2.2 Gauges, Meters, Analyzers, and Trends
Key Takeaways
- Interpret a reading with its location, unit, time, operating state, and expected process relationship; a lone number rarely proves the cause of an upset.
- Trend shape matters: a step, drift, spike, noisy trace, or implausibly flat line points to different process or instrument checks.
- Compare upstream and downstream measurements using realistic process lag, and corroborate suspect online data before changing treatment.
- Chart scale, averaging, logging interval, sample transport, and operating-state labels can hide peaks or create misleading comparisons.
Quick answer: Read process data in context. First identify what was measured, where, when, in what unit, and under which operating state. Then examine direction, rate, duration, and related variables. Verify an abnormal online value before making a treatment change unless the approved emergency procedure requires immediate protective action.
Instruments answer different questions
WPI expects Class I operators to identify abnormal operation by interpreting gauges, meters, charts, and graphs.
| Evidence | Typical information | Interpretation trap |
|---|---|---|
| Gauge | Local pressure, vacuum, or differential pressure | Reading it without checking isolation, location, or range |
| Meter | Flow, totalized volume, level, or electrical load | Confusing instantaneous rate with accumulated total |
| Online analyzer | Turbidity, pH, disinfectant residual, or another quality value | Assuming the sample and signal represent the process instantly |
| Trend or chart | Value plotted through time | Ignoring scale, averaging, missing data, or operating state |
A flow rate answers “how fast now”; a totalizer answers “how much over an interval.” An online analyzer describes the water reaching its sample cell, possibly after sample-line delay.
Use five features of a trend
- Baseline: What range is normal for this plant, source condition, season, and operating mode? Plant SOP limits and jurisdictional requirements control; WPI does not provide one universal setpoint.
- Direction: Is the value rising, falling, or steady?
- Rate: Is the change gradual, sudden, or oscillating? A rapid step differs from slow analyzer drift.
- Duration: Was it a single spike, a repeating cycle, or a sustained departure?
- Relationship: Did connected variables move in an order consistent with flow and process detention time?
Process lag is crucial. A raw-water change should not appear simultaneously at every downstream location. Compare the event time with basin travel, filter status, sample transport, and the historian timestamps.
Recognize common shapes
| Trend shape | Possible process cause | Possible data cause | Useful corroboration |
|---|---|---|---|
| Sudden step | Valve, pump, feed, or source change | Rescaling, maintenance, or tag substitution | Event log, mode, field reading |
| Slow drift | Gradual source or process change | Fouling, reagent aging, or calibration drift | Grab or bench result and maintenance record |
| Repeated oscillation | Unstable control loop or cycling equipment | Signal noise or unsuitable display settings | Related valve position, pump status, and raw signal |
| Isolated spike | Short process disturbance or bubble/particle | Electrical noise or sample disturbance | Adjacent tags and instrument status |
| Exact flat line | Truly stable condition | Frozen value, held output, or communication loss | Timestamp, quality flag, and independent measurement |
Protect the quality of the evidence
An online result is only as representative as its sample and data handling. EPA guidance for treatment-plant turbidimeters emphasizes representative tap location, suitable sample flow, short transport delay, instrument maintenance, aligned timestamps, and accessible historical data. It also warns against data handling that caps true spikes. A chart can therefore look smooth for the wrong reason.
Before comparing trends, ask:
- Are both displays using the same time zone, time window, and sampling or logging interval?
- Is the vertical scale tight enough to show meaningful movement but wide enough not to exaggerate noise?
- Does each logged point show an instantaneous value, an average, or a maximum within the interval?
- Were backwash, filter-to-waste, calibration, shutdown, or maintenance periods identified?
- Does the analyzer controller agree with SCADA, and does an approved independent test support the result?
Sparse logging can miss an event between records. Conversely, an overly narrow graph scale can make harmless noise look dramatic.
Connect variables into a process story
Use upstream input → treatment action → downstream result. For a conventional plant, rising raw-water turbidity is an input. Coagulant dose, rapid-mix operation, floc condition, and settled-water turbidity describe treatment. Individual filter effluent turbidity, head loss, and filter status describe the result. If raw and settled water remain stable while only one online filter analyzer jumps, first verify that analyzer, its sample flow, and the filter's operating state. If several downstream indicators deteriorate in sequence, a process problem becomes more plausible.
A pump start accompanied by current, discharge pressure, and rising flow is coherent evidence. A “running” status with no current, pressure, or flow may be false feedback or failed equipment. Increasing differential pressure through a filter together with declining flow can reflect accumulating resistance; a single pressure gauge with no related change may instead need verification.
A safe decision framework
Use confirm → compare → connect → correct → capture:
- Confirm tag, unit, time, mode, and safe field condition.
- Compare the reading with a second source, expected range, and prior trend.
- Connect related upstream and downstream variables using realistic lag.
- Correct only under the authorized SOP: address the verified cause, maintain barriers, and escalate when water quality or control integrity is uncertain.
- Capture readings, timing, actions, notifications, and response.
Worked scenario
After Filter 3 returns from backwash, its effluent-turbidity trend rises briefly. Other filters and combined effluent remain steady. Do not average the event away or assume immediate filter failure. Confirm that the trend is Filter 3, mark its return-to-service state, inspect the time scale and peak, compare the analyzer controller with SCADA, and follow the plant's filter-to-waste or return criteria. A repeatable, corroborated recovery pattern supports a process decision; an implausibly flat controller value or disagreement between controller and SCADA supports an instrumentation check.
Official source trail
Which trend most strongly suggests a stale or frozen signal rather than a genuinely constant process?
Why should an operator examine graph scale and logging interval before interpreting a turbidity peak?
Only one filter-effluent analyzer rises sharply; raw water, settled water, other filters, and combined effluent remain steady. What is the best initial interpretation?