On drives and energy
The affinity laws have a condition attached, and it is the condition that gets dropped
Power varying with the cube of speed only holds where the system curve is dominated by friction. On a pump working against high static head the saving is modest. On a membrane feed pump, for example reverse osmosis, the membrane sets the required pressure, so the duty behaves as static head and the saving is effectively zero. A proposal that recovers the missing saving by opening the concentrate valve has stopped being an efficiency measure and become a change to the recovery ratio, with consequences for scaling and membrane life.
Full note on VFD retrofit and where the savings are real
Enclosure height is the drive body plus the manufacturer's airflow clearance
Show the sum on the drawing. A clearance quietly absorbed to make a layout fit produces thermal trips that get diagnosed as drive faults for months before anyone measures the air.
Full note on panel and MCC design
On control systems
Redundancy forbids local I/O in the redundant chassis
So "add a second processor for redundancy" is never the real bill of materials. The I/O has to move to a remote chassis, which means network modules, a second chassis, cabling and a rework of every I/O reference in the program. A quotation showing a processor and a redundancy module has been priced, not engineered.
Full note on PLC and DCS migration
A station-class SCADA product may not be able to serve clients at all
On some widely used platforms the station product is standalone by design, so a second operator screen is a second station licence and a second machine, not a client. On the same platforms, read-only web clients can be free. That single distinction can be the whole budget difference between two quotations.
Full note on SCADA licensing and architecture
"It still runs in manual" has to be verified at the wiring
If the manual path runs through a PLC output, a processor in STOP kills manual too. Check it before any recovery plan leans on manual operation.
Full note on troubleshooting method
Read the diagnostic register before clearing it
An error code cleared without being read has destroyed the only evidence of what happened, and the plant now waits for a recurrence to learn what it already knew once.
On procurement and parts
A serial number lookup on a manufacturer's spares portal does not prove authenticity
Those portals confirm that a catalogue number exists and is orderable. They do not confirm that the serial on the label in front of you was ever issued. What carries information is provenance, firmware and date code consistency, physical comparison against a known-genuine unit, and a bench test. And genuineness belongs on the purchase order as a condition of payment, not as a hope.
Non-English labelling is not a red flag
Genuine product is manufactured and distributed globally. Judging authenticity by the language on the sticker produces false alarms, and worse, false confidence whenever the label happens to look familiar.
Full note on spares and obsolete parts
On surveys and monitoring
A thermal survey at light load will pass equipment that fails at full load
Temperature rise goes with the square of current, so the same defect at half load shows roughly a quarter of the rise. Every image needs the load recorded next to it, and a survey taken on a quiet shift should say so on its face.
Check the sensor's own temperature ceiling before specifying continuous monitoring
Many common contact temperature sensors top out below the danger band that electrical maintenance guidance asks you to detect. A sensor that saturates before the alarm condition cannot raise it. Check it against the datasheet at specification, not after installation.
Full note on thermal and PD survey
Report condition on operating hours, not on the calendar
Duty and standby machines rotate. A calendar-based report shows the rested machine as healthy and hides how hard the working one has been driven. This single change makes more difference to a monitoring programme's credibility than any improvement in the analytics.
Guarantee the warning, never the outcome
No monitoring system can honestly guarantee a machine will not fail. What can be committed to is the warning, within a stated detectable envelope, with a stated consequence for a missed warning on an asset reported healthy. A vendor guaranteeing the outcome has either not read their own detection limits or does not expect to be held to it.
Full note on condition monitoring
On fault levels
Work the prospective fault current before accepting a board
A transformer's approximate short-circuit current at its secondary is its rated current divided by its percentage impedance. A 2,500 kVA transformer at 6 % impedance gives roughly 55 kA, which makes a board specified with 50 kA devices non-compliant on the day it is installed, however well it is built. This is arithmetic, it takes a minute, and it is the most common thing found wrong on a panel bought on price.
Full note on panel and MCC design