Standard methods, engineered for continuous use
SKAB does not introduce a new theory of dynamic identification. It applies established Operational Modal Analysis techniques automatically, continuously and at a sustainable cost. The innovation lies in the chain — acquisition, synchronisation, noise rejection, processing and comparison over time — not in the mathematics.
Four layers, from sensor to interface
SKAB stations
Continuous field acquisition, event detection, local buffering on non-volatile storage and managed upload. Firmware updatable remotely.
IoT layer
Device management, telemetry, shared attributes, routing rules and over-the-air firmware distribution.
Ingest and processing
Integrity verification, decoding of the proprietary binary format, conversion to columnar storage, modal analysis pipeline and structured archiving.
Interface
Web application with map, building catalogue, time series, analysis reports and alert management. Real-time updates.
The station
Each station integrates complementary sensing: accelerometers designed for the very low noise levels of ambient vibration, a wideband channel for impulsive events, and a seismometer providing an immediate severity estimate.
| Component | Role | Notes |
|---|---|---|
| Low-noise MEMS accelerometer | Primary modal analysis channel | Two horizontal axes, 16-bit, ≈15 µg/√Hz noise density, selectable full scale. |
| Wideband accelerometer | Impulsive and high-frequency events | Three axes, extended bandwidth. Complementary to the OMA channel. |
| Dedicated seismometer | Spectral Intensity (SI) and PGA | Housner Spectral Intensity is the integral of the pseudo-velocity response spectrum over the period band relevant to buildings, expressed in kine (cm/s). It correlates with damage potential better than peak acceleration alone. Computed on board, available once the event has ended. |
| Environmental sensing | Normalisation of comparisons | Temperature and humidity affect natural frequencies; they are logged so comparison happens under homogeneous conditions. |
| Dual-core MCU | Separate acquisition and connectivity | One core dedicated to real-time acquisition, the other to networking and storage. |
| 4G / Wi-Fi connectivity | Data transmission and remote management | Deployment is possible even where no local network is available. |
How many stations are needed. There is no single answer, and one should be wary of anyone who gives one. On a regular, box-like building, two or three stations are enough to describe global behaviour.
The torsional mode requires a specific condition. Identifying the rotation of the floor diaphragm about the vertical axis requires at least three stations placed on the same level and spaced apart in plan. A configuration with one station at the base and one at the top measures amplification along the height well, but cannot resolve the torsional mode: in that case f₁T remains undetermined. It is a geometric limitation, not an instrumental one, and must be considered when designing the layout.
On irregular geometries — L-shaped or U-shaped plans, adjoining blocks, structural joints — that configuration is not sufficient: different parts of the structure may respond independently. For these cases we are developing a dedicated approach based on recognising the modal pattern of the structure, which drives the number and position of measurement points from the actual geometry. Placement remains defined case by case, with reference to ISO 4866.
From raw sample to natural frequency
STA/LTA detection
Ratio between short- and long-term average of signal energy, computed on the horizontal component. An event is confirmed only if the ratio exceeds the threshold for a minimum duration and the dynamic peak is significant.
Noise classification
Known recurring sources within the building are recognised and separated from genuine transients, reducing analysis load and false positives.
Multi-station alignment
Acquisitions from different stations are brought onto a common time base. The required precision is high: even a modest misalignment degrades the identification of torsional modes. SKAB combines proprietary on-station synchronisation with residual correction during processing.
Spectral estimation and identification
Welch PSD, Frequency Domain Decomposition for mode detection, EFDD for damping estimation, SSI-COV in the time domain with stabilisation diagram.
Quality control
Magnitude-squared coherence between sensor pairs per frequency: values close to unity indicate points vibrating coherently, therefore a genuine structural mode rather than uncorrelated noise.
Baseline comparison and alerting
Computation of the relative shift against the reference and notification when agreed thresholds are exceeded.
| Parameter | Value | Rationale |
|---|---|---|
| Sampling rate | 200 Hz | Nyquist at 100 Hz, wide margin over the band of interest. |
| Analysis band | 0.1 – 10 Hz | Typical range of building natural modes. |
| Analysis windows | 30 – 60 s | Trade-off between frequency resolution and stationarity. |
| Session duration | 1 – 120 min | Remotely configurable per station. |
Standards and literature
SKAB does not define its own criteria: it sits within an existing normative and scientific framework.
Standards
- EN 1998-1 (Eurocode 8, Part 1) — vibration modes and modal response spectrum analysis.
- EN 1998-3 (Eurocode 8, Part 3) — assessment of existing structures and damage state identification from experimental measurements.
- ISO 4866:2010 — measurement of vibration on fixed structures: terminology, procedures, sensor placement.
- ISO 13822:2010 — assessment of existing structures using monitoring-based methods.
- NTC 2018 and Circolare 7/2019 (Italian building code) — modal analysis, participating mass, expected natural frequencies by structural typology.
SHM and OMA literature
- Rytter (1993) — the four levels of damage identification: presence, location, severity, remaining life. SKAB currently operates at the first two.
- SAMCO Guidelines (2006) — European guidelines for Structural Health Monitoring.
- Brincker & Ventura (2015), Introduction to Operational Modal Analysis — primary reference for FDD, EFDD and SSI.
- Peeters & De Roeck (1999) — Stochastic Subspace Identification.
- Housner (1952) — definition of Spectral Intensity; FEMA 306/307/308 and ASCE 41-17 for damage classification and use of vibration measurements.
Methodological honesty. The most widely cited reference ranges derive predominantly from reinforced concrete buildings in Japan and the United States. The Italian building stock is largely historic masonry and pre-seismic-code construction: direct transferability of those thresholds is not demonstrated. Operating thresholds on a real building are therefore always defined together with the appointed structural engineer, and never applied automatically by the system.
Want to see the platform on real data?
We can arrange a technical demo using analysis reports produced by stations currently in service.