Lesson 11

Logger measurements

The audit of lesson 10 rests on figures typed from a nameplate and a watch. This lesson checks them against the site: a gauge and a flow meter on the network, a week of logger records from the compressor, and its measured power. Where the model and the site disagree, the audit says which of them to doubt.

Three kinds of measurement

1. Start from the audit of lesson 10

   Open lesson10.gf9 and save it at once as lesson11.gf9. It opens in audit mode with everything lesson 10 entered:

   Item	Row	Value
   COMP	Audit pressure	6.5 bar
   COMP	Audit capacity	2500 NL/min
   COMP	Audit leak t load / t unload	6 / 54 s
   B1	Audit utilization	50 %
   B2	Audit utilization	30 %
   B2	Audit required pressure	4 bar

   Press Audit once and check COMP against lesson 10: Qn 1960.35, Qn average 995.68 NL/min, Leak share 25.04 %, Min audit pressure 6.1832 bar. If a figure differs, compare the rows of the table above first.

2. Put a flow meter and a gauge on the network

   A site instrument goes on the sheet where it sits on the plant, in line, like the regulator of lesson 8: release a connection, place the item, join both sides.

   Instrument	Where	Name
   A flow meter: VD 500 from Sensors (type VD 500 in the filter)	between COMP and P1, on the main line	FM
   A pressure gauge: CS 16 from Sensors	between P1 and T1, at the end of the main	PG

   Both are CS Instruments cards. CS 16 is an in-line pressure sensor with a 16 bar scale, accuracy 1 % of it. VD 500 is an insertion flow sensor for wet air: accuracy 1.5 % of the reading and nothing of the full scale, so its card has no Range. Rename them (F2) and press Audit. An instrument has no pressure loss: the results of lesson 10 stay as they were. Each gets its own audit rows:

   Row	Meaning
   Audit measured	The reading: bar gauge for PG, a flow for FM.
   Audit range, Audit accuracy class	Full scale and accuracy, % of the full scale. They appear only when the card has no Range or Accuracy row of its own — a site instrument that is not in the catalog.
   Audit load point	When the reading was taken: 1 — at maximum load (empty means 1), 2 — at average load. After the audit the row load_point repeats it as a result.

   PG has a Range and an Accuracy in its card, so it gets no Audit range or Audit accuracy class. FM gets Audit range: its card gives no full scale. Leave it empty — VD 500 is judged by its reading.

   FM shows its Qn even before a reading: 1960.35 NL/min, all the air of COMP — the flow the model expects the meter to see.

3. Readings that agree with the model

   The auditor walked the network at maximum load — both blow-offs on — and wrote down:

   Item	Audit measured
   PG	6.35 bar	the gauge at the end of the main
   FM	1966 NL/min	118 Nm³/h on the meter's display; the row takes the unit of Qn

   Press Audit. Each instrument gets a Tolerance — how far the reading may be from the truth — and, when the model falls inside it, a row Audit: within accuracy with the difference reading − model:

   Row	PG	FM
   Pin / Qn (model)	6.3755 bar	1960.35 NL/min
   Tolerance	0.16 bar	29.49 NL/min
   Audit: within accuracy	−0.0255 bar	5.6524 NL/min

   Check by hand: the tolerance

   The tolerance is u = Accuracy × Range + Accuracy reading × reading. For CS 16: 1 % × 16 bar = 0.16 bar. For VD 500: 0 + 1.5 % × 1966 = 29.49 NL/min. A site gauge without a card is taken as class 1.5 on the nearest standard scale not below 4/3 of the reading — 4/3 × 6.35 = 8.47, so 10 bar and 0.15 bar — and the note of the row marks every such assumption with ~.

   Both agree: the model of lesson 10 describes this plant. That is the answer you hope for — and the one to have on file before anybody acts on the measures.

4. A gauge that does not agree

   Now suppose PG had read 6.0 bar. Type 6 into its Audit measured and press Audit.

   The reading is 0.3755 bar below the model — more than the tolerance. A gauge below the model is believed: the audit holds that node at the reading, solves the network again, and everything downstream follows the gauge. The loss of pressure between the compressor and the gauge is real; the model now has to explain it.

   Item	Row	Expected
   PG	Pressure deviation	−0.3755 bar	alarm: the node is held at 6.0
   PG	Qn unaccounted	alarm	as a leak it would need more air than the compressor delivers; the value shown is only a ceiling, what the capacity leaves for it
   B1	Pressure deficit	≈ 0.06 bar	B1 gets less than its 6 bar
   COMP	Qn	2500	NL/min — at its capacity
   COMP	Pout	6.21	bar — below its setting
   COMP	Demand above capacity	0.29	alarm: bar below the setting

   To pull the end of the main down to 6.0 bar the model needs far more air through P1 than the compressor can give. COMP runs flat out at 2500 NL/min, its outlet falls to 6.21 bar, and the leaks, at lower pressures, shrink to 236 NL/min.

   Qn unaccounted is the air the node would have to lose to fall to 6.0 bar — an equivalent leak between the compressor and the gauge. Here it is an alarm: the 25 mm main loses so little that only a huge leak could pull it down by 0.37 bar, more air than the compressor can give. So the drop is more likely an extra resistance — a half-closed valve, a crushed pipe, a dirty filter nobody drew on the scheme.

   The flow meter settles it. FM still reads 1966 NL/min:

   Row, FM	Expected
   Audit: within accuracy	5.6524 NL/min	the reading matches the flow without the gauge correction
   Diagnosis: resistance		alarm: the flow did not grow by Qn unaccounted

   The meter on the main reads…	Diagnosis	Why
   the flow of the model before the gauge was held	Diagnosis: resistance	No extra air passes: something in the line eats the pressure.
   the flow after it, with Qn unaccounted	Diagnosis: leak	The extra air really goes through the meter and is lost before the gauge.

   The audit does not compare a gauge that reads above the model with anything downstream: it cannot hold a node higher than the air arriving there allows. It only reports Pressure deviation and, when the modelled loss to the gauge is larger than the tolerance, Loss upstream — the measured loss as a share of the modelled one. Here the model loses only 6.5 − 6.3755 = 0.12 bar on the way to PG, less than its 0.16 bar tolerance: the gauge could not tell that loss from zero, so PG gets no Loss upstream at all.

   This plant has no such resistance — it was a thought experiment. Set Audit measured of PG back to 6.35 and press Audit: the alarms go and COMP is back at 1960.35 NL/min.

5. Load a week of logger records

   Single readings say how the plant stands at one moment. For a week the auditor leaves a data logger on the compressor: its power, the network pressure and the flow, every 30 seconds. Save lesson11-ald.csv — an export of an ALD logger, in American units: kW, psig, scfm.

   Click COMP and double-click its row Audit measurement data — it reads no data - double-click to import. The Measurement data window opens. Press Add files... and choose the file. The Import logger file window opens for every file, even when it is recognised, so that you see how it will be read before anything enters the scheme:

   Field	Shows	Why
   Profile	ALD logger export	Recognised by the header Timestamp,kW,psig,scfm.
   Delimiter, Decimal	comma, point
   Header line	1
   Order	M/D/Y	The file holds 2/17/2025: 17 cannot be a month, so the order is decided by the file, not guessed.
   Layout, Time column	wide: one column per channel; Timestamp	One row per moment, one column per channel, date and time in one column.
   Zone	UTC+03:00, no DST — the zone of this computer	The time stamps carry no zone of their own. For a logger from another time zone set its zone here.

   The table of columns at the bottom:

   Column	Use as	Unit	Bound to	Sample min..max
   kW	Power P	kW	K1	0..0
   psig	Pressure p	psig	Collector	0..0
   scfm	Flow Q	scfm	Collector	−82..−82

   A single compressor has one machine, so power goes to it; pressure and flow are measured on the collector — the pipe the compressor feeds. Sample min..max covers only the first lines, here the night before the first shift: the machine is off, the pressure is zero — and the flow reads −82 scfm.

   Press Import. EasyCAS spots the −82 at once:

   scfm: 100.0 % of the values are negative and 76.2 % are exactly -82. This looks like a constant subtracted in the logger or in the export (a stopped machine reads -82 instead of 0). Add b = 82 to this column, so that the stopped state reads 0?

   Answer Yes. ALD exports really do this. The offset goes into the corrections of this file (a = 1, b = 82); the file itself is not changed.

6. Read the week

   The Measurement data window now holds the file:

   Table	What to check	Expected
   Files	Period, Rows, Rejected, Profile	17.02.2025 00:00 – 24.02.2025 00:00; 20161; 0; ALD logger export
   Channels	Quantity, Bound to, Coverage	kW → K1; psig → Collector; scfm → Collector; 100 %
   Notes	Analysis period	17.02.2025 00:00 – 24.02.2025 00:00 (whole weeks)

   Pressure and flow are drawn in the units of the Units tab — bar and NL/min — power in kW. Each channel has a band of its own, and the analysis period is shaded: one whole week from Monday, because the file covers one. Turn the mouse wheel to zoom into a working day, and the week reads like a story:

   • the compressor runs from 07:00 to 15:00, Monday to Friday, and stands the rest of the time: 40 hours a week;
   • while it runs, the power jumps between about 15 kW (loaded) and 4.5 kW (unloaded): a load/unload machine;
   • the pressure saws between 6.0 and 6.5 bar: it unloads at 6.5 — the Audit pressure of lesson 10 — and loads again at 6.0;
   • the flow swings with the blow-offs, never below the leaks.

   The card row of COMP now reads 17.02-24.02.2025, 3 ch. The records are kept in lesson11.gf9, compressed; the CSV file is not needed any more, but keep it for Re-import file....

   What the records do — and do not do yet. In this version the logger records are loaded, bound, shown and stored. Statistics of the machines and the calibration of the model against them come in the next version; the notes of the window say so. Until then the bridge from the logger to the calculation is the next step: the measured mean power.

7. Turn the records into Audit power

   Press Export hourly CSV... and save the file. It holds the hourly means of every channel, one row per hour and channel, in the units of the file:

   timestamp;channel;value;unit
   2025-02-17 07:00;kW;8.706…;kW

   The channels keep the names and units of the logger — kW, psig in psi, scfm — and the correction b = 82 is already in the flow. The file is called hourly_ID7.csv unless you name it otherwise.

   Open it in a spreadsheet (Data → From Text/CSV, delimiter semicolon, decimal point) and average the power over the hours the compressor ran: =AVERAGEIFS(C:C, B:B, "kW", C:C, ">0"). Forty hours, mean 8.71 kW.

   Check: the logger meets the model

   Average the flow the same way: 34.73 scfm while running. One scfm is 28.69 NL/min (101325 Pa against the 1 bar of Normal pressure), so 34.73 × 28.69 = 996 NL/min — and the audit of lesson 10 gave Qn average 995.68. Utilization, leaks and consumers were guessed in lesson 10; the logger says the guesses hold. The pressure averages 6.26 bar, the middle of the 6.0–6.5 band — what an accepted load/unload machine would give as Pout average (lesson 9).

   Click COMP and fill:

   Row	Value
   Audit power	8.7 kW	mean electrical power while running
   Audit hours	empty	40 h a week × 48 weeks = 1920 h — the same as 8 h × 240 days of Input parameters

   Press Audit. COMP gets the energy by measurement, and the estimate of lesson 10 is scaled to it:

   Row, COMP	Before	Now	Unit
   Electricity consumption	22223	22223	kWh	the estimate from Compressed air cost
   Spec power audit		8.738	kW/Nm³/min	Audit power / Qn average
   Annual energy audit		17	MWh	8.7 kW × 1920 h
   Annual cost audit		2172	€	at Electricity cost 0.13 €/kWh
   Power calibration		0.752		measured / estimated energy
   Annual energy calibrated		17	MWh
   Annual cost calibrated		2172	€
   Leak energy	5.6	4.2	MWh	25.04 % of the calibrated energy
   Leak cost	723	544	€
   Artificial demand energy	1.4	1.0	MWh
   Artificial demand cost	180	136	€

   Check by hand: the calibration

   Without a machine, lesson 10 priced the air with Compressed air cost: 2889 € a year, that is 2889 / 0.13 = 22 223 kWh. The meter says 8.7 kW × 1920 h = 16 704 kWh. Their ratio, 16 704 / 22 223 = 0.752, is Power calibration: every energy and money figure of this source is multiplied by it — the leaks are 25.04 % × 16.70 MWh = 4.18 MWh and 25.04 % × 2172 € = 544 €. The specific power, 8.7 kW for 0.9957 Nm³/min, is 8.74 kW per Nm³/min — high, as it is for a load/unload machine that runs loaded only 40 % of the time.

   The estimate from the price of air was a third too high: this plant makes its air cheaper than the 0.024 €/Nm³ of Input parameters says. Every saving of the audit is now counted at the measured energy — the measures in the Problems panel too (the lower setting is now worth 154 €/yr instead of 204), and, in lesson 12, the scenarios, which take the same calibration from this baseline.

   Had the compressor run two shifts, the logger would have shown 80 hours a week: Audit hours 3840 would then set the year of this source, for the measured energy and for the estimate alike.

8. Save it

   Save lesson11.gf9. It keeps the readings, the logger week and the measured power. Lesson 12 applies the measures of lesson 10 in a scenario of their own and counts what they pay back — at the energy you have just measured.

When it does not work

Try this next

• Type 2500 into Audit measured of FM, with PG at 6.0. The meter now agrees with the flow after the gauge was held: Diagnosis: leak, with the equivalent leak in Nm³/h. Set both back.
• In the Measurement data window, drag over Tuesday's shift to set the analysis period by hand, then press Default period to get the whole week back.

Related help topics: