An older machine can still be worth monitoring. The first decision is which useful signal can be obtained without changing how the machine runs. Start with its documentation, installed options and actual operating behavior. A connector on the cabinet does not tell you which data are available or whether another device may safely read them.
Look for supported access first
Ask for the exact controller and firmware version, wiring information and any documented status outputs or communications interface. Where an approved protocol provides the needed values, a gateway can translate the readings into records another system understands. A documented output or a suitable external sensor may be enough when the question is simply whether the machine ran.
Keep the physical connection and the protocol separate in the assessment. For example, RS-485 identifies an electrical interface; it does not establish that a machine speaks Modbus or that a register map is available. Where Modbus is supported, the Modbus application protocol specification distinguishes read functions from write functions. The equipment manufacturer’s documentation still determines the meaning of the values on that machine.
Make read-only a testable boundary
Agree which signals can be collected and how often. Have the responsible controls specialist approve the connection, access restrictions and installation. Restrict the monitoring path to the required reads; do not expose command or write functions to a dashboard. Even a read-only connection can interfere with an existing network if a device polls too aggressively or is installed incorrectly. Prove that the machine runs normally with the monitor disconnected, connected and reconnecting.
Illustrative example: a press with one useful output
In this invented example, a shop wants to know when a press is running outside its scheduled shift. The controller has no approved data port for the project, but the equipment documentation identifies a status output that the installer verifies as “cycle active.” The monitor records start and end times for that signal. It does not report good parts, tool condition, or operator identity because the signal does not establish those facts.
At 18:14 the signal becomes active during an approved setup job. At 18:27 the gateway loses its link. The history shows a confirmed active interval up to the last valid observation and then unknown status. The gap is not counted as stopped time. On reconnect, the system records the new observed state without inventing the intervening transitions.
Hand over a signal map, not just a dashboard
For every collected value, record its asset, origin, meaning, units, expected update behavior and how to recognize invalid or stale data. Verify each state against a witnessed event or the existing operator display. Document how to isolate the monitor, who can change its configuration and what staff should do when the data stops. If the only available signal cannot support the decision, stop and reconsider the scope rather than estimating the missing facts.
See equipment monitoring for a first project, and embedded systems for equipment monitoring for cases that need a device or interface designed around the machine.