Why not just buy new machines
Many factories run equipment that is decades old and still works well. Replacing it with a connected model can cost far more than the machine’s remaining value justifies, and a new machine may need new tooling, retraining and downtime for installation. Meanwhile, the old machine produces no data: nobody knows exactly when it runs, how hard it works or how it is wearing, beyond what an operator notices.
Retrofitting is the practice of adding sensors and a communication path to existing equipment so it can join a connected system. The machine keeps doing its job, and the sensors report on it from the outside. It is one of the most common early steps toward smart manufacturing, partly because it can start small.
Start with the question
It is tempting to buy a box of sensors and see what turns up. A better start is a plain question: Is this machine running when we think it is? Why does it stop? Is the motor bearing getting worse? How many parts does it make per shift? The question decides what to measure. The same thinking applies to any IoT project, and problem definition up front prevents collecting data nobody uses.
Common clip-on and add-on options
Many retrofit sensors attach without opening the machine or cutting into its controls, which keeps risk and warranty questions lower. Typical examples include:
- Current clamps. A clip around a power cable measures how much current a motor draws, without touching the wire’s conductor. Current reveals whether a machine is running, idling or loaded, and sharp changes can suggest a problem.
- Vibration and temperature sensors. Small units stick or bolt to a motor, pump or gearbox housing and report on its health, as described under condition monitoring.
- Light and sound sensors. A sensor that watches a machine’s indicator light stack can record red, yellow or green states without any wiring into the control system.
- Proximity or optical sensors. These count parts passing a point, or detect cycles of a moving component.
- Pressure, flow and level sensors. Added to pipes and tanks, often with fittings that clamp on or screw in.
- Cameras. Pointed at a dial or gauge, with software that reads it, though this works only where conditions are consistent.
Which is right depends on the question. Counting good parts from motor current alone is rarely possible, while a part counter does it directly.
Getting the data out
Add-on sensors are not much use unless their readings reach somewhere. Choices include:
- Wired connections to a nearby gateway, which suit noisy environments and permanent installations.
- Wireless links. The right wireless protocols depend on range and battery needs; a plant full of metal can block or reflect signals, so testing on site is wise. Low-power networks suit sensors that report occasionally, while mesh networking can help where distance or obstacles are an issue.
- A local gateway. IoT gateways collect readings from several sensors, translate protocols and forward summaries, which also reduces the burden on the plant network.
If the machine has a controller, it may already hold useful signals. Tapping into that is sometimes more reliable than adding sensors, but it needs care and agreement from whoever maintains the controller, since changes can affect machine behavior.
Power and installation
Retrofit sensors need power. Options include battery power, which makes sense for sensors that report rarely and where a good battery life estimate is available; nearby mains power; or low-voltage supplies. Some sensors harvest energy from vibration or heat, though they suit limited uses. Where wired data is used, power over Ethernet can carry both power and data on one cable.
Installation details matter more than most people expect. A vibration sensor mounted on a loose cover reports the cover, not the machine. Sensor placement affects readings, and the maker’s mounting instructions are worth following closely. Cables must stay clear of moving parts, heat and fluids, and enclosures should suit dust, moisture and washdown.
Safety and responsibility
Working on live industrial equipment can be hazardous. Installation should be done by qualified people who follow the site’s lockout and electrical safety procedures. Adding sensors should not interfere with safety functions such as emergency stops, and it should be agreed with the people responsible for the machine. Read-only sensing, which watches without controlling, carries much less risk than anything that sends commands back.
Network safety counts as well. Every new connected device is a potential entry point, so changing default passwords, placing devices on a separate network segment and planning for firmware updates are sensible habits.
Expect imperfections
- Indirect measures. Inferring machine state from current or vibration involves judgment and calibration. Thresholds that work on one machine may not carry over to its twin.
- Sensor drift and failure. Add-on sensors need checking, which links to the wider theme of sensor calibration.
- Data quality. Noise, gaps and misleading readings are common, so early data should be cross-checked against what operators see.
- Old machines are variable. Wear, repairs and modifications make each one slightly different.
A sensible approach
Pick one machine and one question. Choose the simplest sensor that answers it, install it carefully, and compare its output with reality for a few weeks. Run it as a small trial before copying it to other equipment. If it answers the question, expand; if not, you will have learned cheaply. That approach lets older equipment contribute useful data without the cost and disruption of replacing it.