HMI trend graph keeps plotting after the machine stops

Why does the trend graph keep plotting after the machine stops?

The graph was never told to stop: your E-stop reset the HMI stages, the reset flipped the stage condition inside the graph-enable rung, and the panel kept writing one sample per interval, at 60 seconds in the common configuration.

The machine has two stop paths in front of it, and only one leaked. A normal cycle stop turns off the start/stop tag and the sample tag, so the trend object goes quiet and the trace holds its last real value. An E-stop, on the machine that produced this fault, reset the HMI stages, and the graph-enable rung was built on a stage condition. The reset moved that condition into its enabled state, the rung switched sampling back on, and the panel began capturing a pen value at every interval while the process value sat still. The E-stop wrote no tag and loaded no screen, which is why nothing looked wrong from the operator's chair. A constant value plotted against time is a flat line, one point per interval; that is the whole mechanism.

What symptoms point at the enable rung and not the graph settings?

The fingerprint is path dependence: the trace survives hours away from the graph screen after a normal stop, and flattens within minutes when the E-stop was pressed from another screen first.

Each symptom points somewhere specific:

• Trace intact after a normal stop and hours off-screen: the graph object, its screen and the tags are sound; leave the display settings alone.
• Flat line only after an E-stop from another screen: a condition that exists only after the E-stop lives inside the enable logic; go to the stage reset.
• Line continues flat from the last process value, with no gap: sampling resumed and recorded a frozen value, so the panel never stopped plotting.
• Clean continuation of the old trace, no jump to zero: the rung toggled while the tags stayed as the stop left them, which is the signature of enable logic outside the tags.
• A trace that spikes while the displayed number holds steady is a different fault with a different feed, covered in why a trend graph spikes while the number stays stable.

What does the start/stop tag actually stop?

A start/stop-style trigger stops the plotting: while the stop condition holds, the object scrolls nothing and posts no data, and the samples missed during the stop are not back-filled when you press start again.

Trend documentation for this class of object says the same in its own words: on stop the display halts and no data is posted to the graph object, and data missed during the stop period is not written retroactively. That is why the cycle-stop path reads clean. The tag drops, the object posts zero points, and the trace holds still, because with the tag false there is nothing to plot at any interval, 60 seconds or 1 second. The interval itself is a fixed-rate capture setting, commonly 60 seconds and configurable down to 1 second, and interval ceilings are their own subject: a graph pushed to update faster than its panel can service stalls instead, as set out in why a trend graph will not update every half second.

How do I trace the enable rung while the machine is stopped?

Watch four things live with the machine stopped and the graph screen closed: the start/stop tag, the sample interval tag, the stage bit that feeds the enable rung, and the 24 V status input. On the faulted machine that input carried the address X4; yours will carry a different one.

Do it once per stop path. After the normal stop, expect the start/stop tag false and the rung false, with nothing in the trace moving. Then press the E-stop from the main screen, leave the graph screen closed, and read the four values again. In this fault the stage reset moved the stage bit, the rung followed it, and the start/stop and interval tags never changed, because the rung enabled sampling without writing them. That is the tell: a graph that keeps plotting while its trigger tags sit untouched is being enabled by logic outside the tags. The E-stop looked unrelated because it loads no screen and writes no tag on purpose; the connection ran through the stage reset, and live values are the only way to see it.

What does the supply-status input really watch?

Before you build with it, prove what the input sees: read its bit during normal running with 24 V present at the input terminal, then again during a controlled stop that drops the supply, and match the bit to the meter both times.

On the faulted machine this input existed to signal loss of that supply, and that real electrical job is what made the fix work. Put a meter on the input terminal, read the bit on the panel, and walk the machine through running and the controlled stop: the bit must follow the measured voltage in both directions. If the bit says powered while the meter reads 0 V, or the reverse, repair the field wiring or the input channel before touching any graph logic. The fix that follows needs a stopped condition the E-stop actually produces, and here that condition was the state of the monitored supply. The input made it visible to the HMI logic.

How do you fix the rung without touching the E-stop circuit?

Insert the supply-status input into the rung that uses the cold stage, so the graph enable now requires the stage condition and the power condition together, and leave the E-stop wiring untouched.

Work through it in this order:

1. List every rung and branch in the project that can write the start/stop tag or the sample tag, and mark the one built on the cold stage.
2. Read the status input live with the machine running and again stopped, and note the bit state and the meter reading at each; the polarity rule comes from those two readings.
3. Insert that contact into the cold-stage rung so the graph enable needs the stage condition and the power condition to agree.
4. Walk the branches from step 1 and correct any path that can still switch sampling on around the new condition.
5. Load the changed project with the machine stopped, then run the verification tests before returning the line to production.

One safety rule bounds all of it: never modify or bypass the E-stop circuit or the supply it acts on to make a graph test easier, and if the input's polarity or electrical meaning is still unclear after the meter work, stop the edit and have the controls design reviewed. The evidence names the input's job but not its contact polarity; take the polarity from the meter, never from the tag name.

Recommendation: Add the supply-status input to the graph-enable rung whenever that rung depends on a stage your E-stop resets. Set the contact polarity from live readings taken running and stopped, and keep the panel out of production until both stop paths produce a trace with zero added points.

What does the fix look like in a worked example?

Take a packaging line whose trend samples every 60 seconds: with the old rung a 20 minute E-stop adds 20 flat points, and with the input in the rung it adds none.

Scenario: one pen on a line trend, 60 second sample interval, the figure trend documentation uses as its example setting, with 1 second the usual fastest. The clock times are chosen to make the arithmetic plain.

Normal stop, 14:00. The start/stop tag goes false and the rung stays false. The line idles for 2 hours: 2 h x 3,600 s/h = 7,200 s, and 7,200 s / 60 s per point = 120 intervals that post nothing. The trace on return at 16:00 is unchanged.

E-stop, old rung, 16:00. An operator presses the E-stop from the main screen and walks away. Stages reset, the cold stage drops, the old rung switches sampling on, and the panel plots a frozen value. The operator returns at 16:20: 20 min x 60 s/min = 1,200 s, and 1,200 s / 60 s = 20 flat points drawn since the reset. That is the flat line.

Same E-stop, fixed rung. The supply contact sits in the rung with polarity matched to the meter, the E-stop drops the monitored 24 V, the rung goes false, and those same 20 minutes post 0 points. The trace ends at the last true sample.

Scaled to a 12 hour weekend shutdown: 12 h x 3,600 s/h = 43,200 s, and 43,200 s / 60 s = 720 flat points with the old logic, 0 with the fix. The interval moves the numbers fast: at 1 s per sample the same 20 minutes would draw 1,200 points.

How do I verify the fix through both stop paths?

Repeat the two tests that produced the fault with the same four bits on screen: a 15 minute E-stop at a 60 second interval must add 0 points, and 15 min is 900 s, so the old logic would have drawn 15.

1. Normal stop from the graph screen, 15 minutes off-screen: return and count new points. Expect 0; the tag path was already clean.
2. E-stop from the main screen with the graph screen closed: the rung must drop the moment the power bit changes, and after 15 minutes the count must be 0.
3. Reopen the graph and read the trace, not just its shape: it should end at the last true sample, continue cleanly on the next start, and carry no flat tail.
4. If points still appear, a second writer exists; the rung list from the fix tells you where to look.

Two loose ends are worth closing while you are at the panel. If the E-stop drops the panel's own supply as well as the monitored one, a restart can clear the trend's stored history, which is a separate problem from plotting; archiving HMI trend data after a power cycle covers it. And if the cell also needs a status or notice board beside the machine, that is a separate screen playing its own content offline; we build shop-floor signage displays for exactly that job.

Questions