Every treatment plant that has been running for more than a few years carries three descriptions of itself. There is the controlled P&ID set in the document system, last issued by a design consultant at a revision that may or may not match what the contractor built. There is the PLC and SCADA tag database, which is what is actually wired and being scanned, complete with every point an operator or an integrator added on a callout. And there is the plant itself, with the transmitters and analysers that are bolted to it today.
All three describe the same works. None of them are fully right. On a brownfield upgrade, the I/O list is built from the first two and a hypothesis is what it is until somebody has walked the third.
Reconciliation is the pass that resolves those disagreements into a single register: one row per tag, a source citation per row, a verification status, and a defensible classification of keep, remove or replace. The register feeds the management-of-change package, the integrator's scope, and eventually the corrected drawings. It is not a full as-built survey of the plant. It is the record of what is known, what was checked, and what was decided.
Why water and wastewater plants drift further than most
The gap between drawings and field opens in every industry, and the general case is in brownfield I/O reconciliation, keep, remove, replace. Treatment plants have a set of habits that widen it.
Instruments are replaced on availability, not on specification. A wet-well level transmitter fails on a night shift. The operator fits the ultrasonic unit from stores, which has a 0 to 8 m range instead of the 0 to 6 m the drawing shows, and rescales the PLC input to keep the pumps starting at the right level. The drawing is not opened that night and is not opened the next morning either.
Analyser sensors change technology under the same tag. The dissolved-oxygen probe in the aeration basin, AIT-4102 on the drawing, was a membrane sensor with a 0 to 20 mg/L output when the plant was commissioned. It is an optical sensor now, on a transmitter with a different output configuration and a Modbus connection the drawing has never shown. The tag is the same. Nothing behind it is.
Small works never reach the master set. A chlorine residual analyser added to meet a tightened discharge condition, a second turbidimeter on the filter outlet, a float switch added after a flooded dry well: each was a minor works order, each was commissioned into SCADA, and each was drawn, if at all, on a single-sheet sketch that lives in the maintenance office.
SCADA migrations rename things. When the plant moved from its original PLCs to the current platform (the RTU versus PLC split matters here, because outstations and the main works often migrated in different years), the integrator renamed the tags to the new site convention. The P&IDs still carry the original numbering. An engineer reading the drawing and an engineer reading the SCADA tag list are looking at the same magnetic flow meter under two different names.
Remote sites are worst. Pump stations and outstations on radio telemetry are visited less, documented less, and modified more freely than the main works. A station's RTU point list is often the only accurate record of its instrumentation, and it was last exported by a contractor who has since left.
The practical result is that the SCADA database contains points the drawings do not show, and the drawings show instruments that are not on the plant. Both conditions have to be resolved before the upgrade scope is set, because both turn into change orders if they are found by the contractor instead of by the engineer.
The three sources and how to rank them
The latest controlled P&ID revision is the nominal paper authority. It has been through a review and approval cycle, it carries a revision number and a date, and it is what a design review references. Its limitation is that controlled revisions lag the plant, sometimes by years.
The PLC and SCADA tag database is what is currently wired and scanned. It includes the points added under minor works and the rescaled inputs from emergency replacements. It does not include anything that never reached the control system: local pressure gauges, sight glasses, a local turbidimeter that only ever fed a chart recorder. Its limitation is that it reflects wiring, not condition. A channel can be scanning a dead transmitter or an abandoned sample line.
A physical walkdown is the only source that reflects the plant directly, and the slowest. A works with a few hundred instruments across headworks, primary treatment, aeration, secondary clarification, tertiary filtration, disinfection and sludge handling cannot be walked in full without a real budget line. The practical approach is to walk selectively: the instruments where the first two sources disagree, and the instruments where being wrong has a consequence.
The precedence rule. On paper, the latest controlled revision governs. If the SCADA database contradicts it, that is a flag for investigation, not a reason to override the drawing at the desk. The walkdown resolves the conflict. Write the rule down, with its exceptions: a SCADA point with live, plausible process data for a tag that appears on no controlled drawing still has to enter the register, because it is on the plant.
Reconciling revisions
The most common conflict is the same tag on two drawing revisions with two different values: a range, a signal type, occasionally a location.
Take the wet-well level transmitter, LIT-1201, on drawing PID-012. Revision B shows 0 to 6 m. Revision D, the latest controlled revision, shows 0 to 6 m as well. The SCADA database shows an engineering range of 0 to 8 m. The drawing governs on paper, but the pump start and stop set points in the PLC were rescaled against 8 m years ago, and the pumps have been starting at the right level ever since.
The desk cannot settle this. What the desk can do is record it precisely: PID-012 revision D, 0 to 6 m; SCADA channel LIT-1201, 0 to 8 m; conflict noted; walkdown required. The walkdown reads the nameplate and the transmitter's configured span. If the installed unit is 0 to 8 m, revision D is wrong and a drawing correction goes into the upgrade documentation package. If it is 0 to 6 m, the PLC has a scaling error that has been quietly moving the pump set points, and that is a control-system change order and an operational finding, not a paperwork one.
The same logic applies to signal-type conflicts. A high-level float in the sludge holding tank, LSH-6301, drawn as a discrete input, appears in SCADA as an analogue point because someone replaced the float with a small pressure transmitter and derived the alarm in the PLC. Flag it. Walk it. Then decide whether the register keeps the discrete tag with a note, or records a replacement.
Keep, remove, replace
Once the sources are cross-referenced and the conflicts flagged, every tag gets one of three classifications.
Keep. The tag is on the latest controlled P&ID, the instrument is on the plant, and the specification is not changing. The channel carries through. Verification is by paper where the sources agree and by walkdown where they did not.
Remove. The tag is not on the latest controlled P&ID, or is not on the plant, or sits in a section being demolished under the upgrade, for instance the chlorine contact tank instruments when the plant moves to UV disinfection. Walk every remove candidate before the classification is frozen. Walking a tag that turns out to be live costs an hour. Having the contractor cut a live effluent flow signal costs a permit exceedance.
Replace. The device is on the plant and is being changed under the upgrade. Replace rows record the old tag, the new tag if it changes, the old and new signal types, the reason, and whether the change needs a new PLC channel or reconfigures an existing one. A typical replace on a treatment works is the effluent flow meter FIT-7001 moving from a mechanical meter with a pulse output to a magnetic flow meter with a 4 to 20 mA output and a digital diagnostic link, because the discharge consent now requires continuous, logged flow.
The register is a living document while engineering is under way. Classifications move as walkdown results arrive and the scope firms up. It is frozen when the change package goes to review.
What to walk, and what to record
Walk the following as a minimum.
- Every tag flagged as a conflict between the latest controlled drawing and the SCADA database.
- Every remove candidate, before the classification is frozen.
- Every compliance-critical tag regardless of whether the sources agree: effluent flow, disinfection residual, final turbidity, anything that feeds a permit report or a regulator's telemetry link.
- Every safety-related tag: chlorine gas leak detection, digester gas pressure and flame detection, confined-space ventilation interlocks.
- Every remote station that has not been formally surveyed in a period the plant manager sets. Five years without a survey is a reasonable trigger for a full-station check instead of spot verification.
For each tag, record whether the instrument is present at the drawn location, the model and serial number from the nameplate, the nameplate range or the configured span, its apparent condition, and whether it sits in a local panel or telemetry cabinet that the drawing does not show. Date and initial each finding in the register. That is the verification record, and it is what the auditor reads.
Walk with the register open on a tablet. Transcribing field notes back into the register later is where tags are lost, and the walkdown results are usually on the critical path for the change package.
A worked example from a coastal treatment works
Three tags, three decisions.
AIT-3305, on the drawing, not on the plant. The latest revision of PID-033 shows a residual chlorine analyser on the contact tank outlet. SCADA has no point for it. The walkdown finds the sample line capped and the analyser cabinet empty; it was removed when the plant moved to UV and the drawing was never revised. Classification: remove. The register cites PID-033 revision C as the source, the SCADA absence as corroboration, and the walkdown date as verification. The change package covers the formal decommissioning of the sample point.
AIT-2210, a different range in SCADA than on the drawing. PID-022 shows a filter-outlet turbidimeter with a 0 to 100 NTU range. SCADA shows 0 to 10 NTU. The walkdown finds a newer instrument configured for 0 to 10 NTU, fitted when the discharge condition tightened, with the narrower range chosen so the compliance trend was readable. SCADA is right. Classification: keep, with the register carrying the confirmed 0 to 10 NTU range and a drawing correction logged for the documentation package.
FIT-7001, being replaced to meet the discharge consent. The existing mechanical effluent meter gives a pulse output totalised in the PLC. The new consent requires continuous flow logging with a diagnostic record, so the design specifies a magnetic flow meter with a 4 to 20 mA output and a digital link to the historian. Classification: replace. The row records the old signal type, the new one, the reason, and the two downstream actions it generates: a remove entry for the pulse channel and an add entry for the analogue channel and the network point, both inside the change package.
Three tags, three distinct decisions the register has to support: a channel coming out, a channel staying with a corrected value, and a channel being removed and replaced under new scope.
Where it goes wrong
Trusting the controlled drawing over the plant. The reconciliation exists because the drawing is not reliable. Using it as if it were defeats the purpose.
No source and revision on each row. A register with a classification column and no source column is an opinion. When the integrator, the commissioning lead or the regulator questions a row, the only defence is the source it came from and the walkdown that resolved it.
Walking everything. A full instrument inventory of a large works delays the change package by weeks. Walk the conflicts, the removes, the compliance-critical and safety-related tags, and the neglected outstations. Accept paper where the sources agree and the consequence of error is recoverable.
Walking nothing. The opposite failure produces a register that looks complete and is built on the assumption that the drawings are right. They are not.
Carrying the reconciliation in someone's head. On a small plant it is tempting to skip the register. It works until the review, where the reviewer asks for the as-found baseline and there is none.
Referencing a register that was never frozen. If the change package cites the register by name without a revision or date, and the register keeps changing, the baseline is undefined. Freeze it, version it, and issue it as a controlled document before the package goes to review.
Getting the instrument list out of the drawings
The pass starts with the instrument list from the latest controlled drawings in a form that can be cross-referenced against the SCADA export. On a works with dozens of P&ID sheets, many of them scanned copies of drawings issued long before the plant had a document system, extracting that list by hand is slow and the tags missed on extraction become rows missing from the register. Tagsight reads the drawing set, scanned sheets included, into a tagged list with signal type and drawing reference per instrument, which gives the register its paper-source column in an afternoon; the reconciliation itself still needs the SCADA export and the walk. How that applies to a treatment works is on the water treatment page, and the scanned-drawing case is in scanned P&IDs versus CAD exports.
The register is not the same thing as an accurate set of drawings. It is a defensible record of what was known, what was verified and what was decided from that. That distinction matters when the change package is reviewed, when the integrator inherits the scope, and when the next upgrade engineer opens the drawing set in five years and finds that, for once, it matches the plant.