging HPLC, a compressor, a motor control center, anything where a slow temperature change is the first sign of trouble. It has eight steps. Most of them take five minutes or less.

The 8-step checklist

1. Write down what “normal” means before you point a camera

Thermal readings are only useful if you know the operating state. A pump scanned during standby can look completely different from the same pump running at full load. If you don’t record the conditions, next month’s comparison is guesswork.

Before I touch a lens, I write down the process setpoints: flow rate, motor speed, column or oven temperature, ambient temperature, and how long the equipment has been running. For the older 1100 HPLC units in our lab, I let the system pump at 1 mL/min for at least 20 minutes before scanning the pump head and heater. That gives the system time to reach its true running temperature.

“If you didn’t write down the load, that number means nothing.” It took me one failed batch to understand he wasn’t being dramatic.

2. Check your own tools first

This step sounds too obvious, which is exactly why it gets skipped. A meter with a low battery, wrong input jack, or expired calibration gives you a precise-looking wrong answer. The same goes for a camera with a dirty lens or a focus set for the wrong distance.

On the FLIR digital multimeters I use in our shop, I check the battery icon and the input terminals before measuring anything. On the FLIR thermal camera E6, I clean the lens, set the focus, and confirm the instrument is within its calibration window. It takes one minute. It has saved me from chasing at least three faults that did not exist.

3. Don’t trust the default emissivity on bare metal

From the outside, a polished stainless steel clamp or an aluminum pump head looks like it is at the temperature the camera reports. The reality is often different. Bare metal is reflective, and the camera may be measuring my body heat, a nearby oven, or a cold window instead of the surface itself.

A simple fix is to put a piece of matte PVC electrical tape on the target and let it reach the same temperature. Emissivity of the tape is close to 0.95, which is what most thermal cameras expect by default. If I can’t add tape, I adjust the emissivity setting in the camera and record what I used. The point is to know what you are measuring, not just to capture a colorful image.

4. Use the same distance and angle every time

A thermal camera measures a spot, and the size of that spot grows with distance. If you stand one meter away this month and five meters away next month, you are not comparing the same area. The larger spot averages in more of the surroundings, so the reading drifts even when the equipment hasn’t changed.

I take my baseline images from the same angle, same distance, and same zoom every quarter. I also keep a rough distance note in the work order. This is one of those small habits that makes later data worth anything.

5. Pair the thermal reading with an electrical measurement

Thermography shows you the consequence. A digital multimeter shows you the cause. A hot circuit breaker might mean the connection is loose, the breaker is failing, or the circuit is simply running near its design limit. You don’t know which one until you measure.

This is where the FLIR digital multimeters earn their place in my bag. If I see a connection running noticeably warmer than an identical adjacent connection, I measure the current through both. If the currents are similar, the problem is usually resistance at the connection. If the current is higher on one phase, the problem is the load balance. Replacing a hot breaker without checking load is a good way to install the same failure twice—I’ve done it, and the callback cost more than the original repair.

6. Add an ultrasonic sensor for the failures that haven’t heated up yet

Heat is a late symptom for some failures. Partial discharge inside a closed electrical panel can make noise weeks before it makes enough heat to be obvious. A failing bearing can sing at high frequency before the housing temperature rises. That’s why I carry an ultrasonic sensor—basically a handheld detector that lets you hear high-frequency sounds, often called an ultrasonic detector.

For motors and pumps, I press the probe against the bearing housing and compare the sound with a healthy unit of the same model. A noisy bearing usually sounds rough or sizzling compared with the clean hum of a good one. For enclosed electrical gear, I scan the outside of the panel and listen for the hiss or crackle of partial discharge. It is not a replacement for thermal imaging; it’s a complement.

7. When a part looks worn, measure it with a caliper

Your eye cannot reliably see 0.02 mm of wear, but a caliper can. If a thermal scan points to a pump or motor problem, I wait until the equipment is locked out, then measure the suspect part directly. For example, a pump piston that is worn on one side will show up far better in three caliper measurements than in a visual inspection.

People in our shop have strong opinions about Mitutoyo vs Starrett calipers. In practice, both brands give repeatable results when they are calibrated and handled properly. The bigger issue is dropped tools, expired calibration certificates, and measurements taken in a hurry. I normally reach for a Mitutoyo caliper, and I keep a Starrett as a second-opinion tool when a reading seems odd. The brand matters less than the habit of checking the calibration date first.

8. Re-image after the repair and save the comparison

The job is not finished when the new part is installed. It’s finished when the repaired equipment shows a healthy thermal signature under the same load. If you don’t re-image, you never actually confirm that the replacement solved the problem.

I save before-and-after images in a folder for each piece of equipment, with the load conditions and emissivity noted. It takes a little extra time, but it gives you a direct answer the next time someone asks, “Has this always run this warm?” If you don’t record the settings, a thermal image is just a nice picture.

A few warnings from the school of hard knocks

First, never take a mechanical or electrical measurement on a live or rotating component unless you are qualified and following your site’s lockout/tagout procedures. No checklist is worth a safety incident.

Second, don’t inspect only the component that failed. When a pump bearing fails, the motor coupling, the shaft seal, and the electrical load can all be involved. Fixing the bearing without checking the alignment just schedules the next failure.

Third, compare trends instead of absolute values. A component may be within its spec limit but still 8°C warmer than it was last quarter. That trend matters. NFPA 70B, the electrical maintenance standard, treats infrared inspection as a core verification tool because it finds problems that do not show up in routine electrical tests.

It took me a few expensive mistakes to understand that prevention is cheaper than correction. A five-minute temperature check before a batch is far less painful than a three-day investigation after one fails. The checklist doesn’t make inspections exciting, but it makes them repeatable. And repeatable is what keeps equipment running.