It's half past three on a Sunday morning. The case packer has stopped with "pusher not home". The pusher is home. You can see it sat there. The operator has pressed reset four times.
Somewhere on that machine is a proximity sensor that cost less than a takeaway. Right now it's costing you a full line of output.
Prox faults are rarely hard. They're slow because people go in the wrong order. They swap the sensor first, find the new one does the same, and only then look at the bracket. Work from the outside in and most of them take ten minutes.
Look at the lights before you touch anything
Most proxes have an LED on the body. The PLC input card has one too. Between them they tell you half the story before you've picked up a tool.
- Sensor LED on, input LED on: the signal is getting to the PLC. Your fault is in the logic, the timing, or a different sensor.
- Sensor LED on, input LED off: the sensor is switching but the signal is lost on the way. Cable, connector, terminal, or the card.
- Sensor LED off with the target in front: the sensor can't see the target, or it has no supply.
- No LED at all, ever: check for 24V at the connector before you blame the sensor.
That last one catches people. A blown fuse or a tripped breaker on the 24V can take out a whole bank of sensors at once. If three proxes have died at the same moment, they haven't. Look at your 24V DC control supply first.
Don't test it with the machine live
Here's the trap. The quickest way to test a prox is to wave a spanner in front of it. On a running machine, that spanner is a signal. The PLC sees "pusher home" and moves the next thing in the sequence.
HSE's code of practice for PUWER lists sensors among the inputs to a control system (L22, paragraph 192). Trigger one by hand and you're driving the machine. HSE warns of machinery that "starts up unexpectedly" during maintenance (HSE, maintenance).
So isolate the motion first. Dump the air, lock off the drives, and leave only the control supply on so you can still see the LEDs. If you can't do that safely on this machine, it needs a written method before anyone tries.
Inductive: check the gap and the target
An inductive prox sees metal, and only at short range. The datasheet gives a rated sensing distance for a mild steel target. Stainless, aluminium and brass all bring that range down. So does a target that's smaller than the sensor face. Check the datasheet for the correction.
In our view, most inductive faults are mechanical:
- the bracket has been knocked and the gap has grown
- the target flag has bent or come loose
- the sensor has been tightened into the target and the face is smashed
- swarf or weld spatter has built up on the face
Set the gap from the datasheet and lock both nuts. A gap right at the edge of the range will work on the bench and drop out when the machine warms up and things shift. Give it margin.
Flush and non-flush sensors are not the same part. A non-flush sensor buried in a steel bracket will see the bracket and stay on all the time. A flush sensor stood proud will lose range.
Capacitive: it sees everything
A capacitive prox sees almost anything, including product, water and your hand. That's why it gets used for level in hoppers and to see product through a plastic wall. It's also why it gives false signals.
When a capacitive sensor goes wrong, look for build-up first. Product dust, sugar, damp cardboard fibre or washdown water on the face will hold it on. Clean it, dry it, then check again.
Then check the sensitivity pot. Somebody may have turned it to "fix" a fault on another shift. If it's set too high, it sees the wall. Too low, it misses the product. Set it with the real product, at the real fill level, and mark the setting.
The cable is usually the culprit
On anything that moves, the cable is the first place we'd look. The bit that flexes on a moving carriage. The M12 connector that gets washed down every night. The bit trapped under a guard someone put back in a hurry.
Wiggle the cable while you watch the input LED. If it flickers, you've found it. Intermittent faults like this are the worst kind, so if it won't show itself, read catching the fault that won't show before you change parts at random.
Like for like means more than the thread size
When you do swap a sensor, match it properly. M18 is not a part number. Check:
- PNP or NPN output
- normally open or normally closed
- flush or non-flush
- sensing distance
- connector or flying lead, and the pin-out
Fit a PNP sensor where the card wants NPN and the LED on the sensor will work fine. The PLC still won't see it. That one has cost plenty of sites a second hour of downtime at 4am.
If the right sensor isn't on the shelf, the night shift can't fix that. Your spares list can. Keep the ones on your critical machines in stock, labelled with what they replace. There's more on that in critical spares: what to hold.
Never bridge out a safety sensor
Some proxes aren't there to run the machine. They're there to stop it. A coded sensor on a guard door is a safety device. So is a prox that proves a gate is shut.
Guards and protection devices must be designed and installed so they can't easily be bypassed or disabled (L22). Bridging one out with a link "until the part comes" is the start of a very bad shift. If a safety sensor has failed, the machine stays down until it's fixed, or until a competent person has set up a safe way to run it. Our guide to safety relays and light curtains covers how to fault-find those circuits without defeating them.
The cost of the cheap part
The sensor costs pounds. The hours it takes when nobody on shift knows the order to check things in cost far more. If your line stops on sensor faults more than you'd like, it's worth reading why the cheapest part stops the most lines. Then ask who on nights can read an input card and set a gap without help.
What to do on Monday
- Walk your worst machine and label every prox with its PLC input address.
- Mark the set gap or sensitivity on each bracket, so the next person can put it back.
- Stock the sensors for your critical machines, with type, output and pin-out on the bag.
- Tell every shift: no bridging links on safety sensors, ever.
Paragraph numbers refer to L22, fourth edition (2014, amended 2018).