Protection · Device 49

Thermal Overload Protection (ANSI 49): Why a Thermal Relay Has Memory — the IEC 60255-149 Replica Explained

How thermal overload (ANSI 49) protection works: the IEC 60255-149 thermal replica, thermal state %, time-to-trip, and why a warm machine trips about eight times faster than a cold one.

Thermal overload protection — ANSI device 49 — is the protection that keeps a motor, transformer, generator or cable inside its thermal limits, and the key to understanding it is one idea: a thermal relay models heat, not just current. A fuse or a definite-time element looks at the current right now. A thermal relay builds a thermal replica of the plant — a running estimate of how hot the winding actually is — and trips on that. That single difference is why the same overload can be safe on a cold machine and dangerous on a warm one.

The thermal replica

Heating in a winding follows a first-order (single-time-constant) thermal model. The relay computes a thermal state, θ, that rises towards a steady-state value set by the load and decays when the load falls. The steady-state thermal state for an equivalent load multiple I_eq is:

θ_ss = (I_eq / k)² × 100%
      

where k is the threshold/overload factor (the per-unit current the element is allowed to carry continuously) and I_eq is the equivalent load multiple of the thermal basis current. The relay trips when θ reaches 100%. The time to get there, from a preload I_p, is the IEC 60255-149 thermal-replica equation:

t = −τ · ln[ (I² − k²) / (I² − I_p²) ]
      

Here I is the load multiple, I_p is the preload (prior-load) multiple, and τ is the thermal time constant. Two consequences fall straight out of this maths:

The hero: a thermal relay has memory

Take a generic single-time-constant replica with τ = 30 min, threshold k = 1.05, tripping at 100%, and apply a 1.5× overload two ways:

Case Time to trip
1.5× overload from cold 20.2 min
Same 1.5× overload, machine already warm (preloaded at 1.0× rated) 2.6 min

That is the whole point of device 49 on one line: the identical 150% overload trips the cold machine in 20.2 minutes and the warm one in 2.6 minutes — about 8× faster (20.2 ÷ 2.6 ≈ 7.8), purely because the winding was already hot. A definite-time relay or a simple I²t fuse sees the same 1.5× current in both cases and cannot tell them apart. The thermal replica can, because it carries the thermal state forward.

Where it never trips — and where it bites harder

Two more reference points from the same replica:

Case Outcome
1.0× rated steady load (k = 1.05) Never trips — θ_ss settles at 90.7%, 9.3% margin
2.0× overload from cold Trips in 9.7 min

Useful sanity check: if the load is at or below the threshold, no trip time exists — the element will sit in alarm territory and never reach 100%.

Negative sequence heats the rotor too

On motors, current unbalance is a thermal problem, not just a current one: negative-sequence current induces double-frequency rotor currents that heat the rotor disproportionately. The replica accounts for it by forming an equivalent heating current before the thermal maths runs:

I_eq = √( I₁² + k_ns · I₂² )
      

where I₁ and I₂ are the positive- and negative-sequence currents and k_ns is the negative-sequence weighting. Feed I_eq into θ_ss and the trip equation and the unbalance shows up as extra thermal state — exactly as it does in the iron.

One engine, many relays — vendor-faithful, honestly bounded

Real relays implement the same physics with their own settings, thresholds and refinements. The thermal model covers a generic single/dual time constant plus vendor-faithful profiles: Siemens 7SJ / 7SD5 / SIPROTEC 5 (3-phase advanced, 1-phase, hot-spot), MiCOM P14x / P34x (generator) / P44x / P54x / P64x (transformer), ABB REX640 motor (MPTTR) and transformer (T2PTTR) — across motor, transformer, generator and cable. On top of the single-time-constant core it adds dual time constants (winding + oil), hot/cold/warm/custom start, preload memory, alarm/trip thresholds, cooling/reset and restart inhibit, ambient-temperature influence, emergency start (Siemens 7SJ — this defeats the trip output only; the thermal image keeps running), and transformer hot-spot/top-oil branches per IEC 60076-7 / IEEE C57.91 with an aging-rate. Thermal-state, trip-curve and cooling curves are all plotted.

Be clear about what these vendor branches are: vendor-faithful but bounded models. The engine itself flags, for example, that the MiCOM P64x transformer path is held "as a bounded transformer study rather than a clause-by-clause IEEE C57.91 replica." The generic IEC 60255-149 replica is the rigorous core; the vendor profiles are interpretations sized to the standard, not certified firmware.

See it on your relay

The Thermal Overload (Device 49) suite models the IEC 60255-149 thermal replica across Siemens 7SJ, MiCOM P64x, ABB REX640 and the generic profiles — showing thermal utilisation %, the time-to-trip cold vs warm, the steady-state thermal state, cooling/reset, negative-sequence heating and the transformer hot-spot/top-oil branches, each in the relay's own vocabulary. It plots the thermal-state, trip and cooling curves, and it does not invent a value you have not entered.

Frequently asked questions

What is ANSI device 49 / thermal overload protection?

ANSI/IEEE device 49 is thermal overload protection: a relay function that estimates the thermal state of the protected plant (motor, transformer, generator, cable) from its load history and trips before the insulation overheats. The modern reference for the thermal-replica behaviour is IEC 60255-149 (legacy IEC 60255-8).

What is the IEC 60255-149 thermal replica equation?

The single-time-constant replica trips when the thermal state reaches 100%, at a time t = −τ · ln[(I² − k²) / (I² − I_p²)], where I is the load multiple, I_p the preload multiple, k the threshold/overload factor and τ the thermal time constant. The steady-state thermal state is θ_ss = (I_eq/k)².

Why does a thermal relay trip a warm machine faster than a cold one?

Because of the preload term I_p. With τ = 30 min and k = 1.05, a 1.5× overload trips a cold machine in 20.2 min but the same overload trips a machine already preloaded at 1.0× rated in 2.6 min — about 8× faster. The thermal state starts higher, so it reaches 100% sooner. A fuse or definite-time element has no memory of prior load and cannot make that distinction.

Does a thermal overload element always trip on overload?

No. It only trips if the load multiple exceeds the threshold (I > k). At exactly 1.0× rated load with k = 1.05, the steady-state thermal state settles at 90.7% and never reaches 100% — it stays in alarm territory but never trips.

How much faster does a bigger overload trip?

From cold, a 1.5× overload trips in 20.2 min while a 2.0× overload trips in 9.7 min — the trip time falls steeply as the load multiple climbs above the threshold.

Does unbalance affect thermal protection?

Yes — especially on motors. Negative-sequence current heats the rotor, so the replica forms an equivalent heating current I_eq = √(I₁² + k_ns·I₂²) before computing thermal state, capturing the extra heating from unbalance rather than just the phase current magnitude.

Which relays does the EI Portal tool cover?

Generic single/dual time constant plus Siemens 7SJ / 7SD5 / SIPROTEC 5, MiCOM P14x / P34x / P44x / P54x / P64x, and ABB REX640 (motor MPTTR, transformer T2PTTR), across motor / transformer / generator / cable. It is a study/interpretation workspace, not certified firmware, and the transformer hot-spot/top-oil branches are bounded studies.

Related on EI Portal