Retrofit machine monitoring instruments equipment you already own from the outside: a split-core current clamp reads run state, a magnet-mount vibration sensor tracks health, a stack light tap captures status, a photo eye counts parts, and a gateway reads the existing PLC. Nothing on the machine is modified.
This is the hands-on companion to connecting machines without replacing them: the actual toolkit, what each device tells you, what the pieces cost, and the order to install them. If your floor spans four decades of equipment, read this alongside connecting mixed-vintage equipment.
What is retrofit machine monitoring?
It is machine monitoring achieved by adding sensing around a machine rather than replacing the machine or rewriting its controls. The defining property is non-invasiveness: a split-core clamp closes around a cable without cutting it, a vibration sensor holds on with a magnet or a stud mount on the frame, a light tap reads the stack light the machine already operates. The machine does not know it is being watched, which is exactly why the approach is fast, cheap, and safe for warranties and certifications.
Retrofit matters because the installed base is old and getting older: U.S. Bureau of Economic Analysis fixed-asset data shows the average age of manufacturing capital stock has climbed for decades. Waiting for the fleet to turn over before monitoring it means waiting a generation.
What is in the retrofit toolkit?
Current clamps (CTs). A split-core current transformer around the motor feed correlates draw with state: off, idle, or working, and roughly how hard. One clamp on the main drive answers the biggest question a supervisor has, and clamps on major motors like spindles, pumps, and drives sketch a surprisingly complete picture of a machine's day.
Vibration and temperature sensors. Magnet- or stud-mounted wireless vibration sensors on bearings, motors, and gearboxes catch developing mechanical faults weeks early, feeding condition-based maintenance. Mounting method matters for signal quality; see vibration sensor types for the tradeoffs.
Stack light taps. If the machine drives an andon stack light, its controls already publish status in three colors. A small sensor or relay tap converts that to a digital feed with zero controls engineering.
Photo eyes and proximity sensors. A beam across the outfeed counts parts; a proximity switch on a moving element measures cycles. Counts plus run state gets you availability and performance for OEE without touching a PLC.
Gateways. Where a PLC exists and IT allows, a gateway reads registers over Modbus or a vendor protocol and translates to modern transports. This is the richest retrofit source and the only one that needs network engineering; the details live in how to connect legacy machines.
How do you retrofit a line, step by step?
- Pick the line that hurts. Start at the constraint or the chronic downtime offender, where the data changes real decisions in week one.
- Choose one question per machine. Utilization? Counts? Bearing health? The question selects the sensor; skipping this step buys hardware that answers nothing.
- Install during planned stops. Clamps, magnets, and taps go on in minutes during changeovers or maintenance windows, under normal lockout procedures, with electricians handling anything in a panel.
- Commission against reality. Stand at the machine and confirm the data matches what your eyes see: run shows run, counts match a hand tally, idle is not reported as down.
- Land everything in one layer. One naming convention, one place, joined to schedule and product context, not five vendor apps; that principle is the core of the machine connectivity guide.
- Put a decision on it in 30 days. A daily downtime review, an alert to the right person, an automated shift report. Monitoring that changes no behavior gets unplugged at budget time.
What does retrofit monitoring cost?
Ranges, from current vendor materials rather than invented precision:
- Split-core current sensors: tens to a few hundred dollars per point, non-invasive install, per suppliers such as Milesight.
- Industrial wireless vibration sensors: roughly 3 to 7 year battery lives depending on sampling interval, per Emerson, KCF Technologies, and Phase IV Engineering; hardware from low hundreds to roughly one to two thousand dollars per point at the industrial high end.
- Photo eyes, proximity switches, and light taps: tens to low hundreds per point.
- Gateways: a few hundred to a few thousand dollars, typically serving several machines each.
Against that, Siemens' True Cost of Downtime 2024 research puts unplanned downtime at roughly 11 percent of revenue for large manufacturers. A retrofitted line usually costs less than one bad Saturday. The other budgeting truth: the hardware is rarely the largest line item. Plan as much for commissioning, naming, and the weekly habit of acting on the data as for the sensors themselves, because that is where the return is actually earned.
What mistakes derail retrofit projects?
Instrumenting everything, deciding nothing. Two hundred sensors and no owner for the Monday review is the most common failure. Sensors are the easy 20 percent. Believing the first week of data. Uncommissioned sensors misclassify idle as down and inflate counts; validate on the floor before showing anyone a chart, or the floor will spend a year distrusting the numbers. One dashboard per vendor. Each sensor brand ships its own app; five apps means nobody looks at any of them. Insist on one layer. Skipping the operators. The people who run the machine can tell you which signals matter and will spot bad data instantly; monitoring installed on people rather than with them earns quiet sabotage.
How does Harmony AI handle retrofit monitoring?
Retrofit-first is Harmony AI's default model, not a fallback. Our engineers deploy white-glove and in person: walk the line with your team, pick the cheapest reliable tap for each asset, commission every signal against what the floor actually sees, and land it all in one operational layer beside your ERP, quality, and paperwork data, where dashboards, alerts, and AI workflows run on top. No rip-and-replace, no controls rewrite, no dashboard sprawl. The CLS case study shows the approach across multiple shops of varied ages, and sensors in manufacturing covers the wider sensing landscape beyond monitoring.