Structural Health Monitoring, Data Analysis & Remote Monitoring Notes

Last reviewed 30 Sept 2026 · 7 min read

What SHM is

Structural health monitoring (SHM) is the continuous or periodic measurement of a structure's condition with permanently installed sensors, followed by analysis that converts the data into information about its safety and serviceability. It complements inspection: inspectors see surface defects at intervals; SHM sees behaviour — stress, movement, vibration — all the time, including hidden parts and during extreme events.

Objectives

Verify design and construction (stress in stay cables, deflection during erection).
Track long-term behaviour — creep, shrinkage, settlement, prestress loss, temperature response.
Detect damage or deterioration early — cracking, bearing failure, cable loss, foundation scour.
Warn when a safety threshold is reached — after a flood, earthquake, ship impact or overloading.
Support maintenance decisions — replace what is needed, when it is needed.
Measure real loads — traffic (weigh-in-motion), wind, temperature — for future design.

What is monitored

Type	Parameters and sensors
Loads and environment	Wind speed and direction (anemometers), temperature, humidity, traffic (WIM, cameras), earthquake (accelerometers)
Structural response	Strain, stress (strain gauges), deflection and displacement (LVDT, GNSS, laser), tilt and rotation, cable force (load cells or vibration method), acceleration
Local condition	Crack width, acoustic emission, corrosion (potential, resistivity), fatigue counters
Foundations and scour	Scour depth sensors, tilt of piers, pore pressure
Bearings and joints	Movement, tilt, load

From raw data to information

The chain: acquisition → validation → processing → analysis → interpretation → decision.

Data validation and cleaning

Real data are imperfect. Before analysis:

Remove spikes and outliers (a sensor glitch) by limit checks and median filters.
Detect drift and offsets by comparison with reference sensors or known zero conditions.
Fill gaps or flag them (power failure, communication loss); never invent data.
Check status channels — battery voltage, sensor health and cable resistance.
Compensate for temperature — strain and displacement have a large temperature-driven component (daily and seasonal). Regression against measured temperature separates the thermal from the load effect: $ε_{cor r ec t e d} = ε_{m e a s u r e d} - α_{T} (T - T_{0})$ or with a fitted regression.

Filtering

Filter	Use
Moving average	Smooths noise in slowly varying data
Median filter	Removes spikes without blurring the edges
Low-pass (Butterworth)	Removes noise above a cut-off; extracts static or slow response
High-pass / band-pass	Removes drift; isolates a vibration band
Digital filter order and phase	Zero-phase (forward–backward) filtering avoids phase distortion in analysis

Statistical processing

Long records are summarised by mean, standard deviation, RMS, maximum and minimum, percentiles, and by extreme value analysis for design load statistics. Trends (slope of daily mean over months) show creep, settlement or deterioration. Rainflow counting turns a stress history into cycles of stress range for fatigue damage assessment (with Miner's rule).

Frequency-domain analysis

The Fast Fourier Transform (FFT) converts a time record to a spectrum, showing which frequencies carry the energy. For a record with sampling frequency $f_{s}$ and $N$ points, the frequency resolution is

Δ f = \frac{f _{s}}{N}

and the highest usable frequency is $f_{s} /2$ . Power spectral density (PSD), spectrograms (frequency vs time) and transfer functions are also used; windowing (Hanning) reduces leakage, and averaging many segments reduces noise.

Ambient vibration data give the bridge's natural frequencies, mode shapes and damping:

Peak-picking from the spectrum (simple, good for well-separated modes).
Frequency Domain Decomposition (FDD) — separates close modes.
Stochastic Subspace Identification (SSI) and Eigensystem Realization — time-domain methods.

Worked ExampleExample — frequency resolution and damage indication

A bridge is monitored at $f_{s}$ = 100 Hz with records of $N$ = 4096 samples, so $Δ f = 100/4096 = 0.0244$ Hz.

The baseline first vertical frequency (after construction) is 2.50 Hz. A later record shows a peak at 2.44 Hz — a drop of 2.4 %.

Since $f \propto k / m$ , $\frac{Δ f}{f} \approx \frac{1}{2} \frac{Δ k}{k}$ when the mass is unchanged, so the stiffness has fallen by about 4.8 % — more than the scatter caused by temperature (which is 0.5–1 % in most concrete bridges). The engineer looks at the temperature-corrected frequency, other modes, and then inspects the bridge for cracking or a change in the bearing behaviour.

Downloaded from civilgyantests.com · Not for resale. Verify every code clause and date against the official source.