AHU-0037 — Economizing when it should not (damper position)

Preconditions (host-enforced): Supply fan running, and the unit past its start-up transient — a morning warm-up purge or a scheduled pre-occupancy flush opens the damper by design and needs a host NO_EVAL, not a fault. min_damper_sp must be set to the unit’s own commissioned minimum damper position before deployment: 20% is PNNL’s default, and a unit whose design minimum is 30% alarms continuously against it. OAT sensor quality is a host gate, since a sensor reading high manufactures this fault out of a working economizer. Do not bind the rule to a unit with no return-air path (100%-outdoor-air or makeup-air): it has no minimum position to hold. The rule reads the damper command, so it is silent on a damper commanded to minimum and stuck open — that is AHU-0020 and AHU-0028. When any gate is unmet the verdict is NO_EVAL, not healthy.

Points: oat, rat, oa_dmpr_cmd

Outputs:

• yFault — True while outdoor conditions have not favored economizing and the OA damper command has stayed above min_damper_sp × excess_damper_multiplier, both continuously for at least alarm_delay

Parameters:

Description

Nothing about the weather favors free cooling and the outdoor-air damper is open well past its minimum position anyway. Whatever the coils are doing, they are doing it to more outdoor air than the ventilation calculation asked for, and outdoor air no cooler than the air it displaces is a load, not a resource. This is the raw-damper half of the economizer question: it never divides one temperature difference by another, so it keeps working in the conditions where the outdoor-air fraction cannot be computed at all — which is most of the band it fires in. AHU-0021 and AHU-0030 measure how much extra air arrives; this rule asks only whether the damper had any business being open, and answers from the command the controller wrote.

Detection Logic

econ_unfavorable = (rat − oat)          < temp_deadband    when econ_type_is_ddb
                 = (econ_hl_temp − oat) < temp_deadband    otherwise

yFault = econ_unfavorable
     AND oa_dmpr_cmd > min_damper_sp × excess_damper_multiplier
     sustained continuously for alarm_delay

Block graph (rule.cxf.jsonld):

Both changeover branches are computed every tick and unfavSel (Logical.Switch, y = u2 ? u1 : u3) picks one, exactly as AHU-0017 does — same subtractions, same operand order, same temp_deadband, with LessThreshold where that card has GreaterThreshold. econ_unfavorable is therefore the strict complement of AHU-0017’s econ_favorable, in both branches, which is what lets a site read the pair as one policy. minDmpr and excessThr keep PNNL’s two published numbers as two live knobs: the 30% line is computed on the wire, so retuning a site’s minimum position to 30% moves the threshold to 45% without anyone redoing the arithmetic. Both comparisons are strict, so a damper commanded to exactly 30% and a changeover gap of exactly 1.0 °C both read healthy — the second is one boundary point more conservative than PNNL’s NOT (rat − oat > temp_deadband). persist requires 30 continuous minutes, the length of PNNL’s averaging window, which rides out the damper stroke and the minutes either side of changeover; delayOnInit = true holds that window across a controller restart.

Possible Diagnoses

PNNL-27338 §3.4 emits a diagnostic message rather than a fault set, so this list is authored from the mechanisms that hold an OA damper command above minimum when nothing favors economizing:

1. Minimum position parameter set well above the ventilation design minimum — dialled up during a complaint or a commissioning shortcut, and a $0 desk fix. Check it before anything else.
2. Economizer enable with no disable path: the sequence opens on a favorable comparison and never re-tests it
3. Changeover setpoint or high limit left at a factory default that does not fit the climate
4. A changeover device — dry-bulb or enthalpy switch — failed in its “economize” state, a single point of failure with no other symptom
5. OAT sensor reading low, so outdoor air still looks worth importing
6. A mixed-air low-limit or freeze-protection loop holding the damper open, its setpoint never re-tuned for cooling weather
7. An override left in place after service (AHU-0027 finds the flag itself)

Energy Impact

EXCESS_CONSUMPTION, HIGH confidence, DIRECT_MEASUREMENT. Every point of damper command above the minimum imports outdoor air with no ventilation benefit, and the rule fires only when that air is within temp_deadband of return temperature or hotter, so the waste is a cooling penalty: (oa_dmpr_cmd − min_damper_sp)/100 × supply_airflow × ρ·cp × (oat − rat), every term but the airflow already on this rule’s wires. Read it as a floor — sensible-only, and in the near-changeover band where this rule does its distinctive work the latent load of humid outdoor air is the larger half. Cooling-dominant and sharply seasonal.

Emissions Impact

Scope 2, DIRECT_EMISSIONS, HIGH confidence; typical 1,000-6,000 kg CO₂e/yr, matching AHU-0034’s range for the same equipment and the same mechanical cooling. The whole impact is electric compressor or chiller work, and the hours are the grid’s worst — hot afternoons coincident with cooling peaks — so value it at the marginal operating emissions rate (MOER) rather than an average grid factor.

Deviations

• This card is a library extension, not a transcription: chapter 9 specifies AHU-0001..065 and stops. Name, severity 3, phase 2 and method: rule are assigned here — severity 3 to match every other economizer card, phase 2 because the rule presupposes a site that has already configured a changeover type for AHU-0017. The algorithm and every threshold are PNNL-27338 §3.4’s.
• The economizing-unfavorable gate is in the graph, where AHU-0021 and AHU-0030 leave it to the host. Those cards gate on the unit’s economizer mode, an operating state no measurement establishes. PNNL-27338 §3.1 computes econ_condition from two temperatures at every time step, which is a measurement — the library’s standing line (AHU-0030: measured points in, modes out) — and AHU-0017 and AHU-0034 already carry that comparison in-graph. Here it is load-bearing: host-side, this rule would be a bare damper threshold.
• The changeover-type switch is carried even though §3.4’s prose is written around the dry-bulb case. §3.1 defines both branches, and AHU-0017 and AHU-0034 ship them; a site on fixed high-limit changeover would otherwise have this rule contradicting its two siblings on the same unit. econ_type_is_ddb, econ_hl_temp and temp_deadband are their parameters unchanged.
• min_damper_sp × excess_damper_multiplier ships as two parameters, not one 30% threshold. PNNL publishes the minimum as the site value and the excess threshold as its derivative; collapsing them makes a site with a 30% minimum remember to type 45. Feeding the minimum as Reals.Sources.Constant.k and the factor as MultiplyByParameter.k is algebraically identical and keeps both of PNNL’s published names retunable alone (AHU-0021’s precedent for desired_oaf).
• Persistence replaces PNNL’s window average. §3.4 averages damper_signal over a 30-minute data_window (with no_required_data = 10) and tests the mean; this card tests the instantaneous comparison through a 30-minute TrueDelay. The latency matches, the semantics do not: a damper spending half the window wide open and half closed would trip the average and does not trip the delay. Persistence is the house form and the stricter of the two.
• temp_deadband ships at 1.0 °C, not PNNL’s 1 °F (0.56 °C). The value is AHU-0017’s and AHU-0034’s, and one deadband across all three keeps their changeover geometry coherent: that pair brackets a symmetric ±1.0 °C dead zone and this card’s gate is the exact complement of AHU-0017’s. Coherence is worth more than half a degree of fidelity to a threshold PNNL itself calls adjustable. A site with untrimmed sensors must raise it on all three together.
• No evaluability output. AHU-0021 and AHU-0030 must publish yTempDeltaOk because their quotient becomes meaningless as |oat − rat| shrinks. This rule has no quotient, so there is no in-rule condition under which its own answer is unknown, and per SCHEMA.md there is nothing to expose as a y…Ok.
• Both comparisons are strict (> on the damper, < on the changeover gap). At rat − oat exactly equal to temp_deadband this rule stays silent where PNNL’s NOT (rat − oat > temp_deadband) would evaluate — one boundary point, in the direction the library’s strict convention always takes.
• fan_sp is not bound. §3.4.1 lists it among the process’s required points, but it appears only in the energy estimate of Step 7, never in the detection test; fan speed belongs to the host’s energy accounting, and binding a point a rule does not read would misstate the rule’s data requirement.
• The unoccupied case in §3.4’s prose is left to AHU-0026. PNNL extends the check to unoccupied periods, where the damper should be closed rather than at minimum; AHU-0026 already tests exactly that against occ_schedule, so this card’s operating_states restricts it to occupied hours instead of carrying a second expectation for the same actuator.
• The rule reads the damper command, not a measured position (PNNL’s own damper_signal, from the controller). A damper commanded to minimum and mechanically stuck open produces the entire physical fault and none of this signature; run AHU-0020 on the OA damper alongside, as AHU-0034 does.
• suppresses and suppressed_by are empty although this rule nests AHU-0034. At shipped defaults its changeover gate implies this one, its damper threshold is higher, and it adds a cooling conjunct — so every AHU-0034 assertion is also one of these, in both branches. Both findings are true and separately actionable — AHU-0034’s is the sharper diagnosis, this one the wider net — and any suppression edge would be an index-level decision declared on both cards, which single-writer authoring cannot do from here.
• savings_range is inherited, not transcribed. PNNL-27338 publishes no annual percentage for any of its five economizer diagnostics; the band quoted is AHU-0034’s, carried so the family reads consistently, and the card’s own claim is the runtime integral.
• persist.delayOnInit = true (Modelica/CDL default is false), the library’s standing choice: a damper already open past changeover when the controller restarts waits out the full 30 minutes instead of alarming on the first tick.
• Every vector is authored — PNNL-27338 publishes an algorithm and a flow chart, not test cases. With default parameters they reach only the dry-bulb branch, since vectors/v1 stages inputs and not parameters, so hlConst, hlGap and hlUnfav are structurally verified but never reach yFault through u3 (as on AHU-0034); econ_type_is_ddb = false commissions itself.

Notes

The temperature spread decides which of the three excess-outdoor-air cards can speak. AHU-0021 and AHU-0030 infer the fraction from (mat − rat) / (oat − rat) and report NO_EVAL whenever |oat − rat| is under 6 °C. This rule’s gate points the other way — it needs rat − oat under 1 °C — so it owns a roughly 7 °C-wide band, from a degree below return temperature up to where the ratio becomes reliable, in which it is the only one of the three that can answer at all. Above that band both forms work and both should fire: PNNL-27338 §3.5 does exactly that, running the raw damper test beside its OAF check against the same 30% line and reporting both causes when both trip. Below it, in genuine free-cooling weather, this rule is silent by construction. The 1-to-6 °C gap where none of them speaks is honest — mild outdoor air with an unreliable ratio is when an open damper is defensible.

Check the minimum-position parameter before sending anyone to the roof, then command the damper to minimum and watch it move. If the command changes and the fault clears, the sequence never asked for minimum and the fix is at a desk (economizer-failure, remote steps). If oa_dmpr_cmd already sits at minimum while the space says otherwise, this rule was never going to see it, and AHU-0020 is the one to run.

Test Vectors

11 scenarios, clock step 60 s over 3600 s.

{
  "schema": "cxf-library/vectors/v1",
  "clock": {
    "step_s": 60,
    "horizon_s": 3600
  },
  "scenarios": [
    {
      "name": "damper_at_minimum_unfavorable",
      "description": "Healthy: OAT 28 \u00b0C against RAT 22 \u00b0C leaves nothing to economize with, and the damper sits at its 20% minimum \u2014 the correct response",
      "inputs": {
        "oat": 28.0,
        "rat": 22.0,
        "oa_dmpr_cmd": 20.0
      },
      "expect": [
        {
          "output": "yFault",
          "from_s": 0,
          "to_s": 3600,
          "equals": false
        }
      ]
    },
    {
      "name": "damper_above_minimum_unfavorable",
      "description": "The fault: same unfavorable outdoor condition, damper commanded to 60% \u2014 twice the 30% excess threshold; alarms after alarm_delay (1800 s)",
      "inputs": {
        "oat": 28.0,
        "rat": 22.0,
        "oa_dmpr_cmd": 60.0
      },
      "expect": [
        {
          "output": "yFault",
          "from_s": 0,
          "to_s": 1740,
          "equals": false
        },
        {
          "output": "yFault",
          "from_s": 1860,
          "to_s": 3600,
          "equals": true
        }
      ]
    },
    {
      "name": "favorable_for_economizing",
      "description": "Econ gate off on its own: OAT 12 \u00b0C is 10 \u00b0C below return air, so a wide-open damper is free cooling doing its job and no damper position is excessive",
      "inputs": {
        "oat": 12.0,
        "rat": 22.0,
        "oa_dmpr_cmd": 95.0
      },
      "expect": [
        {
          "output": "yFault",
          "from_s": 0,
          "to_s": 3600,
          "equals": false
        }
      ]
    },
    {
      "name": "damper_exactly_at_threshold",
      "description": "Edge case: damper parked exactly on min_damper_sp \u00d7 excess_damper_multiplier (20 \u00d7 1.5 = 30%) \u2014 the strict `>` keeps it out of the alarm",
      "inputs": {
        "oat": 28.0,
        "rat": 22.0,
        "oa_dmpr_cmd": 30.0
      },
      "expect": [
        {
          "output": "yFault",
          "from_s": 0,
          "to_s": 3600,
          "equals": false
        }
      ]
    },
    {
      "name": "damper_just_above_threshold",
      "description": "Edge case, other side: 30.1% clears the threshold by a tenth of a point and the alarm follows at 1800 s",
      "inputs": {
        "oat": 28.0,
        "rat": 22.0,
        "oa_dmpr_cmd": 30.1
      },
      "expect": [
        {
          "output": "yFault",
          "from_s": 0,
          "to_s": 1740,
          "equals": false
        },
        {
          "output": "yFault",
          "from_s": 1860,
          "to_s": 3600,
          "equals": true
        }
      ]
    },
    {
      "name": "deadband_exactly_at_threshold",
      "description": "Edge case: rat \u2212 oat sits exactly on temp_deadband (22.0 \u2212 21.0 = 1.0 \u00b0C) with the damper wide open \u2014 the strict `<` leaves the outdoor condition counted as still favorable, so no alarm",
      "inputs": {
        "oat": 21.0,
        "rat": 22.0,
        "oa_dmpr_cmd": 95.0
      },
      "expect": [
        {
          "output": "yFault",
          "from_s": 0,
          "to_s": 3600,
          "equals": false
        }
      ]
    },
    {
      "name": "deadband_just_inside_threshold",
      "description": "Edge case, other side: rat \u2212 oat = 0.9 \u00b0C closes the deadband, the outdoor condition counts as unfavorable, and the wide-open damper alarms at 1800 s",
      "inputs": {
        "oat": 21.1,
        "rat": 22.0,
        "oa_dmpr_cmd": 95.0
      },
      "expect": [
        {
          "output": "yFault",
          "from_s": 0,
          "to_s": 1740,
          "equals": false
        },
        {
          "output": "yFault",
          "from_s": 1860,
          "to_s": 3600,
          "equals": true
        }
      ]
    },
    {
      "name": "close_temperatures_where_the_ratio_is_blind",
      "description": "The complementarity case: OAT 23.0 \u00b0C against RAT 22.5 \u00b0C is a 0.5 \u00b0C spread, far inside the 6 \u00b0C evaluability gate that puts AHU-0021 and AHU-0030 into NO_EVAL \u2014 this rule still reads a 70% damper as excess and alarms at 1800 s",
      "inputs": {
        "oat": 23.0,
        "rat": 22.5,
        "oa_dmpr_cmd": 70.0
      },
      "expect": [
        {
          "output": "yFault",
          "from_s": 0,
          "to_s": 1740,
          "equals": false
        },
        {
          "output": "yFault",
          "from_s": 1860,
          "to_s": 3600,
          "equals": true
        }
      ]
    },
    {
      "name": "changeover_arrives_midrun",
      "description": "Gate edge: the damper is open at 95% throughout, but outdoor air only rises from 12 \u00b0C to 28 \u00b0C at t=600 \u2014 persistence starts from the changeover, not from t=0, so the alarm lands at 2400 s",
      "inputs": {
        "oat": [
          {
            "t": 0,
            "value": 12.0
          },
          {
            "t": 600,
            "value": 28.0
          }
        ],
        "rat": 22.0,
        "oa_dmpr_cmd": 95.0
      },
      "expect": [
        {
          "output": "yFault",
          "from_s": 0,
          "to_s": 2340,
          "equals": false
        },
        {
          "output": "yFault",
          "from_s": 2460,
          "to_s": 3600,
          "equals": true
        }
      ]
    },
    {
      "name": "damper_closes_before_delay",
      "description": "Transient: the damper drives from 60% back to its 20% minimum at t=1200, clearing the condition short of alarm_delay, so no alarm",
      "inputs": {
        "oat": 28.0,
        "rat": 22.0,
        "oa_dmpr_cmd": [
          {
            "t": 0,
            "value": 60.0
          },
          {
            "t": 1200,
            "value": 20.0
          }
        ]
      },
      "expect": [
        {
          "output": "yFault",
          "from_s": 0,
          "to_s": 3600,
          "equals": false
        }
      ]
    },
    {
      "name": "damper_cycle_resets_persistence",
      "description": "Transient: the damper closes at t=900 and reopens at t=1200 under an unbroken unfavorable outdoor condition \u2014 persistence restarts from the reopen, so the alarm lands at 3000 s, not 1800 s",
      "inputs": {
        "oat": 28.0,
        "rat": 22.0,
        "oa_dmpr_cmd": [
          {
            "t": 0,
            "value": 60.0
          },
          {
            "t": 900,
            "value": 20.0
          },
          {
            "t": 1200,
            "value": 60.0
          }
        ]
      },
      "expect": [
        {
          "output": "yFault",
          "from_s": 0,
          "to_s": 2940,
          "equals": false
        },
        {
          "output": "yFault",
          "from_s": 3060,
          "to_s": 3600,
          "equals": true
        }
      ]
    }
  ]
}