Your Trend Interval Is Hiding the Problem

Most trends in most buildings were set up once, at whatever interval the template defaulted to, and never reconsidered. They exist to satisfy a specification, not to answer a question. Which is why they so often show you a smooth, reassuring line right through the middle of a real problem.

There is a documented case that makes this concrete: a chilled water valve trended at fifteen-minute intervals looked unremarkable. The same valve trended at one-minute intervals showed the position changing every single sample — classic hunting, wearing out an actuator and destabilising the loop. Nothing about the valve changed between those two trends. Only the sampling rate did.

Match the interval to the question

Stop thinking about trends as record-keeping and start thinking about them as an instrument you tune to the thing you're measuring.

The pattern

Slow questions want long durations at coarse intervals. Fast questions want short bursts at fine intervals. The mistake is using one setting — usually fifteen minutes, forever — for both.

The two-pass method

When you don't yet know what you're looking for, don't guess at an interval. Work in two passes.

This is the same logic as the diagnostic method generally: localise first, then look closely. Fine-interval trending across a whole building for a month produces a data set nobody will ever read.

What to trend together

A single point in isolation rarely tells you anything. Trends earn their value from what sits next to them.

• Command and feedback together. Valve command alongside valve position, or fan command alongside status. The gap between them is where faults live.
• Cause alongside effect. Discharge air temperature next to the valve that controls it. If the valve moves and the temperature doesn't, that's a different problem than the valve not moving.
• The thing that gates it. Trending economizer position without outdoor air temperature tells you nothing about whether the economizer should have been open.
• Setpoint alongside process variable. A loop that looks unstable may be tracking a setpoint that is itself moving.

Practical constraints worth knowing

Where the platform supports change-of-value trending rather than fixed intervals, it is often the better instrument for exactly this reason: you capture every actual change without sampling on a clock. It is particularly good for catching intermittent behaviour that a fixed interval would only occasionally land on.

Before you leave

If you set a trend to one-minute intervals to chase something, write down that you did, and set it back when you're finished. The next technician inheriting a controller with its history full of month-old one-minute data will not thank you, and neither will the trunk.

A trend is an instrument. Pick the resolution for the measurement you're taking — and put the instrument away when you're done.

Found a wrong number or a broken link? Tell me and I’ll fix it.