Blast Vibration Forensics

Headline metrics live

Velocity traces & limit line

All three components plus vector sum. The dashed line is your limit; the marker is the peak vector sum.

Computed indicators live

Install representativeness live

Separates electronic health (from SNR) from how representative the mount is of free-field ground. Driven by your Step 1 mount, ground, crest and nearby-monitor answers. A monitor can be electronically perfect and still be reading a local slab, not the rock mass.

First-arrival particle motion (orientation)

Horizontal hodogram of the first strong arrival. If the geophone faces the blast, early body-wave motion should line up along the source bearing (radial). A horizontal azimuth far off the bearing hints at mis-orientation or a horizontally-driven local response.

Scaled-distance prediction vs measured live

Uses PPV = K·(D/√W)^−B from your Step 1 inputs. Blair (2004) shows the single-W/single-D choice is unsound near-field, so treat this as a band, not a line.

Free-surface & geology context live

Cadence → frequency: source vs monitor live

The firing cadence of the blast design sets the frequency leaving the pattern; the monitor records what survives the trip. Load a hole pattern in Step 8 (with Detonation_ms) to populate the design cadence.

Source → monitor animation live

Indicative wave-type energy split live · estimate

A single triaxial trace can't fully separate wave types, so this is an indicative split from band energy, horizontal/vertical balance and motion rectilinearity — not a definitive decomposition. It flags which family likely drove the peak.

The waves in the system learn

Spectrogram — time vs frequency live

Where each frequency's energy sits in time. Early high-frequency body-wave packets separate from later low-frequency surface/local energy here in a way the whole-window FFT can't show.

Energy packets live

Auto-detects vibration packets from a smoothed vector envelope: did the ground settle between arrivals, which axis drove each packet, and at what frequency. Repeated non-settling packets point to stacking; one isolated packet points to a single hole/event.

Sliding-window polarisation live

Eigen-analysis of the 3-component motion in a moving window. High linearity = rectilinear motion (body / Love-type); high planarity with low linearity = elliptical motion (Rayleigh-type). The class shares summarise the whole event.

Rayleigh signature — V vs radial phase live

Rayleigh waves show retrograde elliptical motion: vertical and radial roughly 90° out of phase. This plots the instantaneous V−R phase (via Hilbert transform); sustained bands near ±90° are a Rayleigh signature. The score is the fraction of event energy spent near ±90°.

Frequency content (FFT) live

Hann-windowed amplitude spectrum per axis; dotted lines mark each axis's ISEE zero-crossing frequency.

Filtered re-analysis live

Splits slow low-frequency movement from sharp packet energy (zero-phase RC filters). Interpretation only — raw values remain the compliance values.

Signal vs noise live

Pre-trigger background vs the event. High SNR confirms real recorded motion — not that it's representative free-field ground.

Compare against a reference monitor live

Load a nearby or known-good monitor's CSV. If a sensor metres away doesn't reproduce the same low-frequency motion, local mount amplification becomes the leading explanation.

Wavelet scalogram — multi-resolution time–frequency live

A continuous Morlet wavelet transform. Unlike the fixed-window spectrogram, it resolves sharp high-frequency onsets and slow low-frequency surface energy at the same time — so the early body-wave packets and the late low-frequency arrival both stay crisp. Bright ridges trace each arrival's dominant frequency through time.

Response spectrum — what a structure feels live

Everything else here describes the ground; this describes the building. It sweeps a damped single-degree-of-freedom oscillator across natural frequencies and plots its peak response — so a structure whose resonance sits on a spectral peak feels far more than the raw PPV. This is the structural/regulatory lens behind most damage questions. Pseudo-velocity, 5% damping, shown from 3 Hz up (below that, integration drift dominates).

Energy build-up & spectral drift live

Left: cumulative energy (Arias-style) — what fraction of the total arrived by each instant, and how long the 5–95% energy took. A single late pulse looks very different from sustained shaking. Right: spectral centroid over time (band-limited 2–150 Hz) — quantifies the high-to-low frequency drift the cadence animation shows qualitatively.

You-are-here — synced time traces live

The same moment as the 3D dot, marked on the three component traces. Watch how the spatial shape above corresponds to the wiggles here.

Mechanism weighting live

Relative evidence for each candidate mechanism, normalised to 100%. Driven by axis frequencies, energy balance, SNR and your Step 1 crest / nearby-monitor answers. These are weighted indicators to structure the argument — not a hard classifier.

Radial / transverse rotation live

Rotate the two horizontal axes into a true radial (toward blast) / transverse frame. If you entered a blast bearing in Step 1 it's used as the starting angle. Radial should carry the body-wave energy; strong transverse points to shear / Love / local rocking.

Hole-timing arrival overlay live

Overlay detonation/arrival markers on the trace to test whether recorded packets line up with the firing sequence. Paste time_seconds, label per line (trigger-relative), or auto-place markers at the detected packet peaks.

Drill-pattern vibration heat map live

Upload the hole file (HoleID, X, Y, Z, Detonation_ms, Charge_kg), enter the monitor X/Y/Z, and the tool computes each hole's 3D distance, scaled distance, predicted contribution (your Step 1 K/B) and predicted P-arrival. Shows which holes most likely drove the reading.

Local resonance / slab-rocking simulator live

Pass the recorded motion through a single-degree-of-freedom oscillator (a loose slab, boulder or bedding layer). If a low natural frequency reproduces your low-frequency peak with amplification, local rocking becomes a credible explanation for the exceedance.

Source & physics live

Holes, monitor and VoD come from the loaded pattern / Paradigm file (Step 8). Site-law K, B and velocities come from Step 1. Override anything here.

Synthesis result live

Synthetic vs recorded live

Synthetic radial/transverse/vertical overlaid on the recorded traces (rotated to radial using the blast bearing). Shapes and peak timing matter more than exact amplitude.

Synthetic spectrum live

FFT of the synthetic vector trace, with the design firing-cadence frequency marked. Compare against the measured spectrum in Step 6.

Charge, primer & wave types live

What's happening

Arrivals at the monitor live

Idealised P / S / surface arrivals at their true relative travel times. The amber line tracks the animation.

③ Timing & wave stacking — real blast sequence live

The actual blast, to scale in metres. Each hole fires on its real .blst detonation time and sends a wavefront out at the rock velocity; the wavefront reaches the monitor at det_ms + distance ÷ velocity. The trace below sums every hole's arriving wavelet (charge-scaled, distance-attenuated) on a real millisecond axis — this is the predicted stacked waveform and peak at the monitor, the same engine as Forward Synth (Step 9).

The physics, briefly learn

Load the records live

Waveform CSVs from the same blast, ideally co-located within a few metres. Instantel Micromate / MiniMate exports; time column auto-computed if absent.

Nominate your reference yourself. The tool can pick one blind — that is the right mode when you are validating whether the indicators rank correctly without being told the answer. But in an investigation you know which mount is the good one, and that knowledge should not be thrown away. A reference chosen on vertical-axis agreement is chosen on consensus, and consensus is wrong whenever the majority of mounts share the same fault.

Verdict — which record is best coupled live

Two-stage test. Stage 1 mutual coherence: co-located sensors that are properly coupled must agree with each other in the main blast band; a badly-mounted sensor adds motion of its own and agrees with nobody. Stage 2 displacement residual: integrate velocity — real ground returns to zero displacement, a sensor that slid or tilted does not.

What each mount cost you live

Two different things, side by side. Reads high by is how much the peak reading was inflated — the number that decides a pass or a breach. Disagrees by is the largest instantaneous departure from the trusted record, as a share of its peak. A mount can score low on the first and high on the second: that is a rattle, buzzing constantly without ever building into a big number. The reverse — high over-read, lower disagreement — is a swing, quietly amplifying everything.

Sensor orientation — solved from the data live

Rotating a sensor about vertical swaps energy between its Long and Tran channels but leaves Vertical untouched. So the tool picks a reference using the vertical axis alone, then solves each record's horizontal orientation in closed form and rotates everything into one frame before anything is compared. Without this step a sensor installed facing a different way looks exactly like a badly coupled one.

Coupling fingerprint live

Two independent measurements, plotted against each other. Across: how much extra slow motion the sensor carries versus the reference — a swaying mount adds low-frequency energy the ground never had. Up: how badly its agreement collapses at high frequency — a loose mount follows the slow motion it can't avoid, then decorrelates where the buzzing lives. Where a record lands names its fault.

Agreement matrix live

Magnitude-squared coherence between every pair, averaged over the three axes across 10–100 Hz (the main blast band). Bright = the two sensors saw the same ground motion. A record that is dark against everything is the odd one out — and the mount is the reason.

Indicator table live

Every measurement behind the verdict. Hover the header notes below for what each one means physically.

See the difference live

What the evidence means learn

What this costs you live

Can you spot it in a single waveform? tested

A fair question, and worth testing rather than assuming. Using two field trials where the mounting of every sensor was known, these waveform-shape measures were checked to see whether any of them separates good coupling from bad within a single record.

How coupling goes wrong learn

① The shake test — same ground, three answers live

One ground motion drives all three mounts at once. Watch the sensor pucks: the buried block moves with the ground; the free-standing sensor swings on its soft footing; the loose fixing buzzes. The trace under each is what that monitor would report.

② Every mount is a spring — and every spring has a resonance live

This is the physics behind everything else. A mount passes ground motion faithfully below its resonant frequency, amplifies near it, and blocks motion above it. If that resonance lands inside the blast frequency band, your monitor over-reads. Drag the stiffness and watch the curve move across the band.

③ Return to zero — the movement test live

Add up a velocity trace over time and you get displacement — how far the ground actually moved. Real ground shakes and comes back to where it started, so displacement returns to zero. A sensor that slid, tipped or settled does not come back. That leftover offset is the single strongest evidence of a mount that moved.

④ Do they agree? live

Two monitors a few metres apart are hit by the same wave. So any motion one of them records that the other doesn't cannot be ground — it has to be the mount. Coherence measures that agreement frequency by frequency, and it's the strongest evidence the tool has.

⑤ The mounting methods, honestly learn

What each method actually does, and how it fails. Ground type matters as much as method — a spike is only as good as what it's driven into.

⑥ Why it matters live

Measured field trial, five monitors, one blast, all within a few metres of each other on the same ground. Only the mounting differed.

① The signature hole live

Everything here is built from the waveform of one hole. Upload a real single-hole record for a quantitative prediction, or use a modelled signature to explore the behaviour. A measured signature carries your geology, path and monitor response — a modelled one cannot.

①b One hole vs the whole blast live

The most direct question a timing design can be asked: did firing all those holes together produce more than one hole on its own? If the single-hole record and the production blast come from the same monitor, the ratio of their peak vector sums is a measured stacking gain — no model, no site law, no assumptions. Load the single hole in panel ① and the production blast in Step 1.

② Re-time the blast live

The design as loaded is the baseline. Change the sequence below and everything downstream updates. Firing direction is the big lever — sweeping away from a receiver spreads arrivals in time, sweeping toward it bunches them.

③ What it does to the reading live

④ Delay sweep — the biggest lever you have live

The same holes, the same explosive, only the inter-hole delay changed. This scans every delay and predicts the peak at your receiver for each. The result is resonant, not gradual — good and bad delays sit close together, so this is not something you can judge by eye or by rule of thumb.

④ Where the energy goes live

Predicted peak vibration on a grid around the blast, computed from the same superposition. Watch the quiet lobe swing as you change firing direction — this is what "shielding" a receiver actually means.

⑤ Detonator scatter — how certain is that number? live

Real detonators do not fire exactly on time. This runs the same prediction many times with realistic scatter, so the answer is a range rather than a single optimistic figure. It is also the clearest case for electronic initiation you can put in front of someone.

⑥ Timing sets the frequency live

Holes firing at a fixed interval reinforce at that interval's frequency and its harmonics — a comb. Move the delay and you move the energy. The value of that: you can shift vibration away from the frequency a nearby structure responds to, without reducing the explosive at all.

⑦ The wavefield, in motion live

Each hole fires on its real time and sends its wave out at the rock velocity, to scale. Watch the fronts sweep and stack. Drag to rotate.

How this works, and where to be careful learn

① Resonance from ambient noise — HVSR live

Nakamura's horizontal-to-vertical spectral ratio, standard practice in earthquake site characterisation and almost unused in blast monitoring. A site or a mount with a resonance shows a peak in the ratio of horizontal to vertical motion at that frequency. It needs no blast — only background noise. Drop in a waveform (its pre-trigger window is used) or, far better, a dedicated ambient recording of 10–20 minutes.

② Is a wave coming back? — reflection detector live

A wave that reaches a free face does not vanish. It reflects — at normal incidence with its polarity inverted and essentially all of its energy — and travels back into the rock, where it can meet the next incoming wave. If that is happening, the record contains the same arrival twice, separated by the two-way travel time to the reflector. Autocorrelation finds that repeat.

③ Will that slab ring? — predicted resonance live

A near-vertical slab of rock bounded by open bedding or joint planes, held at depth and free at the crest, behaves as a cantilever. If a monitor is bonded to such a slab it will record the slab's motion, not the ground's — and a slab that is still swaying cannot let the trace return to zero between panels. Enter the geometry you can see in the wall and compare the prediction against the HVSR peak above.

The physics, and the limits learn