Consider a hypothetical milling machine. Its demand rises during some operations and falls while it waits. That pattern could help an owner ask better questions about the working day. It cannot, by itself, explain what was made, whether it passed inspection or why work stopped.
Start with a decision
Choose the question before choosing the dashboard. You might want to investigate a recurring quiet period, understand whether a machine is active when expected, or compare activity before and after a scheduling change.
Each question needs different context. To assess activity during a shift, you need the shift schedule. To investigate a pause, you may need job records or an operator’s explanation. To assess output, you need production records.
Write down what you would do differently if the answer changed. If the reading would not change a decision or prompt a useful conversation, collecting more of it may add little.
Understand what the meter sees
A power reading measures electrical demand at the point being monitored. If that point includes the whole machine, the reading may include auxiliary equipment as well as the process you care about. Coolant equipment, for example, can contribute to the demand.
A NIST co-authored study of machining power models describes contributions from the machine body, coolant system and moving axes. It also explains why changes in machine and machining configuration matter to the model. Its findings are a reason to check an interpretation locally, rather than assume one threshold will work everywhere.
For the hypothetical mill, a higher reading might be consistent with cutting under the conditions observed. Calling it “likely active” preserves that distinction. Calling it “good production” would require evidence about the output.
Check the signal against the work
Compare the recorded trace with someone’s observations of the machine. Include ordinary working periods and the situations likely to confuse the interpretation, such as preparation, waiting and different jobs. Align the timestamps so a visible change can be compared with the activity that caused it.
Then test the proposed rule on other periods. Note where it gets the interpretation wrong. If two activities produce similar readings, the display may need a broader category or another source of information.
Agree how short changes should be treated. A brief dip should not automatically become a separate stoppage, and an average over a long interval may conceal a short interruption. The appropriate level of detail depends on the decision.
Keep the original readings available alongside the interpretation. That makes it possible to revisit the rule when the machine, job mix or operating practice changes.
Make uncertainty visible
A gap in readings leaves a period you cannot classify from that data. Showing it as zero activity would turn missing evidence into a conclusion.
Show when the latest reading was taken, the period covered and any gaps. If a summary excludes missing periods, explain that beside the result. Otherwise, a percentage can look comparable across days even when the amount of observed time differs.
The same care applies to the labels. If the system estimates activity from power, say so. Give the person reviewing it a way to see the evidence and add context when an apparent idle period has an ordinary explanation.
Keep activity and output distinct
Activity monitoring can help identify periods worth investigating. Measuring productive performance needs more information. Overall equipment effectiveness, or OEE, includes availability, performance and quality; NIST’s manufacturing monitoring research describes these as separate components.
A machine can draw power while making an unacceptable part. It can also be quiet during a planned changeover. Power alone does not establish either outcome. Use production and quality records when those are the questions you need to answer.
A checklist before expanding
Before extending a monitoring setup to more machines, ask:
- Which decision will this help someone make, and who will review it?
- What equipment does the measurement include?
- Which observed activities support the labels, and where does the interpretation remain uncertain?
- Are timestamps, gaps and the basis of any percentage clear?
- What other records are needed to explain activity or assess output?
- What change would trigger another check of the interpretation?
A small pilot should leave you with a clearer answer to those questions. That is a useful basis for deciding whether to expand monitoring.
If you are considering machine monitoring, tell Athonics which decision you want it to support. We can discuss what needs measuring and how to check that the result is useful.