Emanuele Burgognoni

Monday, May 27, 2024 · 0 min read

Structural and Geotechnical Engineering Department, Sapienza University of Rome

Comparing MEMS and IEPE accelerometers for structural vibration behavior testing

Elastoplastic behavior studies through force-displacement hysteresis cycles are essential to understanding how structures behave when exposed to cyclic loads, e.g., those generated by earthquakes. The Sapienza University in Rome studied the plastic behavior of a scale model comparing the performance of IEPE and MEMS accelerometers. The aim was to identify the type of accelerometer providing the signals most consistent with those of a laser displacement sensor.

The Structural and Geotechnical Engineering Department of Sapienza University in Rome, Italy, studied plastic behavior by analyzing the cyclic loads applied to a scale structure placed on a shaking table.

This analysis through force-displacement hysteresis loops allows us to examine the structure's energy dissipation capacity and evaluate its plastic behavior in detail.

Figure 1. The laboratory with machinery for material stress testing and the structure scale model.

The laboratory

The Materials and Structures Laboratory of the Structural and Geotechnical Engineering Department of the Sapienza University of Rome came to being in the early sixties. Nowadays, the activities it’s hosting focus on three different areas;

  • the field of teaching,

  • the scientific field, with experiments on materials and models,

  • the services to public or private bodies by providing analysis of the mechanical characteristics and any related official certification of materials and prototypes related to civil and industrial constructions.

The Laboratory is an official research and testing laboratory for the execution of tests on construction materials and carries out:

  • Tests on steels, concrete, mortars, bricks, profiles, and metal mesh.

  • Tests on structural elements and models, even at full scale.

  • Calibration and verification of the reading scales of the measuring instruments.

The fundamental testing machines present within the laboratory are:

  • Presses, maximum capacity 1500kN and 5000kN.

  • Traction machines, maximum load of 500kN and 1000kN.

  • Electronically controlled MTS universal machine for static and dynamic tests, maximum load 500kN (static load), piston stroke ± 100 mm, maximum frequency up to a few tens of Hz.

  • A double-acting Schenck jack (traction and compression) with electronic control, maximum load 250kN, piston stroke 400 mm, maximum frequency in the order of a hundred Hz.

  • A Pulsator with a pair of 20t jacks and a 100t jack. Also, for cyclic tests at a maximum frequency of 10 Hz.

  • Electrodynamic stirrer, complete with power supply/amplifier, peak sinusoidal force 100 N, maximum acceleration 588 m/s2, maximum displacement 10 mm.

  • MOOG L081-324-11 shaking table.

Figure 2. The model on the shaking table.

The vibration shaker

The vibration shaker or shaking table is used in engineering to subject structures to controlled dynamic loads similar to those generated by earthquakes and other stresses. This machine is fundamental for laboratory experimentation. It allows engineers to evaluate structural behavior under simulated seismic or dynamic loading conditions over time.

The main components of a vibration table are:

  1. Vibration platform: The shaker has a flat, solid surface, subject to controlled vibrations. It can vary in size according to needs, allowing the positioning of structures of different scales.

  2. Vibration system: The shaker is equipped with a drive system, for example, electric or hydraulic motors, that generates vibrations. This system applies dynamic loads in a controlled manner to the tested structure.

  3. Control system: An integrated control system regulates the frequency, amplitude, and energy of vibrations applied to the table. Precise control is essential to reproduce the desired load conditions during testing.

  4. Measurement instrumentation: The shaker is equipped with sensors and instrumentation to measure the dynamic response of the tested structure, for example, accelerometers, strain gauges, and force transducers.

The Moog L081-324-11 shaking table used by the laboratory has the following characteristics:

  • dimensions of the vibrating table are 1.5 x 1.5 m

  • maximum stroke of ± 200 mm

  • maximum load capacity of 2 tons

  • 4-story frame with a floor of 0.60 m x 0.60 m and an inter-story height of 0.60 m

  • total height of 2.40 m.

This frame is connected to a steel plate of dimensions 0.68 m x 0.76 m, connected to the table via an elastomeric isolator and four-ball load transfer elements.

Shaker testing applications

  • Seismic Behavior Analysis: The shake table simulates seismic loads, allowing engineers to study the response of buildings and other structures to seismic forces.

  • Structural Model Validation Testing: Structures can be subjected to vibration to validate and calibrate numerical models and simulations to ensure an accurate match between experimental data and model predictions.

  • Evaluation of Structural Resistance: Tests to evaluate the seismic resistance of structures, identifying possible weak points and evaluating the ability to absorb energy without suffering irreversible damage.

  • Testing the Efficiency of Anti-seismic Devices: The shaking table tests the effectiveness of anti-seismic devices, such as seismic isolators and dampers, designed to reduce the transmission of seismic forces to structures.

  • Study of Plastic Behavior: Tests to study the plastic behavior of structures in the presence of cyclic loads similar to those generated by earthquakes.

  • Characterization of Materials: The shaking table can characterize materials, evaluating their response to vibrations and cyclic stresses.

  • Structural Design and Optimization: The results obtained from shaking table tests provide crucial data for designing and optimizing structures, improving their safety and seismic resilience.

Overall, using a shake table allows engineers to perform various controlled tests. They can replicate realistic dynamic conditions and provide data critical for seismic design and construction safety.

Figure 3. Force-displacement hysteresis cycles obtained by the dynamic load tests.

Plastic behavior: single-frequency cyclic loads

How do you study the plastic behaviors of structures subjected to earthquakes? With the vibration shaker, it is possible to carry out cyclic load tests to identify the elastic-plastic behavior of the structures. 

Performing integration, you can obtain the displacement values and the force-displacement ratios through accelerometers. The analysis of force-displacement hysteresis cycles is the core of the analysis that leads to understanding the plastic mechanisms of the structure.

The shape and width of the hysteresis curves reflect the structure's ability to dissipate energy during dynamic events. This information is critical for designing energy dissipation systems, such as viscous or pendulum dampers, which can improve the structure's earthquake resilience.

Analysis of crucial points of hysteresis curves, such as the yield and collapse point, offers a detailed understanding of structural performance. This data is the basis for developing plastic behavior models that improve the accuracy of predictive analysis of cyclic loading situations.

Figure 4. The correlation between the force-displacement hysteresis cycles and the moving structure

The shaking table plays a crucial role in structural engineering. It is a versatile means for extrapolating broad-spectrum information. It is possible to improve the response capacity of structures in dynamic situations, such as earthquakes. Through these cycles, engineers can observe how the structure reacts under cyclic loads, providing a detailed analysis of the structure's limitations and capabilities.

Measurement setup

The models proposed in Figures 5 and 6 represent a four-story frame with an insulated base. We limited its path through limit switches (bumpers). Our testing involved subjecting the model to time-variable dynamic forces through a vibrating table. The use of bumpers was fundamental to avoid alterations to the measurements due to the collision of the base with the platform.

We installed two accelerometer types on each of the floors of the scale model:

  1. PCB 393A03: IEPE type accelerometer (Integrated Electronics Piezo-Electric)

  2. Dewesoft IOLITEi-3xMEMS-ACC-8g: MEMS type accelerometer (Micro Electro Mechanical Systems).

We measured the displacement of the base through an optoNCDT 1420-500 laser transducer to ensure high measurement precision.

The data acquisition was done by:

We forced the structure mechanically through single-frequency waves to obtain detailed information on the structure's ability to absorb dynamic loads. The nature of this test also brings advantages from the perspective of analyzing the quality of the response of the transducers in the domain of the structure.

Figure 5. Schematic from the upper view of the vibrating table and the acquisition system
Figure 6. Schematic from the lateral view of the vibrating table and the acquisition system.

Analysis objectives

This study had two well-defined objectives: 

  1. The first aimed at identifying the most suitable accelerometer instrumentation to provide a signal in line with that obtained from the laser displacement sensor. The fidelity of this signal is evaluated as a function of the temporal response of the transducers, generating single-frequency mechanical waves (harmonics).

  2. The final aim instead concerns the study of the elastic-plastic behavior of the structural model as the structural elements characterizing the model itself vary.

Figure 7. The two types of accelerometers, PCB 393A03 and Dewesoft IOLITEi-3xMEMS-ACC-8g.

Accelerometer technologies

A couple of decades have passed since geotechnics led the entry of geophones (velocimeters) into the world of dynamic analysis. Now, we have overcome the problems of signal phase shift imposed by the velocity meter resonance. Accelerometers, which are less sensitive but extremely more reliable in phase response thanks to their higher resonance frequency, have been introduced. 

This project allowed us to understand the importance of analyzing the signal’s phase response, not only depending on the sensor’s resonant nature but also on applying filters to the signal.

The fidelity of a signal generated by an acquisition system concerns the module and the phase. The module indicates how much the perceived wave intensity it reconstructs. The phase is the alter ego of the sensor's temporal response to the event. Both applied sensors generate a signal with very high module and phase fidelity. However, a non-negligible factor occurs in the data processing line generated by one of the two technologies.

The electrical signal generated by piezoelectric technology is in itself very attenuated. It is represented by the movement of the surface electrical charges of the crystal as the mechanical compression forces experienced vary. To use this signal, we must amplify it by increasing the number of charges subject to the displacement imposed by the crystal’s mechanical compressions.

We superimposed the power supply on the signal line to ‘dope’ the crystal with an excess of charges and obtain a broader signal. On the other hand, we got an imposed offset voltage by this charge excitation exactly. You must apply an HP (High Pass) filter to eliminate this offset voltage, generally at 0.5Hz. 

This filter has negligible consequences if the frequency band in concern is less than 1Hz. However, if the analyzed structures will be studied continuously across the entire spectrum up to the DC, it is an entirely different matter.

Figure 8. Bode plot of a 6th-order HP filter frequency response at 0.5 Hz.

The signal alteration follows different trends for the module and the phase as the order of the filter varies. Theory imposes a module attenuation of -20dB/Decade for each degree of the filter, while the phase undergoes a variation of -90° for each additional filter order. This last detail is not negligible since you, through the order of six, obtain a phase shift of 540°.

It is now possible to understand how the signal generated by the IEPE sensor and then cleaned by the HP filter undergoes non-negligible influences. On the other hand, MEMS technology does not require a filter for signal processing and is without these problems.

Sensor Characteristics

Comparison of IEPE and MEMS sensors characteristics.
ModelPCB 393A03Dewesoft IOLITEi-3xMEMS-ACC-8g
TypeIEPE - Integrated Electronics Piezo-ElectricMEMS - Micro Electro Mechanical Systems
Resonance frequency> 55 kHz2.4 kHz
Bandwidth7 kHz1.5 kHz
Applied filtersHP (~0.5Hz)None

Phase shift - the key role of MEMS

The effect of the attenuation and phase shift of the signal generated by the IEPE sensor is visible in the graphs below – see Figure 9. Thanks to the Bode diagram of the HP filter, we have seen how the effect of attenuation and phase varies as the frequency taken into consideration varies – see Figure 8. 

We carried out tests by forcing the structural model with single-frequency mechanical waves. This method gave us a detailed view of the attenuation and phase shift effects of the IEPE sensor’s HP filter: they were strong at 0.25 Hz and then became negligible around 1 Hz.

Figures 9 to 13 show the signal attenuation and phase shift generated by the three types of sensors. The signals shown are as follows: 

  • Purple - IEPE accelerometer

  • Orange - MEMS accelerometer

  • Magenta - Laser displacement sensor

Figure 9. The phase shift of 67.68° with the forcing imposed by the vibrating table at 0.25 Hz.
Figure 10. The phase shift of 39.78° with the forcing imposed by the vibrating table at 0.5 Hz.
Figure 11. The phase shift of 21.06° with the forcing imposed by the vibrating table at 0.75 Hz.
Figure 12. The phase shift of 20.88° with the forcing imposed by the vibrating table at 1 Hz.
Figure 13. The phase shift of 16.2° with the forcing imposed by the vibrating table at 1.25 Hz.
Figure 14. Software view during the dynamic load test.

Results - behavior at low frequencies

In an elastic-plastic test, you can divide deformations in the structure into two main phases: the elastic and the plastic phase. 

In the elastic phase, the deformation is reversible and linear concerning the load application, following Hooke's law. However, once it exceeds the material's yield point, a transition to the plastic phase occurs, and deformations become permanent and non-reversible.

During force-displacement cycles under these conditions, you often observe a change in slope and shape. The material or structure exhibits hysteretic behavior, i.e., the relationship between the applied stress and the resulting deformation is not unique. It means that the structure response depends not only on the current value of the applied stress but also on its loading and deformation history.

This phenomenon can have significant implications in the design and analysis of structures. It affects their ability to withstand repeated loads over time, not suffering permanent damage or structural failure. Understanding and characterizing the hysteretic behavior of materials and structures is critical to ensuring the safety and reliability of engineering applications.

This test highlighted an alteration of the hysteresis cycle. The change did not derive from the structure but from the different signal processing of the two sensor types. The IEPE sensor’s signal was attenuated and out of phase due to the high-pass filter in the data processing chain. On the other hand, the MEMS sensor generated a signal in phase with the laser displacement sensor, showing no variations in module and phase.

Figure 15. The scale structural model with the MEMS and IEPE sensors.


When considering MEMS and IEPE accelerometers for vibration acquisition to test structures’ elastoplastic behavior, you must consider the impact of the IEPE accelerometer’s high pass filter. It is an element that could lead to the loss of information at low frequencies.

In this application, MEMS sensors offer high precision and remarkable resolution. They allow the detection of even small changes in acceleration at very low frequencies. Despite being equally precise, IEPE accelerometers may offer lower precision in the vicinity of the continuous component of the frequency spectrum due to the effects of the high pass filter. 

IEPE accelerometers can alter the hysteresis loop due to the high-pass filter effect. They may not be ideal for tests that require maximum precision in characterizing the elastoplastic behavior of structures - tests where accurately capturing variations at low frequencies is crucial.

In conclusion, both types of sensors have their advantages and limitations. However, you must pay attention to the impact of the EPE accelerometer’s high-pass filter when accuracy at low frequencies is critical for your testing. 

You may use MEMS accelerometers or adopt compensation strategies to mitigate the effect of the IEPE accelerometer’s high-pass filter. Careful consideration of such factors is crucial for selecting accelerometers based on the specific requirements of the test.

For the material provided, many thanks to

Additional external articles on hysteresis loops: