Blog

How to Choose the Right Sensor for Structural Health Monitoring

How to choose a sensor for structural health monitoring: noise density, RMS, dynamic range and modal analysis, compared across the four StructSense models.

July 12, 2026
StructSense sensor family for choosing the right sensor in structural health monitoring

Choosing the right sensor for structural health monitoring directly affects how reliable the natural frequency, period, damping ratio and modal analysis results will be. Instead of using the most expensive or the lowest-noise sensor on every structure, consider the structure’s vibration level, expected frequency range, measurement goal and environment together.

Recording the high accelerations of an earthquake and identifying a building’s natural frequency from everyday ambient vibration are not the same measurement problem. The right question is therefore not “Which sensor has the lowest noise?” but “What noise level is sufficient for this structure and this analysis goal?”

The StructSense family developed by StructHEALTH has four models that differ mainly in noise level:

  • StructSense Seismic – SH-A1100
  • StructSense Elite – SH-A1000
  • StructSense Ultra – SH-A900
  • StructSense Core – SH-A020

This article explains sensor selection through noise density, RMS vibration level, dynamic range and structural analysis goals.

Why sensor selection matters

Accelerometers on a structure do more than record earthquakes. With a suitable sensor and a long enough record, the following engineering parameters can be determined:

  • Natural frequencies
  • Structural periods
  • Dominant vibration directions
  • Mode shapes
  • Damping ratios
  • Differences in dynamic behaviour between floors
  • Frequency changes before and after an earthquake
  • Long-term changes in stiffness and performance
  • Vibrations from machinery, traffic, wind and human movement

However, these parameters cannot necessarily be identified with the same sensor on every structure. A strong earthquake produces accelerations that are easy to measure, whereas a building’s micro-vibrations in normal use can be lost inside the sensor’s own noise.

Noise density, RMS, resolution and accuracy are different things

Four terms are often confused when choosing a sensor:

  • Noise density (µg/√Hz): the level of the sensor’s own electronic noise per unit of frequency.
  • RMS noise: the total noise level over a chosen frequency band; it depends on the bandwidth.
  • Resolution: the smallest step the digital output can distinguish (for example the ADC bit count); it is not the same as noise density.
  • Accuracy: how close the measured value is to the true value; it depends on factors such as scale factor, non-linearity and temperature effects.

This article only gives a selection logic based on noise density and dynamic range.

How noise density affects sensor selection

The noise density quoted in µg/√Hz in sensor datasheets describes the sensor’s own noise at a given frequency. In simple terms, it is the electronic noise the sensor adds to the measurement.

The lower the noise density:

  • smaller structural motions can be detected;
  • natural-frequency peaks are clearer in FFT and PSD plots;
  • frequency changes can be tracked more reliably;
  • operational modal analysis results become more stable;
  • long-period, low-amplitude structures are easier to monitor.

When noise density is high, low-amplitude real vibrations can stay inside the sensor’s own noise. So when choosing a sensor, look not only at the measurement range but also at the noise density.

What the RMS value means for civil engineers

RMS expresses the total vibration level over a given time and frequency range as a single number. In other words, it answers: “How much did the structure vibrate in total during the period examined?”

A RMS value measured in the field can include all of the following together:

  • Sensor noise
  • The real motion of the structure
  • Traffic and machinery vibration
  • Wind effects
  • Mounting and cable movement

Therefore a RMS value measured in the field should not be read directly as the sensor’s internal noise.

An approximate band RMS value can be calculated from the datasheet noise density. If the noise is assumed constant (white) between 1 and 10 Hz:

a_RMS ≈ n × √(10 − 1) = 3 n

Here n is the sensor’s noise density in µg/√Hz.

Approximate 1–10 Hz noise levels of the StructSense models

  • StructSense Seismic (SH-A1100): 0.02 µg/√Hz → about 0.06 µg RMS
  • StructSense Elite (SH-A1000): 0.2 µg/√Hz → about 0.60 µg RMS
  • StructSense Ultra (SH-A900): 0.9 µg/√Hz → about 2.70 µg RMS
  • StructSense Core (SH-A020): 25 µg/√Hz → about 75 µg RMS

These are not field measurements. They are approximate equivalents obtained by converting the datasheet noise density to the 1–10 Hz band under a white-noise assumption. Real measured levels can be higher because of the structure, ground, mounting, wind, traffic and other environmental effects.

What is dynamic range?

Dynamic range is the ratio, expressed in decibels (dB), between the smallest motion a sensor can detect and the largest motion it can measure without saturating. The dB here is not a sound level.

Datasheet dynamic range values for the StructSense family:

  • StructSense Seismic: 170 dB
  • StructSense Elite: 148 dB
  • StructSense Ultra: 120 dB
  • StructSense Core: 80 dB

A high dynamic range makes it easier for the same sensor to measure both everyday micro-vibrations and the larger motions of an earthquake. But dynamic range should not be the only criterion: noise density, sampling rate, time synchronisation, mounting quality and measurement range should be considered together. For the conditions under which these values are specified, refer to the product datasheet.

Sensor selection for modal analysis

To determine a structure’s period, its natural frequency must first be identified: T = 1 / f. For example, if the dominant frequency is 0.5 Hz, the period is T = 2 seconds.

For the natural frequency to be clearly visible, the structure’s vibration level must be sufficiently above the sensor noise.

StructSense models and where they fit

StructSense Seismic

StructSense Seismic has the lowest noise density in the product family. It is a candidate for:

  • High-rise buildings
  • Long-period structures
  • Hospitals with seismic isolators
  • Long-span bridges
  • Dams
  • Historic structures
  • Very low-amplitude micro-vibration measurements

It offers the lowest noise level in the family for long-term tracking of natural frequency, period and small frequency changes.

Model: SH-A1100 · Noise density: 0.02 µg/√Hz · Dynamic range: 170 dB

StructSense Elite

StructSense Elite is a balanced option for general structural health monitoring projects that need low noise.

  • Reinforced concrete and steel buildings
  • Bridges and viaducts
  • Hospitals
  • Public buildings
  • Industrial facilities
  • Operational modal analysis applications

Its noise level is not as low as Seismic, but for many building and bridge applications it can be sufficient for identifying natural frequency and period; the vibration level of the specific structure should be confirmed with a preliminary measurement.

Model: SH-A1000 · Noise density: 0.2 µg/√Hz · Approx. 1–10 Hz noise: 0.60 µg RMS · Dynamic range: 148 dB

There is not yet a field data set for StructSense Elite, so this assessment rests on datasheet values.

StructSense Ultra

StructSense Ultra is intended for monitoring typical building vibrations together with earthquake and service vibrations.

  • Standard reinforced concrete buildings
  • Schools
  • Hospitals
  • Factories
  • Medium-span bridges
  • Retaining structures
  • General structural health monitoring projects

If the ambient vibration is high enough, natural frequencies and periods can be identified. If the structure’s vibration level is very low, Seismic or Elite will be more advantageous.

Model: SH-A900 · Noise density: 0.9 µg/√Hz · Dynamic range: 120 dB

StructSense Core

StructSense Core is intended for strong vibration and event monitoring rather than very low-level micro-vibration.

  • Earthquake records
  • Threshold exceedance
  • Construction activity
  • Strong machinery vibration
  • Impact and shock events
  • Traffic-induced vibration
  • Field alarm systems

Natural frequency can be identified if the structure has sufficient vibration or forced vibration is applied. For very quiet structures, however, it is not the first choice for detailed modal analysis from ambient vibration alone.

Model: SH-A020 · Noise density: 25 µg/√Hz · Dynamic range: 80 dB

Starting point by project type

The pairings below are a general starting point; each structure needs a preliminary measurement to confirm.

  • Hospital with seismic isolators: StructSense Seismic
  • High-rise building: StructSense Seismic or Elite
  • Long-span bridge: StructSense Seismic or Elite
  • Historic structure: StructSense Seismic
  • Standard reinforced concrete building: StructSense Elite or Ultra
  • School, hospital and public building: StructSense Elite or Ultra
  • Factory and machinery vibration: StructSense Ultra or Core
  • Earthquake event recording: StructSense Core, Ultra, Elite or Seismic
  • Threshold alarm and impact monitoring: StructSense Core
  • Operational modal analysis of low-amplitude structures: StructSense Seismic or Elite

What are NHNM and NLNM?

Structural and seismic measurement reports sometimes compare spectra against the NHNM and NLNM reference curves (the New High Noise Model and New Low Noise Model of Peterson, 1993).

  • NLNM is the envelope of the lowest background noise observed at seismic stations worldwide.
  • NHNM is the envelope of the highest background noise observed at seismic stations.

These curves can be used to judge how quiet a sensor, a site or a measurement environment is. One distinction matters:

NHNM and NLNM are not structural safety, damage or code limits.

A building’s PSD curve lying above the NHNM does not mean the building is damaged. In an urban structure, high spectral levels can come from traffic, lifts, air-conditioning and fans, pumps, human movement and nearby machinery.

Is a low-noise sensor always necessary?

No. High-grade low-noise sensors are not technically or economically necessary for every project.

In a factory building where machinery makes the vibration level high, StructSense Core or Ultra may be sufficient. In a long-period tall building or a historic structure, where the natural motions can be very small, StructSense Seismic should be preferred.

Questions to answer when selecting a sensor:

  1. What is the smallest vibration level you want to measure?
  2. What is the structure’s estimated natural-frequency range?
  3. Will only earthquakes be monitored, or continuous ambient vibration as well?
  4. Besides natural frequency, will damping and mode shapes be extracted?
  5. How many sensors will be used?
  6. Will the sensors be time-synchronised?
  7. Will measurement be continuous or event-based?
  8. How will the sensor be fixed to the concrete element?

Beyond the sensor: what else matters for modal analysis?

A low-noise sensor is important but not sufficient on its own. For reliable modal analysis:

  • Sensors must be rigidly fixed to the load-bearing element.
  • Cable movement must be prevented.
  • Sensor directions must match the structure’s axes.
  • Multiple sensors must work on a common time base.
  • Sampling rate and filters must preserve the expected frequencies.
  • Records must be long enough.
  • Sensor locations must be planned to represent the mode shapes.
  • Methods such as FDD, EFDD, ERA or SSI should be evaluated together.

Natural frequency is usually the easiest and most stable of the modal parameters to identify. Damping ratio is more sensitive to record length, ambient vibration and the algorithm used.

Conclusion: do not use the same sensor for every structure

If the aim is to track a structure’s period and natural-frequency changes from very low-amplitude ambient vibration, StructSense Seismic or StructSense Elite should be preferred.

For standard buildings and bridges, StructSense Ultra is a strong option for general structural health monitoring and modal analysis.

If the goal is to monitor earthquakes, impacts, shocks, threshold exceedance and strong service vibration, StructSense Core is an economical and suitable option.

The basic selection principle is:

The smallest structural motion you want to measure must be sufficiently above the sensor’s own noise level.

This article is a general guide to sensor selection. Sensor measurements alone do not settle the condition of a structure; judgements about structural safety require adequate engineering analysis in addition to the measurements.

Frequently asked questions

Which sensor measures a structure’s period?

If the structure’s vibration level is sufficient, the period can be identified with any StructSense model. For very low-amplitude ambient vibration, StructSense Seismic and Elite are likely to give more reliable results.

Why is a sensor with low noise density preferred?

Because the sensor’s own noise floor is lower, small structural movements can be distinguished more clearly, which improves natural-frequency and modal analysis quality. It is not needed in every project.

Are RMS and noise density the same?

No. Noise density is a spectral value in µg/√Hz. RMS is the total vibration level over a chosen frequency band.

Are NHNM and NLNM damage limits?

No. They are general seismic background-noise reference curves and are not used as structural safety or damage limits.

Is the lowest-noise sensor needed for earthquake measurement?

For strong earthquake motion, a low noise level is usually not the primary criterion. Measurement range, saturation behaviour, time synchronisation and data continuity matter more.

Contact the StructHEALTH engineering team to determine the right StructSense model for your structure, plan sensor placement and design your structural health monitoring infrastructure.

Reference: Peterson, J. (1993). Observations and Modeling of Seismic Background Noise. U.S. Geological Survey Open-File Report 93-322.