What Is Trending?

Back to Blog

Real-Time and Historical Trends

A real-time trend scrolls live, showing the last minutes or hours of a value as new samples arrive, which is ideal for watching a startup, a fill, or the immediate reaction to a control change. A historical trend pulls stored data from a historian to plot days, months, or years, which is what you need to see a compressor slowly losing efficiency or a tank's fill-and-draw cycle over a season. Most SCADA systems let you flip between the two on the same chart.

Good trend displays let you overlay several tags on shared or independent axes, so you can see, for example, wellhead pressure, casing pressure, and flow together and read the relationships between them. Cursors, zoom, and value readouts turn a picture into numbers when you need to quantify a change.

Why Trending Matters in Oil & Gas

Trending is where slow problems are caught before they become failures. A gradual rise in bearing temperature, a creeping increase in separator level, or a slow decline in gas-lift injection all show up as a trend line long before they trip an alarm. Overlaying trends is also how engineers separate cause from coincidence - if tubing pressure falls every time a nearby well is put on, the relationship is obvious on a shared time axis.

The quality of trending depends on the data behind it: the sampling rate, whether the historian applies compression, and how gaps from telemetry outages are handled. A cloud SCADA platform that historizes field data can serve both the live scrolling trend an operator watches and the multi-year historical trend an engineer uses for analysis, from the same stored dataset and reachable in a browser.

Building Trends People Can Actually Read

Most bad trends fail on the same few choices. Too many pens: past a handful of traces, a chart stops being readable, so split related groups across stacked charts instead of piling them onto one. Careless axis scaling: full autoscale re-zooms on every spike and makes a flat line look dramatic, while a fixed 0-to-max scale can flatten the variation that matters; pick scales that show the normal operating band with room to see excursions. Mixed units on a shared axis - pressure and temperature fighting over one scale - hide both; give each unit family its own axis or its own panel.

The other half is consistency. Build trend templates for repeated equipment - every wellhead gets the same pens, colors, and scales - so an operator moving between sites reads each chart instantly instead of re-orienting. Keep color meanings stable across the system, and resist decorating: gridlines, markers, and fills that do not carry information are noise. A good test for any trend display is whether a competent operator who has never seen this particular chart can say what is normal, what is abnormal, and roughly when it changed, within a few seconds of looking.

What the Historian Does to Your Data

A trend is only as honest as the stored data behind it, and a SCADA historian rarely stores every raw sample. Collection deadbands discard changes smaller than a threshold, and deadband compression further thins what is archived. Tuned well, this is invisible; tuned carelessly, it clips the tops off short excursions and turns a lively signal into a staircase. If a trend looks suspiciously smooth or misses an event a field tech swears happened, the compression settings are the first suspect.

Rendering choices matter too. A trend client drawing straight lines between sparse stored points shows a clean ramp where the process may have stepped; stair-step rendering is more truthful for slowly sampled or compressed data. Zooming changes the picture again, because many clients serve aggregated values at wide time ranges and only fetch raw samples when zoomed in - so an event can appear, disappear, or change shape as you zoom. Knowing whether you are looking at raw samples, interpolation, or aggregates is part of reading a trend correctly, and it is the difference between diagnosing the process and diagnosing the plotting library.

A Worked Diagnostic Walkthrough

Symbolic example: a transfer pump's discharge pressure has drifted down over several weeks from its normal value P toward some lower value, while motor current holds steady and flow at the downstream meter has fallen slightly. Overlay the three on one time axis. If discharge pressure and flow decline together while current is unchanged, the pump is doing the same work for less result - consistent with internal wear or recirculation. If instead current fell along with pressure and flow, the pump is doing less work, pointing upstream: suction restriction, a closing valve, falling tank level, or gas entrainment.

Now add context tags: suction pressure, tank level, and a parallel pump's status. If the drift began exactly when the parallel pump changed state, the correlation reframes the problem entirely - system curve, not pump health. This is the everyday power of trending: each overlay either strengthens or kills a hypothesis, without a site visit. The habit to build is to ask what else changed at that moment, and to answer it by dropping the candidate tag onto the same time axis rather than by memory, because memory reliably invents cleaner timelines than the historian records.

Trend Habits for Shift Operators

Trends earn the most value when they are consulted routinely, not only during upsets:

1. Start of shift, open the overview trends for your key sites and scan the last day for anything drifting from its normal band.
2. After any control change or restart, watch the relevant real-time trend long enough to confirm the response, then check it again later in the shift.
3. Before acknowledging a recurring alarm, pull the trend of the alarming tag over a multi-day window - repetition patterns identify chattering or cyclic causes instantly.
4. Annotate events where the system supports it, so the next shift inherits the context.
5. When handing over, show the trend, not just the number: where it has been is the story.

One caution belongs with the habit: the live HMI value and the trend can legitimately differ, because one shows the latest poll while the other shows what survived collection and compression into the historian. Small discrepancies are normal; systematic ones are a configuration problem worth chasing through the guide to HMI and historian readings that differ. Operators who understand that pipeline - sensor to poll to store to chart - read trends with appropriate trust, which is high but not blind.

Frequently Asked Questions