Why an underground train can become sound inside a room

From the track to the room

This is usually described as ground-borne vibration and ground-borne sound. The two are related but they are not interchangeable. A resident may feel vibration in a floor or object, hear a low rumble radiated by the building surfaces, experience both, or notice neither at a particular position.

That distinction changes the investigation. Airborne traffic noise is commonly assessed at a façade or through a window. A ground-borne problem requires attention to the source, the tunnel and track, the soil, the building foundations and structure, and finally the room in which structural movement becomes audible sound.

What the new study adds

A September 2026 peer-reviewed study combined a train-track-tunnel-soil-building model with a room-acoustic model. The researchers compared predictions with measurements in a five-storey residential building close to a metro tunnel. They reported reasonably strong agreement for both vibration acceleration and indoor sound pressure level, with coefficients of determination of about 0.80 and 0.78 respectively.

The value of the work is the joined-up path. It does not stop at vibration at the foundation; it follows how structural response can radiate sound into occupied rooms. The simulations also showed that the response was not uniform between floors or positions within a room.

A higher floor is not a simple escape

It is tempting to assume that the effect must always reduce with height. In the modelled buildings, some middle floors were quieter while upper floors could show amplification at particular frequencies. That is plausible because a building is a dynamic system: its mass, stiffness, spans, supports and natural modes influence how energy travels and where surfaces respond most strongly.

This does not mean that upper floors are generally worse beside a railway. The field validation covered one five-storey building and the taller cases were simulations. Construction, geology, tunnel depth, track form, train operation and room geometry all matter. Floor number alone is not a prediction.

Why one reading can miss the pattern

Low-frequency sound in a room can vary substantially with position because room modes create areas of reinforcement and cancellation. A measurement near one wall or at one height may therefore differ from another position in the same room. Structural vibration can also vary across a floor slab.

A useful survey needs repeatable train events, suitable vibration and acoustic instrumentation, frequency analysis and clearly recorded measurement positions. It should separate what is felt from what is heard and distinguish the railway event from lifts, pumps, ventilation equipment and other building services.

Mitigation has to follow the path

Changing a window is unlikely to address sound generated by floors or walls. Effective controls may instead relate to the railway source, track isolation, the transmission path, foundations or the dynamic behaviour of the receiving structure. In an existing home, the first task is to confirm the dominant mechanism before discussing building work.

The study explored structural changes in its model, but its numerical results should not be treated as a general retrofit promise. An intervention that changes one building mode can leave another path unchanged. UK project guidance for new railways similarly treats ground-borne sound and vibration as specific effects requiring prediction, design and, where appropriate, monitoring.

A practical investigation

For a home near an underground railway, a short evidence record can help turn a vague rumble into a testable problem.

• Record the times, rooms and positions in which the event is heard or felt.
• Compare the pattern with train movements rather than relying on proximity alone.
• Note whether windows open or closed make a material difference, without assuming that this identifies the whole path.
• Measure vibration and indoor sound during the same events and retain frequency information.
• Sample more than one useful position where low-frequency variation is suspected.
• Base mitigation on the confirmed source-path-receiver mechanism, not on a generic product.