Victim Locator Triangulation From Passive Breath and Tap Signals
The Physics and the Gap
USGS researchers published Finding Trapped Miners Using Music-Recording Hardware describing a prototype system that detects trapped-miner tap signals using off-the-shelf audio equipment. The older NIOSH/Bureau of Mines Seismic Detection of Trapped Miners Using In-Mine Geophones report documented in-mine geophone arrays with more than 1000 feet of detection range. An Academia-hosted analysis of calculating surface seismic signals from trapped miners models how tap signals propagate through fractured rock to surface arrays. Australian trials in Investigating Microseismic Monitoring for Trapped Miners documented 100 percent detection at 65 meters and 90 percent at 84 meters using pre-existing microseismic arrays.
The physics is solved. What is not solved is translating detection into a map location the command post can act on. A Springer review in Through-the-Earth Communication Systems describes TTE seismic fingerprint triangulation as a promising direction, but commercial kits from vendors like Delsar LifeDetector and LEADER USAR seismic sensors still produce bearings rather than 3D positions. The rescue coordinator gets "signal detected, estimated 40-80 meters, east-northeast" and has to map that cone onto a collapse geometry that itself is uncertain.
The operational consequence is wasted rescuer hours. A cone projected onto an outdated map intersects multiple possible miner locations; a crew must be sent to each. If the map is wrong by 20 meters, the crew may be searching the wrong pillar. Post-Sago and post-Upper Big Branch investigations have repeatedly flagged the gap between detection and localization as an area with room for improvement.
The detection-to-localization gap also has a temporal cost that compounds with the rescue clock. Each false-direction advance consumes 30 to 90 minutes of rescuer time, and breathing-apparatus duty cycles cap how many false-direction advances any single team can absorb before the team must rotate. A four-team mutual-aid response budget can be exhausted by three or four bad bearings on a single suspected miner location. Coordinators who track localization-quality metrics across past incidents often find that the median trapped-miner rescue spends more time chasing false bearings than executing the eventual successful approach. Tightening probability cones on the live map is therefore not just a precision improvement; it is a duty-cycle improvement that determines how many independent localization attempts the response budget can sustain.
Stitching Detection Events Into Probability Cones
EchoQuilt's approach is to attach triangulation data as probability cones directly onto the live 3D quilt, updating in real time as more sensors report. When a rescuer's belt node detects a candidate breath-rate signal or tap pattern, the system logs the arrival time, bearing, and signal characteristics. As additional rescuers move through the collapse zone, their nodes pick up the same signal from different positions. The quilt intersects the cones, tightening the probability volume patch by patch until the command post sees a bounded region rather than a bearing.
The stitching logic treats tap signals and breath-rate signals differently. Tap signals are transient, discrete events with characteristic frequency content in the 50-500 Hz band that distinguishes them from mine equipment noise. Breath-rate signals are sustained periodic patterns at 0.1-0.4 Hz envelope modulation over a broadband carrier. Both can be detected; both can be localized; but they require separate classifiers because the physics of propagation differs. Tap signals are compressional waves through rock; breath signals propagate through air gaps in the rubble. The quilt's sensor-fusion layer maintains both as separate probability cones that can be combined only when the command post confirms both are attributable to the same miner.
Localization accuracy depends on rescuer geometry. Three rescuers listening from three different bearings can typically bound a tap signal to within 10-15 meters. Four rescuers in a tetrahedron configuration can typically bound it to within 5-8 meters — fine enough to direct the next advance to a specific pillar or crosscut. The command post can see the probability volume shrink in real time as rescuers move, which also tells the IC when the current geometry has extracted all available information from a given signal and additional rescuer positioning will not help.
The geometry-of-the-rescuers question is itself worth dedicated planning: a captain who positions four rescuers in a near-collinear arrangement may technically have four bearings but will not gain the spatial diversity needed to tighten the cone, while a captain who sends one of the four to climb to a higher elevation in an adjacent crosscut adds the vertical baseline that pulls the probability volume down to a small ellipsoid.
This is where advanced triangulation at chokepoints becomes practical — narrowing the probability volume near chokepoints where trapped miners often congregate saves hours.
Advanced Tactics for Tap and Breath Localization
Three tactics separate usable triangulation from theoretically-valid triangulation. First, synchronize clocks across all rescuer nodes to sub-millisecond precision. Tap-signal arrival time differences between nodes are in the millisecond range for short distances. If node clocks drift by more than a few milliseconds between synchronizations, the intersecting cones come apart. EchoQuilt uses a combination of mesh-network time sync and dead-reckoning clock correction to hold node clocks within tolerance.
Second, maintain a library of expected tap and breath signatures. A trapped miner striking a steel rib with a rock produces a different signature than the same miner striking a coal pillar. Known signature templates can be matched against incoming candidate signals to reduce false positives from natural mine noise. The library should include SCSR regulator signatures too, because a trapped miner's SCSR will emit a regulator tone as the miner breathes — and that tone is a reliable confirmed-presence signal even when the miner is not actively tapping.
Third, filter for the window after a tap sequence ends. Trapped miners typically tap in bursts separated by minutes of silence. The triangulation system should keep the candidate cone active for at least 20 minutes after a tap burst, because the next burst from the same miner will tighten the localization. Discarding the cone after a single burst loses this accumulation. The same underlying principle runs real-time detection workflows that flag possible miner signals as they emerge rather than waiting for post-shift analysis.
A common mistake is to treat a single tap signal from a single sensor as actionable. It is not; it is a trigger to reposition rescuers so that subsequent taps can be triangulated. The command post must resist the urge to dispatch crews on a single-bearing signal. A two-rescuer, two-bearing confirmation before advance is the minimum discipline. This is what the post-Sago and post-Crandall Canyon rescue-team best-practice reviews pushed toward, and what EchoQuilt's probability-cone display makes enforceable at the tablet level.
Coordinators running entrapment exercises should also stress-test the system against ambient mine noise patterns specific to their operations. A longwall section produces continuous shearer noise that can mask faint tap signals; a room-and-pillar section idle after the production shift may produce intermittent rib-creep noise that resembles tap patterns. The classifier's false-positive and false-negative rates depend on the noise environment, and tuning the thresholds against the mine's actual noise profile is the difference between a triangulation system that works in the lab and one that works at 3 a.m. during a real entrapment. Coordinators who invest two or three half-day sessions in noise-environment characterization see meaningfully better classifier performance during real responses than coordinators who use vendor-default thresholds.
Join the Waitlist for Mine Rescue Coordinators
Mine rescue coordinators preparing for entrapment scenarios can request early access to the victim-triangulation module. We pair the trial with SCSR signature calibration for your team's current regulator inventory, historical tap-pattern templates from MSHA post-incident archives, and node-clock sync testing against your existing comms infrastructure. Priority goes to teams with active entrapment drills on the calendar and to operators whose ERPs include trapped-miner response protocols. Send us your drill schedule and we will scope a triangulation field test against your next exercise. The passive-listening pattern parallels passive sound trends that cave divers use when mapping unexplored conduits — both rely on continuous passive observation rather than active pings, and both benefit from the same triangulation math.