A flow switch confirms one condition at its installation point. It does not prove that each cooled lead receives the intended flow or transfers heat adequately. A blocked branch, bypass or poor local heat transfer can coexist with a normal total-flow indication. IEC 61378-1 covers industrial converter-transformer specification, design and testing [1].
Define the cooling circuit for the actual lead. Identify branch flow, inlet and outlet temperatures, conductor measurement positions and the operating current. Keep the monitoring boundary local enough to observe the failure of concern.
Compare heat input and removal
For a steady coolant stream, Qwater = ṁcp(Tout − Tin), with mass flow ṁ in kilograms per second, specific heat cp in joules per kilogram-kelvin and Q in watts. If flow is measured volumetrically, use the appropriate density to obtain mass flow.
The relation assumes representative temperatures and a defined stream. Mixing several branches before measuring return temperature can conceal a poorly cooled individual lead.
| Observation | What it indicates | Important blind spot |
|---|---|---|
| Branch flow | Water movement through that route | Heat transfer may still be inadequate |
| Return temperature rise | Heat carried by the measured stream | Mixed returns can hide local variation |
| Conductor temperature | Local thermal state | Sensor may miss the hotspot |
| Current and resistance basis | Approximate heat input | Additional losses and joints may differ |
| Sensor health and timing | Trustworthiness of the alarm evidence | A stale signal can resemble stable operation |
Establish the available reaction time
A simplified lumped balance is C dT/dt = Ploss − Qwater − Qother, where C is effective heat capacity in joules per kelvin. If the net heating power is approximately constant and positive, time to a temperature rise ΔT is CΔT/Pnet.
That estimate is a screening calculation. Actual local temperatures, resistance growth and changing cooling can produce a different trajectory. Validate the relevant thermal response rather than selecting an alarm delay from the formula alone.
Include sensor lag, data processing and the approved system response in the timing assessment. An alarm that appears before an average temperature limit can still be too late for a small joint hotspot.
Reconcile branch water heat with the reaction interval
Assume a steady branch at approximately 25 °C inlet, mass flow 0.10 kg/s, temperature rise 5 K and water heat capacity 4180 J/(kg·K). Removed heat is 2090 W. Against assumed local loss 3000 W, net heating is 910 W. With assumed effective heat capacity 20 kJ/K and 20 K remaining temperature margin, constant-net-power screening gives 20,000 × 20/910 = 440 s. Reserve detection and response time within a validated local thermal model; a normal common-loop flow signal cannot supply the missing 910 W.
Challenge the monitoring logic
Use approved representative tests or validated fault models for reduced branch flow, higher inlet temperature and sensor failure. Do not deliberately remove cooling from an operating high-current lead without an authorized qualification procedure.
Set thresholds from the actual lead limits and demonstrated monitoring capability. Avoid treating one universal flow or return-temperature value as adequate for every load.
Base the protection decision on the lead’s local thermal response and the demonstrated detection interval. A normal main-loop flow signal cannot clear a branch whose conductor temperature or heat balance shows that cooling has deteriorated.
Retain an explicit measurement model and input uncertainties when comparing measured and calculated duty [2].
References
- IEC, IEC 61378-1:2011, Converter transformers – Part 1: Transformers for industrial applications.
- JCGM, JCGM 100:2008, Evaluation of measurement data — Guide to the expression of uncertainty in measurement, 2008.