Chapter 2

MEMS Accelerometers: Testing and Practical

Approach for Smart Sensing and Machinery
Diagnostics

A. Albarbar and S.H. Teay

Abstract Micro-Electro Mechanical Systems (MEMS)-based sensing elements are

gaining wider acceptance and adoption for static and dynamic (mobile) applications.
Recent increase in demands for reliable wireless sensing nodes has necessitated
seeking alternatives to expensive conventional accelerometers to perform multi-
control and monitoring tasks. Owed to their size and cost, MEMS accelerometers is
one of the alternative options.
This chapter provides insight into the fundamental design, working princi-
ples and practical guidance to MEMS accelerometers. Details of experimental
set-ups, signal conditioning and data processing are also provided to construct
integrated performance assessment system. Performance assessments are carried out
using sinusoidal excitations, impulsive (hummer testing) and random excitations.
Subsequently, calculations and comments on frequency response functions, signal-
to-noise ratios and phase distortions are outlined. Finally, guidelines to practical
adoption of MEMS accelerometers such as packaging, establishing smart vibration
sensing nodes and extraction of condition-related information are given.

Keywords MEMS accelerometers design and testing • Performance assessment

of MEMS accelerometers • Vibration measurement using MEMS sensors • Smart
sensor design and implementation • Machinery-condition monitoring

1 Introduction

Vibration analysis is one of the most usable methods in machinery-condition

monitoring. It plays a significant role in the dynamic qualification of newly designed
structural components, prediction of faults and structural aging-related problems,
and several other structural dynamics studies and diagnosis [1–3]. One reason for

A. Albarbar () • S.H. Teay

School of Engineering, Manchester Metropolitan University, All Saints Building,
All Saints, Manchester M15 6BH, UK
e-mail: a.albarbar@mmu.ac.uk

© Springer International Publishing Switzerland 2017 19

D. Zhang, B. Wei (eds.), Advanced Mechatronics and MEMS Devices II,
Microsystems and Nanosystems, DOI 10.1007/978-3-319-32180-6_2
20 A. Albarbar and S.H. Teay

its wide use is its capacity to monitor vibrating machines without interrupting
normal operations. In addition, the vibrating mechanisms of most machineries and
structures are fundamentally well known, giving rise to the possibility of detecting
many faults in accordance with the characteristics of the vibration responses.
Furthermore, the progress of vibration signal processing techniques, computing
capabilities, reliable performance of vibration instrumentation such as wide-band
transducers and portable analysers has caused this technique to be extensively used
around the world.
As an example, in machinery-condition monitoring, accelerometers are often
used to measure their vibration (acceleration) signals, which may lead to the detec-
tion of any deviation from normal signatures. Vibration waveforms are interpreted
and processed in a variety of ways such as peak values and variance of the signal in
the time domain, and power spectral analysis in the frequency domain [4].
Vibration has traditionally been sensed using piezoelectric accelerometers. These
are accurate and reliable, but have some inherent problems: they are difficult
to mass-produce, and they have high source impedance, which means that their
signals must be very carefully amplified. Moreover, the use of the traditional
piezoelectric accelerometers for simultaneous multiple data collection points was
considered to be impractical; this is mainly because of their cost as well as the
costs of the associated electronic signal conditioning units. Other types such as
piezoresistive accelerometers have limited resolution and can be used only for low
and medium frequencies. Low resolution is also a disadvantage of electrodynamic
accelerometers. The capacitive type accelerometers have low resolution and are
fragile. Commercially, several types of accelerometers manufactured by many well-
known manufacturers are available all over the world. In vibration-measurement
filed tests, accelerometers of mainly Integrated Circuit Piezoelectric (ICP) type, are
found not to be able to measure the non-stationary impulsive responses of structures
accurately all the time. However, confidence in measuring impulsive signals using
the charge-type piezoelectric accelerometers is very good [4, 5].
Obviously, the quality of the vibration-based diagnosis and/or identification
of vibration-related problems mainly depends on the measured responses using
accelerometers. Therefore, good performance and high reliability of the com-
mercially available accelerometers is very important. Additionally, the use of
conventional accelerometers for multiple data collection points may increase the
complexity of monitoring system because of the associated electronic units that are
externally connected to the accelerometers. Consequently, the need for cheaper and
more reliable devices is well recognised.
With increasing demands for wireless sensing nodes used for assets control and
condition monitoring; the need for alternatives to those expensive conventional
accelerometers in vibration measurements has risen. Micro-Electro Mechanical
Systems (MEMS) accelerometers are one of the available options because of their
small size, newer technology and low cost, e.g. MEMS accelerometer may be
within 10 % of the cheapest commercially available conventional accelerometers
that come with a signal conditioning unit [4, 5]. According to Yole’s Development
report, the total market for MEMS is expected to exceed $20 Billion in 2020.
2 MEMS Accelerometers: Testing and Practical Approach for Smart Sensing. . . 21

The average price of MEMS accelerometers across all applications decreases, from
an average of $2.50 in 2004 to less than $1.70 in 2015, with consumer applications
driving the price down [6]. MEMS technology is used in some sectors such as
automotive industry for measuring pressure, temperature and air bags systems.
A few earlier researches compared the performance of MEMS accelerometers with
the conventional ones. The main differences are mainly in actuation frequencies,
amplitudes and phase shifts. In general, the selection of accelerometers is based on
their technical specifications and the measurement requirements. The main technical
specifications that decide the use of accelerometers are as the following:
(a) Sensitivity is the ratio of the electrical output of the accelerometer to its
mechanical input. The output usually is expressed in terms of voltage per
unit acceleration. The specification of sensitivity is sufficient for instruments,
which generate their own voltage, independent of an external power source. The
sensitivity of an instrument requiring an external voltage usually is specified in
terms of output voltage per unit of voltage supplied to the instrument per unit
displacement, velocity, or acceleration, e.g. milli-Volts per g of acceleration.
(b) Frequency range is the operating frequency range over which the sensitivity
of the transducer does not vary more than a stated percentage from the
rated sensitivity. The range may be limited by the electrical or mechanical
characteristics of the transducer or by its associated auxiliary equipment.
(c) Amplitude limit specifies the maximum range of acceleration that can be
measured by the accelerometer.
(d) Shock limit represents the maximum level of acceleration the accelerometer can
withstand without any damage to the unit.
(e) Linearity indicates the accuracy of the measured acceleration amplitude within
the corresponding frequency range.
(f) Natural frequency which is indirectly indicative of the measuring frequency
range. In general, a higher natural frequency allows a larger measuring fre-
quency range of an accelerometer.
However, the use of vibration analysis in condition-monitoring is based on three
key points:
1. A frequency component identifies the basic problem.
2. The amplitudes of this component and its harmonics indicate the severity of the
problem.
3. Phase relationships are used to distinguish between looseness and eccentricity
(fault diagnosis).
While a vibration spectrum can reveal much about the ranges of motion and
flexibility (resonance) of a machine, the frequency domain does not yield all the
answers. The time domain is the only place we can identify peak and peak-to-
peak amplitudes of each cycle, phase relationship between signals, and the presence
of such distinctive characteristics as truncated waveforms, pulses and modulation.
The calibration procedure adopted, generally uses the sinusoidal vibration generator
(shaker) with varying frequencies and amplitudes to characterise the accelerometer
22 A. Albarbar and S.H. Teay

being calibrated. This is achieved by comparing its measured responses with

other well-calibrated accelerometers. Accuracy of sinusoidal response measurement
satisfies the calibration procedure requirements. However, accelerometers are in
general used for measuring periodic signals (e.g. sinusoidal), as well as impulsive
and random signals.

2 Types and Design of MEMS Accelerometer

Generally, MEMS accelerometers are divided into two groups: piezoresistive

MEMS accelerometers and capacitive MEMS accelerometers.

2.1 Piezoresistive MEMS Accelerometers

Piezoresistive MEMS accelerometer is also known as strain gauge MEMS

accelerometer. Unlike piezoelectric-typed sensor, piezoresistive MEMS accelerom-
eter changes in resistance across the piezo material when stress is applied. The
history of piezoresistive MEMS accelerometer could be found in late 1970s.
Roylance et al. have successfully developed a first prototype of silicon-based strain
gauge accelerometer at Stanford University [7]. The mechanism of piezoresistive
MEMS accelerometer is simple. The first design of silicon piezoresistive MEMS
accelerometer consists of a frame, a proof mass and a thin beam as shown in
Fig. 2.1. The strain gauge is placed on the top layer of the beam, forming a half
bridge connection. When acceleration is applied to the sensor, the normal force
causes the beam to bend, leading to a change in resistance of the strain gauge. The
half bridge configuration gives proportional output to the acceleration. The proposed
design has become a fundamental topology of piezoresistive MEMS accelerometer.
Multiple beam-mass system has been introduced to improve the accelerometer’s
merit of figures [8]. Technically, piezoresistive MEMS accelerometer provides good
sensitivity. However, their low tolerance on ambient temperature conditions and
self-heating characteristics have been the development bottleneck [9]. Benefitting
from recent micro-fabrication technology, such difficulties could be overcome and
nowadays, there are relatively robust piezoresistive MEMS accelerometers available
in the market [7].
Using similar design techniques, higher number of axis, e.g. three-axis piezore-
sistive accelerometers, could also be manufactured as shown in Fig. 2.2. Four
beams with eight piezoresistors configuration are integrated to sense acceleration
in three different directions. This configuration could be visualised in an equivalent
Wheatstone bridge topology as shown in Fig. 2.2c. The parallel accelerations along
x, y and z-axis introduce beam deformation and cause compressive and tensile stress
2 MEMS Accelerometers: Testing and Practical Approach for Smart Sensing. . . 23

A’
A’
a b

Frame
Proof mass

Proof mass
Frame

Beam
Beam

Frame
A

A
Fig. 2.1 Schematic diagram of piezoresistive MEMS accelerometer (a) top view, (b) cross-
sectional view [7]

in each piezoresistor. The corresponding changes are outlined in Table 2.1. One
additional rationale of using Wheatstone bridge is to nullify the defects caused by
temperature drifts in piezoresistors [7, 8].
For packaging design, the piezoresistive MEMS accelerometer is wafer encapsu-
lated with etched glass caps, preventing over-range protection and fluidic squeezed-
film damping. To ensure better measurement experience, the transfer function
between sensor surface and glass gap can be evaluated using their damping
coefficients based on mass-spring-damper mechanical system analysis [8].

2.2 Capacitive MEMS Accelerometer

Silicon design of capacitive MEMS accelerometer allows low cost mass production
due to the mature technology on surface micromachining. ADXL series from
Analog Devices is one of dominant capacitive MEMS accelerometers in the market
[4]. Application of capacitive MEMS accelerometer include smartphone devices,
smart sensor system and low cost embedded monitoring system. Essentially, the
schematic diagram of capacitive MEMS accelerometer is illustrated in Fig. 2.3.
The movement of the proof mass due to an acceleration of the device changes
the capacitance value between electrodes on stationary fingers. This difference in
24 A. Albarbar and S.H. Teay

a b

Fig. 2.2 Structural schematic of three axis piezoresistive MEMS accelerometer [9] (a) model
view, (b) top and cross-sectional view, (c) equivalent Wheatstone bridge model

capacitance can be used to measure acceleration. An equivalent electronic circuit

incorporated with the capacitive model in Fig. 2.3 is shown in Fig. 2.4. The part
inside the rectangular box represents which its output is the differential capacitance
value from electrodes. The equivalent circuit contains signal source, a buffer
amplifier, a filter, a synchronous demodulator and electro-mechanical feedback
loop with a feedback amplifier. This model has the sensitivity of 20 mV/g for an
operational range of 0–50 g [9, 13]. To improve accuracy, symmetric design and
differential sensing are able to reduce the effect of thermal mismatch to a minimum
and linearises the capacitance different to acceleration relationship. In modern
design, Zhou et al. have proposed an MEMS capacitive accelerometer with fully
symmetrical double-sided H-shaped beam, of which the sensitivity of the device is
0.24 V/g with nonlinearity of 0.29 % over the range of 0–1 g [14].
Optimisation problem is always a concern in designing a robust capacitive
MEMS accelerometer. General design criteria include manufacturing costs,
Table 2.1 Piezoresistance value corresponding to acceleration direction
Acceleration
direction R1 R2 R3 R4 R5 R6 R7 R8
X axis Decrease Increase Decrease Increase Unchanged Unchanged Unchanged Unchanged
Y axis Unchanged Unchanged Unchanged Unchanged Decrease Increase Decrease Increase
Z axis Decrease Increase Increase Decrease Decrease Increase Increase Decrease
2 MEMS Accelerometers: Testing and Practical Approach for Smart Sensing. . .
25
26 A. Albarbar and S.H. Teay

Stationary fingers

Anchor to
substrate

Spring
legs

Proof
mass
Inertia force

Fig. 2.3 Schematic diagram breakdown respective circuit model of capacitive MEMS accelerom-
eter [10]

Differential capacitor

Filter

Signal Output
source

Amplifier Synchronous
demodulator

Inverter Feedback
amplifier

Fig. 2.4 Equivalent circuit of the integrated MEMS accelerometer [11, 12]

full-scale range, structural geometry and threshold acceleration [15]. Due to the
complexity of the mechanism, the performance of capacitive MEMS accelerometer
is still far from trivial. Various techniques have been carried to study the variables
of capacitive MEMS accelerometer to optimise its performance and robustness
[16, 17].
2 MEMS Accelerometers: Testing and Practical Approach for Smart Sensing. . . 27

3 Testing and Calibration

Calibration is carried out on a variety of equipment in different situations to ensure

that the equipment will produce results, which meet pre-defined operating criteria.
In the context of calibration of devices (MEMS accelerometers), a calibration
involves determining the performance of the sensor in reference to a known
standard. Adjustment of the overriding system may be a part of a calibration, but
not always.
In practical applications, accelerometers are usually used for measuring the
periodic (sinusoidal, sweep-sine, step-sine, multi-sine, etc.), impulsive and random
signals. Hence, these tests are usually carried out on a similar test set-up to that
shown in Fig. 2.5, and subsequently the outputs of MEMS accelerometers are
compared with well-calibrated accelerometers.

3.1 Performance Evaluation

In order to assure meaningful outputs, the performance of MEMS accelerometers

have to be critically assessed and calibrated. Albarbar et al. have conducted a
comprehensive study on performance of MEMS accelerometer for machinery-
condition monitoring [4, 5]. Three MEMS accelerometers (denoted as A, B
and C) have been validated by comparing their performance with a well-calibrated,
conventional accelerometer and subsequently were utilised to measure the vibration
signals of CNC machine in real industrial environment. Their respective technical
specification is listed in Table 2.2. The tested devices output frequency response
functions (FRS) and phase of the outputs of the tested devices are shown in
Fig. 2.6a, b, respectively.

MEMS accelerometers back to Signal processing and data

back with a conventional Charge amplifier
Display
accelerometer

S.

NI DAQ Card
Shaker

Signal
Power amplifier
generator

Fig. 2.5 MEMS accelerometer’s calibration experimental set-up

28 A. Albarbar and S.H. Teay

Table 2.2 Technical specifications of accelerometers

Conventional
accelerometer MEMS (A) MEMS (B) MEMS (C)
Sensitivity 0.1 Supply Supply Supply
(V/g) voltage/7 voltage/5 voltage/5
Frequency 1–2000 1–6000 1–10,000 1500
range (Hz)
Amplitude ˙50 ˙5 ˙3 ˙3
limited (g)
Linearity Less than ˙5 % ˙1 %/1 kHz ˙1 %/5 kHz ˙1 %/1 kHz
Shock 5000 250 1000 100
limit (g)

As anticipated, there were clear differences between the tested accelerometers

due to their design, sensitivity and noise contents of their outputs. For more details,
the reader is advised to see [4, 5].

3.2 Calibration of MEMS Accelerometers

In this section, the calibration is performed using recently developed MEMS

accelerometer (ADXL001-70) (http://www.analog.com/media/en/technical-
documentation/data-sheets/ADXL001.pdf). An accelerometer calibrator type-
4291 by Brüel & Kjær is used to generate reference mechanical vibration at 1 g
(10 ms2 ) peak ˙2 % with 79.6 Hz. A conventional Brüel & Kjær piezoelectric
accelerometer was used as a reference and a commercial data acquisition system
by National Instruments recorded the outputs. Both accelerometers are attached on
the mechanical shaker, which excites at different frequency and amplitude. A photo
of the calibration set-up is shown in Fig. 2.7. The calibration results are shown in
Table 2.3 and in Fig. 2.8.
In the other test, the mechanical shaker was set to oscillate at 50 Hz. The
sampling frequency is set to 5 kHz. Accelerometers outputs are shown in Fig. 2.9.
Note that both the power spectrums have their peak values at 50 Hz, which comply
with the dominant mechanical frequency from the shaker. The coherence of the
measured signals of both ADXL001-70 and conventional accelerometer is shown
in Fig. 2.10. It has the highest correlated coefficient when the frequency is close
to 50 Hz. This shows that both sensors have successfully obtained the important
dominant frequency information. The magnitude for harmonic frequencies between
the two signals deviates when the frequency is getting larger (100 and 150 Hz)
probably due to cross talks and noise contents [18].
2 MEMS Accelerometers: Testing and Practical Approach for Smart Sensing. . . 29

a 30 MEMS(A)/PCB Accelerometer
FRF Amplitude

20
10
0

−10
−20
0 100 200 300 400 500 600 700 800 900 1000
Frequency(Hz)

30 MEMS(B)/PCB Accelerometer
FRF Amplitude

20
10
0

−10
−20
0 100 200 300 400 500 600 700 800 900 1000
Frequency(Hz)

30 MEMS(C)/PCB Accelerometer
FRF Amplitude

20

10

−10

−20
0 100 200 300 400 500 600 700 800 900 1000
Frequency(Hz)

b 500
MEMS(A)/PCB Accelerometer
FRF Phase(deg)

−500
100 200 300 400 500 600 700 800 900 1000
Frequency(Hz)
500
MEMS(B)/PCB Accelerometer
FRF Phase(deg)

−500
100 200 300 400 500 600 700 800 900 1000
Frequency(Hz)
500
MEMS(C)/PCB Accelerometer
FRF Phase(deg)

−500
100 200 300 400 500 600 700 800 900 1000
Frequency(Hz)

Fig. 2.6 MEMS accelerometer’s response (a) magnitude and (b) phase of the frequency [4]
30 A. Albarbar and S.H. Teay

Fig. 2.7 Accelerometer

calibrator with mounted
ADXL001-70

Table 2.3 ADXL001-70 calibrated result

Parameter
Specification 0 Hz, 0 ms2 peak 79.6 Hz, 10 ms2 peak
Average voltage 2.62 V 2.64 V
Peak–peak voltage 200 mV 800 mV
RMS voltage 2.66 V 2.62 V
Dominant frequency 0 Hz 80.4 Hz (1 % deviation)

4 Packaging Technique

The delicate nature of MEMS accelerometers structures necessitates the use of

suitable packaging techniques to allow them to survive harsh conditions they would
face when used to acquire machinery vibrations in real industrial environments.
Benefits of proper packaging are to provide:
• Mechanical protection for the chip and components
• Stable temperature operating environment
• Ease of identification of sensors
• Means of mounting the sensor to equipment under test
Details of packaging typical MEMS accelerometer chip (ADXL202AE) are
explained in Figs. 2.5 and 2.6. The connections (Fig. 2.11) and components
(Fig. 2.12) for implementing the MEMS’ chip in an analogue voltage output
configuration were determined from the supplier data sheet.

5 MEMS Implementation on Machinery Diagnostics

In this section, MEMS accelerometers are used to collect vibration data from two
types of rotating machinery: a DC motor and a CNC machine. These two types of
machines are widely used in today’s industry, and it would be of great importance
2 MEMS Accelerometers: Testing and Practical Approach for Smart Sensing. . . 31

a 5 b 1
4.5 0.9
4 0.8

Amplitude (arbitrary units)

3.5 0.7
Voltage (V)

3 0.6
2.5 0.5
2 0.4
1.5 0.3
1 0.2
0.5 0.1
0 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0 50 100 150 200 250
Time (sec) Frequency (Hz)
5 1
c d
4.5 0.9
4 0.8
3.5 Amplitude (arbitrary units) 0.7
Voltage (V)

3 0.6
2.5 0.5
2 0.4
1.5 0.3
1 0.2
0.5 0.1
0 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0 50 100 150 200 250
Time (sec) Frequency (Hz)

Fig. 2.8 Calibrated vibration waveform of ADXL001-70: (a) 0 Hz, 0 ms2 , (b) frequency domain,
(c) 79.6 Hz, 10 ms2 , (d) frequency domain

if their incipient failures were successfully detected by the cost-effective MEMS

accelerometers. In other words, the adopted MEMS accelerometer has to clearly
show fundamental frequencies and harmonics of the monitored machines.

5.1 Vibration Measurements of DC Motor Test Rig

The DC motor test rig is shown in Fig. 2.13. It consists of a DC motor attached to a
shaft supported by three roller bearing and holds a metallic disk. The motor’s speed
is controlled by a speed controller, which varies from 0 rpm to a maximum rotational
speed of 2864 rpm. An ADXL001-70 MEMS accelerometer, placed on drive end
bearing housing, is used to collect the vibration data at five different speeds: 280,
914, 1602, 2281 and 2864 rpm. Time and frequency domains of collected vibration
data are shown in Fig. 2.14. The respective harmonics could be seen, in particular,
at rotational speeds higher than 1602 rpm.
32 A. Albarbar and S.H. Teay

1
ADXL-001-70
0.9 Conventional accelerometer

0.8

0.7
Normalised PSD

0.6

0.5

0.4

0.3

0.2

0.1

0
0 50 100 150 200 250 300 350 400
Frequency (Hz)

Fig. 2.9 Frequency spectrum of acquired vibration signal using ADXL001-70 and conventional
accelerometer at 50 Hz

0.9

0.8

0.7

0.6
Coherence

0.5

0.4

0.3

0.2

0.1

0
0 50 100 150 200 250 300 350 400
Frequency (Hz)

Fig. 2.10 Coherence between the vibration signals acquired using ADXL001-70 and conventional
accelerometer
2 MEMS Accelerometers: Testing and Practical Approach for Smart Sensing. . . 33

+5v

0v

1nF 1nF

8 5
0.1µF VDD XOUT

AADXL202AE
3 6
COM YFILT
2 7
1MΩ T2 XFILT
4 2
YOUT ST

Fig. 2.11 MEMS accelerometer wiring diagram

Fig. 2.12 MEMS packaged sensor components: (a) sensor outer casing (mild steel), (b) threaded
mounting bottom plate, (c) flat mounting bottom plate with MEMS affixed, (d) MEMS with
components attached, (e) Packaged MEMS sensor

Fig. 2.13 Experimental set-up of DC motor vibration monitoring

34 A. Albarbar and S.H. Teay

a
2 4.88Hz
Voltage (V)

|Y(f)|
0
0 10

−2

0 0.2 0.4 0.6 0.8 1 0 100 200 300 400 500

Time (s) Frequency (Hz)
b
2 14.65Hz
Voltage (V)

|Y(f)|
0
0 10

−2

0 0.2 0.4 0.6 0.8 1 0 100 200 300 400 500

Time (s) Frequency (Hz)

c 26.86Hz 53.71Hz 80.57Hz

2
Voltage (V)

|Y(f)|
0
0 10

−2

0 0.2 0.4 0.6 0.8 1 0 100 200 300 400 500

Time (s) Frequency (Hz)

d 39.06Hz 78.13Hz 109.9Hz

2
Voltage (V)

|Y(f)|

0
0 10

−2

0 0.2 0.4 0.6 0.8 1 0 100 200 300 400 500

Time (s) Frequency (Hz)

e 48.83Hz 97.66Hz 148.9Hz

2
Voltage (V)

|Y(f)|

0
0 10

−2

0 0.2 0.4 0.6 0.8 1 0 100 200 300 400 500

Time (s) Frequency (Hz)

Fig. 2.14 DC motor vibration and power spectrum measured by ADXL001-70 at different rotating
speeds: (a) 280 rpm (4.67 Hz), (b) 914 rpm (15.23 Hz), (c) 1602 rpm (26.7 Hz), (d) 2281 rpm (38
Hz), (e) 2864 rpm (47.73 Hz)

5.2 Vibration Measurements of CNC Machine

The MEMS accelerometer is placed on CNC machine in a typical industrial

environment as shown in Fig. 2.15. The CNC machine operates under a speed
of 2400 rpm. Respective time and frequency domains of the vibration signals are
plotted in Fig. 2.16. Experimental results show that the important fundamental
frequencies and their harmonics are captured by the MEMS accelerometer.
The actual vibration amplitude for the machine was about 0.2 g peak (˙0.01 g—
using hand-held commercial vibration monitoring device) which gives clear evi-
dence on the correctness of the vibration amplitude indicated by the MEMS
accelerometer.
2 MEMS Accelerometers: Testing and Practical Approach for Smart Sensing. . . 35

Fig. 2.15 Experimental set-up of CNC machine vibration monitoring

a b

Fig. 2.16 CNC machine vibration measured by ADXL001-7 at speed of 2400 rpm (a) time
domain and (b) frequency domain [4]

The CNC machine was run at 2400 rpm, and at this speed the fundamental speed
and harmonics are 40, 80 and 160 Hz which were clearly presented in the frequency
domain of Fig. 2.16.

6 Design and Implementation of MEMS-Based

Smart Sensors

Previous sections have outlined the characteristics and fundamentals of MEMS

accelerometers. MEMS accelerometers are envisioned as future candidates in
realising the practical deployment of smart sensing technology. Smart sensor/trans-
ducer provides more complex functions than raw signal acquisition by integrating
embedded system into the sensing architecture. Possible features of smart sensors
include programmable signal conditioning, on-board signal processing and analysis,
simple decision-making and remote communication. Generally, there are two major
advantages of using MEMS accelerometers over their conventional counterparts for
smart sensing development:
36 A. Albarbar and S.H. Teay

• Low power consumption: A conventional accelerometer requires charge amplifier

to amplify the induced acceleration by conventional accelerometer. Such an
amplifier usually requires a supply voltage higher than that of an embedded
system (which average operational voltage is between 1.65 and 5 V and about
15 mA).
• Low cost of deployment: The cost of an MEMS accelerometer could be less
than 10 % compared with commercially available accelerometers. The average
price of MEMS has been declining since 2009, and it is believed that it will
continue to do so in the next few years (predicted figure from $1.01 to $0.90).
High performance accelerometer such as ADXL001 has an average market price
at $30, which is at least 50 % cheaper than commercially available conventional
accelerometer.
To validate the feasibility of MEMS accelerometer for smart sensing application
on rotating machinery diagnostics, working system architecture is designed and
developed. Its respective hardware architecture could be partitioned into five main
modules: sensors, embedded microcontrollers (peripheral integrated circuit and core
computational unit), external data storage module and wireless communication
module. The architecture design is shown in Fig. 2.17. The working prototype is
also developed as shown in Fig. 2.18.
The sensing element—MEMS accelerometer—is responsible for transducing
physical movement into electronic signals. The resulting electronic measurements
are then sent to the peripheral computational unit, where necessary analogue signal
conditioning is carried out. The processed analogue signals are then transmit-
ted to computational core for signal processing, analysis and fault diagnostics.
The quantised data is logged and interrogated in removable data storage device
(universal serial bus (USB) flash drive or secure digital (SD) card) through data
storage module. The end-result and diagnosis report is transmitted to another system
platform through wireless communication module. A flowchart to represent the
workflow of the developed smart sensor is illustrated in Fig. 2.19.

Main computational unit for signal

Peripheral IC for anti-aliasing filter processing, condition monitoring Information communication with
and power management. and health assessment other modules/human experts
Core Wireless
Sensors Peripheral IC
computational unit communication
(MEMS) (8bit MCU)
(32bit MCU) module–XBee
Removable
memory unit

Saved files for user preferences and

data logging

Power supply

Fig. 2.17 System architecture of MEMS-based smart sensor

2 MEMS Accelerometers: Testing and Practical Approach for Smart Sensing. . . 37

a b Passive electronic
components incl.
transistor bank

Data
storage
module

Wireless
communication
Sensing device
channel

Printed circuit Peripheral

Main computational computational
board core development kit unit

Fig. 2.18 Working prototype of MEMS-based smart sensor system (a) prospective view and
(b) top layout view

On-board vibration measurement is one of the crucial elements for MEMS-based

smart sensor. For example, to sample a 16-bit (2-byte) vibration signal for 3 s in
5 kHz requires at least 30 kB data memory (2  3  5000 D 30 kB). Such data load is
hectic to be logged into internal memory of commercially available microcontroller,
as tabulated in Table 2.4. To tackle the above design challenge, an external data
storage module is necessary. The end terminal of data storage module is removable
mass storage device, such as USB thumb-drive or secure digital (SD) card. This has
extended high volume of data storage, which has flash memory counted in gigabyte.
The developed smart sensor has the ability to collect raw vibration, compute sta-
tistical features and send analysed result to user interface autonomously. Advanced
signal processing or artificial intelligence could be performed on-board for complex
machinery diagnostics. Such system has significant advantage in reducing power,
cost and communication overhead in future sensing technology.
In terms of sustainability, MEMS-based smart sensor has potential to integrate
energy harvesting technology in the future, which is another trending topic in mod-
ern engineering. Most commercial digital integrated circuit provide programmable
sleep/hibernate mode to conserve energy consumption. Respectively, the power
characteristics of developed smart sensor are tabulated in Table 2.6.
MEMS-based smart sensor integrates a wireless communication module such
as XBee or XBee-PRO. Its modular design architecture and direct digital interface
have enabled relatively fast transmission between the smart sensor and other sensing
platform. Several customised commands are pre-programmed to access smart sensor
through computer’s user interface as tabulated in Table 2.5. They are coded in 8-bit
ASCII character data-type, providing good readability for the user to monitor the
console.
The developed smart sensor has the ability to collect raw vibration, compute sta-
tistical features and send analysed result to user interface autonomously. Advanced
signal processing or artificial intelligence could be performed on-board for complex
machinery diagnostics. Such system has significant advantage in reducing power,
cost and communication overhead in future sensing technology.
38 A. Albarbar and S.H. Teay

Start

Acquire signals using

sensors

Signal conditioning in
peripheral
computational unit

Signal processing in
main computational
unit

Signal analysis with

feature extraction and
data fusion

Data classification

Send ‘Healthy’ signal No Yes

Is the signal Send ‘Alarm’ signal to
to sensing node
faulty? sensing node gateway
gateway

Log data into external

data storage device

Turn sensor to
hibernate mode

Waiting for the next

operation dutycycle

Fig. 2.19 Workflow of smart sensor

In terms of sustainability, MEMS-based smart sensor has potential to integrate

energy harvesting technology in the future, which is another trending topic in mod-
ern engineering. Most commercial digital integrated circuit provide programmable
sleep/hibernate mode to conserve energy consumption. Respectively, the power
characteristics of developed smart sensor are tabulated in Table 2.6.
2 MEMS Accelerometers: Testing and Practical Approach for Smart Sensing. . . 39

Table 2.4 Available memory capacity of various commercial microcontrollers

Maximum flash Data memory
Microcontroller Manufacturer memory (kB) (RAM) (kB)
PIC32MX575F512L Microchip 512 64
dsPIC33EP512MU814 Microchip 536 52
AT32UC3C0512C Atmel 512 64
MC56F84789 Freescale 256 32
CY8C5868AXI-LP032 Cypress 256 64

Table 2.5 Commands to communicate with IMSEM through XBee

module
Commands Description
Menu Shows menu of the IS
Initialise Reset or initialise the IS together with the SD card module
Read Read data that has been stored in the SD card of IMSEM
Auto Start monitoring process with time-driven mode
Calculate Calculate feature components of target signals in SD card
Debug Reserved for debugging purpose

Table 2.6 Power budget of smart sensor

Parameter Value
Maximum rated voltage (Identified from the system 7.5 V
modules’ data sheet)
Rated voltage 5.6 V
Minimum voltage 4.8 V
Current drawn (Operation mode) Average 0.22 A
Current drawn (Idle mode) Average 0.2 A
Current drawn (Inactive mode) 8 A
Rated power consumption (Operation mode) 1.232 W
Rated power consumption (Idle mode) 1.12 W
Rated power consumption (Inactive mode) 44.8 W

7 Concluding Remarks

MEMS accelerometers offer excellent alternatives to conventional types of those

transducers as there is no need to carry heavy charge amplifiers, but the choice has
to be made according to specifications and through validation tests. MEMS sensors
have also to resist harsh environments using an appropriate packaging.
There are a number of MEMS accelerometers that could be used for monitoring
machinery vibrations. The technical specifications for each of them should be well
understood and must be suitable for the intended application. MEMS accelerometers
are of sensitive and breakable structure, and thus they should be suitably packaged
40 A. Albarbar and S.H. Teay

to withstand harsh environments without affecting their specifications. Sensitivity,

resolutions and frequency range are the most important specifications that should be
taken into account when choosing MEMS accelerometers
Adequate performance tests of MEMS accelerometers should be conducted
under different excitations signals including sinusoidal, impulse and random. The
measured responses of the MEMS accelerometers shall also be compared with a
well-calibrated type accelerometer. Moreover, their outputs contain high amount of
noise and extra un-interpretable peaks which necessitate full understanding of the
measured vibration signals.

References

1. Albarbar A (2012) Vibration and noise of diesel engines. Lap Lambert Academic, Saarbrücken.
ISBN 978-3-659-11039-9
2. Shi H-Y, Wang W-L, Kwok N-M, Chen S-Y (2012) Game theory for wireless sensor networks:
a survey. Sensors 12(7):9055–9097
3. Ibrahim G, Albarbar A, Abouhnik A, Shnibha R (2013) Adaptive filtering based system for
extracting gearbox condition feature from the measured vibrations. Measurement 46:2029–
2034
4. Albarbar A, Sinha J, Starr A (2009) Performance evaluation of MEMS accelerometers. J Meas
42(5):790–795
5. Albarbar A, Mekid S, Starr A, Pietruszkiewicz R (2008) MEMS accelerometers! Are they
suitable for machinery condition monitoring? Int J Sensors 8(9):784–799
6. Growing in maturity, the MEMS industry is getting its second wind Status of the
MEMS Industry report, Yole Development. http://www.yole.fr//iso_upload/News/2015/PR_
MEMSMarkets_YOLE_May2015.pdf. Accessed Jan 2016
7. Plaza J, Collado A, Cabruja E, Esteve J (2002) Piezoresistive accelerometers for MCM
package. J Microelectromech Syst 11(6):794–801
8. Biefeld V, Buhrdorf A, Binder J (2000) Laterally driven accelerometer fabricated in single
crystalline silicon. Sensors Actuators A Phys 82(1):149–154
9. Liu C, Kenny TW (2001) A high-precision, wide-bandwidth micromachined tunneling
accelerometer. J Microelectromech Syst 10(3):425–433
10. Kim S-C, Wise KD (1983) Temperature sensitivity in silicon piezoresistive pressure transduc-
ers. IEEE Trans Electron Devices 30(7):802–810
11. Doll JC, Corbin EA, King WP, Pruitt BL (2011) Self-heating in piezoresistive cantilevers. Appl
Phys Lett 98(22):223103
12. Zhou X, Che L, Liang S, Lin Y, Li X, Wang Y (2015) Design and fabrication of a MEMS
capacitive accelerometer with fully symmetrical double-sided H-shaped beam structure.
Microelectron Eng 131:51–57
13. Xie H, Fedder G (2000) CMOS z-axis capacitive accelerometer with comb-finger sensing. In:
Proceedings of IEEE Micro Electro Mechanical Systems (MEMS), pp 496–501
14. Bao M (2005) Chapter 1—Introduction to fMEMSg devices. In: Bao M (ed) Analysis and
design principles of MEMS devices. Elsevier Science, Amsterdam, pp 1–32
15. Analog Devices (2015) MEMS accelerometers. Analog Devices. http://www.analog.com/en/
products/mems/mems-accelerometers.html. Accessed 2 Oct 2015
16. Roylance LM, Angell JB (1979) A batch-fabricated silicon accelerometer. IEEE Trans Electron
devices 26(12):1911–1917
17. Shaoqun S, Jian C, Minhang B (1992) Analysis on twin-mass structure for a piezoresistive
accelerometer. Sensors Actuators A Phys 34(2):101–107
18. Badri AE, Sinha JK, Albarbar A (2010) A typical filter design to improve the measured signals
from MEMS accelerometer. Measurement 43(10):1425–1430
http://www.springer.com/978-3-319-32178-3