Machines older than the software can still tell you everything that matters.
Which signals a conventional machine already offers, how the retrofit works, and an honest table of what you get compared with a networked CNC.
A large share of Indian manufacturing capacity sits on machines built before machine data was a product category. Radial drills, capstan lathes, power presses, hydraulic shearing machines, VMCs with controls whose Ethernet port was an option nobody ordered. These machines are frequently the most reliable equipment on the floor and the least visible in any report.
The instinct is to exclude them from a monitoring project and cover the CNCs first. That is usually the wrong order, because the older machines are where the unrecorded losses concentrate — nobody has ever had a timeline for them, so nobody knows.
The signal exists; it just is not on a network.
Motor contactor state
The contactor that closes when the main drive runs is the cleanest available proxy for machine-on. It is a dry contact, it is already in the panel, and it costs nothing to read.
Current on the supply
A split-core CT around the supply cable distinguishes running from idling from cutting, without any panel modification. Useful where contactor state alone cannot separate powered-on from working.
Stack light
Already wired, already meaningful, already understood by everyone on the floor. Three signals that map directly to running, attention and stopped.
Existing counters and limit switches
Presses, shears and SPMs frequently have a stroke counter or a proximity switch already fitted. Reading it converts a machine into a part-counting machine.
An afternoon per machine, not a project.
A MachineWise Gateway is mounted in or beside the panel and wired to whichever signals that machine offers. Inputs are optically isolated, which matters more in Indian plants than the specification sheet suggests: welding sets, VFDs and contactor coils create an electrically hostile environment that destroys consumer-grade hardware. The board carries a real-time clock so a power cut does not corrupt the timeline, and 4G, Wi-Fi or Ethernet uplink so machines in a shed with no cabling can still report.
No machine control is modified. Nothing is written to the machine. The wiring taps existing signals and, where a signal does not exist, adds a sensor — a proximity switch on a ram, a CT around a cable — without changing how the machine operates. If the Gateway is removed, the machine is exactly as it was.
Survey the panel
Half an hour per machine to identify which signals exist. Most machines offer more than the owner expects.
Wire and mount
Screw-terminal I/O, DIN-rail or panel mount, scheduled around a planned stop rather than interrupting production.
Verify against reality
Watch the machine and the dashboard together for one shift. States that do not match get corrected before anyone is asked to trust the data.
Being precise about this prevents disappointment later.
| Metric | Networked CNC | Legacy machine via Gateway |
|---|---|---|
| Running / idle / stopped states | Yes | Yes |
| Accurate stop durations and timeline | Yes | Yes |
| Part or stroke counts | Yes | Yes, where a counter or sensor exists |
| Downtime reasons from the operator | Yes | Yes, via kiosk, QR or keypad |
| Availability and a usable OEE | Yes | Yes |
| Active program or job number | Yes | No — unless selected by the operator |
| Spindle load, feed override, alarms | Yes | No |
| Cycle-level performance against an ideal cycle | Yes | Partial — depends on count granularity |
The right way to read this table: a legacy machine gives you the availability and downtime picture in full, which is where most of the recoverable loss sits. It does not give you controller-level detail, and no vendor can make it.
The machines nobody was measuring.
A forging unit with eight hammers and presses, none of them monitored, and twelve CNCs that were. The assumption was that the CNCs were the constraint. Wiring stroke counters and contactor states to Gateways on the older equipment showed the opposite: the hammers were idle for long stretches waiting on billet heating, and the pattern repeated at the same times every day.
Nothing about that finding required controller-level data. It required a timeline, which the machines had been capable of producing for thirty years and nobody had asked for. At a typical forging machine-hour rate the recovered time on a floor of that size runs into tens of lakhs annually — but the number that changed the decision was simpler: how many hours a day the most expensive machines in the building spent waiting.