Beruflich Dokumente
Kultur Dokumente
in Fig. 3(b), the fans of the matrix are controlled by a relay interface board that allows
turning on and off the fans of the matrix independently, in order to create different
situations of faulty fans in fault diagnosis tests.
Fig. 3. (a) Foxconn D90SM-12 fan. (b) Fan matrix built for the test
It has been noticed that if one fan fails, the maximum position and value of the
acoustic image change. One of these effects can be observed on Fig. 4, which shows
the maximum positions of different acoustic images of the fan matrix with one faulty
fan, whose real position is showed as a red cross.
Fig. 4. Maximum positions of the acoustic images of fan matrix with one faulty fan (the position
of this faulty fan is represented with a red cross).
After obtaining these geometrical parameters of the acoustic images, they are given
to the classification algorithm, based on a linear SVM, in order to detect which fan is
not working properly. The obtained accuracy rate for the SVM algorithm is 95.6%.
This result shows that the purposed methodology is reliable when one fan of the matrix
fails, because it detects the faulty fan position.
The accuracy rate results, obtained by the SVM algorithm, in this robustness
analysis are shown in Table 1. These results show that as the two faulty fans are more
separated, the accuracy of the algorithm to detect one of the fans that is not working as
faulty decreases. In these tests, although in most cases the SVM classifier can’t identify
correctly which are the faulty fans, but it can alert if there is a problem with the fan
matrix, it alerts if the matrix is not working properly.
4 Conclusions
This paper shows a fault detection methodology developed to detect if one fan of a fan
matrix is not working properly. This methodology, based on geometrical parameters of
acoustic images of the fan matrix and in a Support Vector Machine algorithm, is
reliable when only one fan of the matrix is not working.
This methodology has been designed for these specific situations, obtaining a really
good accuracy rate. But its robustness under other unexpected circumstances, i.e. with
more than one faulty fan on the matrix at the same time, has also been analyzed. This
analysis has shown that this methodology fails if more than one fan fails at the same
time. Under this circumstance, the methodology does not detect which are the faulty
fans. But it detects successfully that the matrix is not working properly.
Future work must be done to solve this limitation. One solution could be training
the SVM algorithm to detect individual failures for each fan of the matrix.
Acknowledgements. This work is supported by the Spanish research project SAM TEC 2015-
68170-R (MINECO/FEDER, UE). The authors also gratefully acknowledge the helpful com-
ments and suggestions of the reviewers, which have improved the presentation.
Conflict of Interest. The authors declare that there is no conflict of interest regarding the
publication of this paper.
References
1. Reppa V, Tzes A (2009) Fault detection and diagnosis based on parameter set estimation.
IET Control Theory Appl 5(1):69–83. https://doi.org/10.1049/iet-cta.2009.0202
2. Jardine AKS, Lin D, Banjevic D (2006) A review on machinery diagnostics and prognostics
implementing condition-based maintenance. Mech Syst Signal Process 20(7):1483–1510.
https://doi.org/10.1016/j.ymssp.2005.09.012
3. Haddar M, Bartelmus W, Chaari F, Zimroz R (2016) Applied Condition Monitoring. Book
Series. Springer. ISSN 2363-698X
4. Dalpiaz G, Rubini R, D’Elia G, Cocconcelli M, Chaari F, Zimroz R, Bartelmus W,
Haddar M (2013) Advances in condition monitoring of machinery in non-stationary
operations. In: Proceedings of the third international conference condition monitoring of
machinery in non-stationary operations, CMMNO 2013. ISBN 978-3-642-39348-8
5. Lu Y, Tang J, Luo H (2012) Wind turbine gearbox fault detection using multiple sensors
with features level data fusion. J Eng Gas Turbines Power 134:042501-1-8. https://doi.org/
10.1115/1.4004438
6. Ravetta P, Muract J, Burdisso R (2007) Feasibility study of microphone phased array based
machinery health monitoring. Mecánica Computacional 26:23–37
7. Lu W, Jiang W, Yuan G, Yan L (2013) A gearbox fault diagnosis scheme based on near-
field acoustic holography and spatial distribution features of sound field. J Sound Vib
332:2593–2610. https://doi.org/10.1016/j.jsv.2012.12.018
8. Herold G, Sarradj E (2015) Microphone array method for the characterization of rotating
sound sources in axial fans. Noise Control Eng J 63(6):546–551. https://doi.org/10.3397/1/
376348
228 L. del Val et al.
9. Zenger F et al (2016) Sound source localization on an axial fan at different operating points.
Exp Fluids 57:136. https://doi.org/10.1007/s00348-016-2223-8
10. Van Trees H (2002) Optimum array processing: part IV of detection, estimation and
modulation theory. Wiley, New York
11. Van Veen BD, Buckley KM (1988) Beamforming: a versatile approach to spatial filtering.
IEEE ASSP Mag 5:4–24. https://doi.org/10.1109/53.665
12. Izquierdo A et al (2016) Design and evaluation of a scalable and reconfigurable multi-
platform system for acoustic imaging. Sensors 16(10):1671. https://doi.org/10.3390/
s16101671
13. del Val L et al (2017) Using a planar array of MEMS microphones to obtain acoustic images
of a fan matrix. J Sens. Article ID 3209142. https://doi.org/10.1155/2017/3209142
14. del Val L et al (2017) First steps on fan matrix condition monitoring and fault diagnosis
using an array of digital MEMS microphones. In: Proceedings of meetings on acoustics, vol
30, p 030014. https://doi.org/10.1121/2.0000656
15. del Val L et al (2018) Analysis of a maxima position estimator on acoustic images of a fan
matrix for failure detection. In: Proceedings of the Euronoise 2018 Conference, Crete, May
2018
Experimental Characterization
of Metal-Mesh Isolator’s Damping
Capacity by Constitutive
Mechanical Model
1 Introduction
The metal mesh isolator essentially consists of a cushion of stainless steel, woven
using a knitting machine, rolled and/or pressed into the required geometric shape
via a press mold in order to achieve the desired geometric shape, so that different
geometries can be manufactured depending on the application by changing the
mold process. It has been widely used in vibration control engineering applica-
tions such as isolation mounting of machine tools.
In order to evaluate the damping capacity including the loss factor, damp-
ing ratio, and inverse quality factor, common measurement methods were used.
In the literature, the measurement of material’s damping capacity consist in
estimating either its inverse quality factor or loss factor via the half-power
bandwidth (HPB) [1]. The convenient HPB formula provides a method used
to determine the inverse quantity factor for linear system. For nonlinear system
that exhibits a jump discontinuities in their response, the use of this method is
invalid since one of the half power points ceases to be observed. To overcome
this issue, the quantity factor is obtained from the measurements by fitting the
response curve with the Lorentzian function [2]. However, this estimation lead
to be poor for strong nonlinearity and when the unknown parameters depend on
the motion amplitudes. Davis [3] provided a new formula derived from the HPB
method and used in the same manner but with the physical validity significantly
improved accuracy. The Rational Fraction Polynomial (RFP) method [4] is one
of the traditional techniques in the modal analysis applications for identifying
the modal parameters of the predominant modes. Curve fitting has been per-
formed on the measurement to estimate the natural frequency and the damping
ratio.
The common feature of these aforementioned methods is that the physical
constitutive model is not taken into consideration and the dynamic mechan-
ical properties are not reflected in these methods. Experiments have showed
that the mechanical properties of metal mesh devices, such as the stiffness and
the damping, are related with a number of factors including the material and
structural parameters, installation conditions, molding direction. However, a few
publications exhibit the dynamic hysteretic behavior of mechanical systems in
the microscopic scale where the input-output dynamic relations of the structure
components are represented in the form of hysteresis loops. The modeling of the
hysteretic behavior is a prerequisite for predicting the overall dynamic response
which is highly dependent on the ability of energy dissipation of its members.
In past decades, various mathematical models have been developed to describe
the behavior of hysteretic processes and investigate the hysteresis phenomena.
These models are classified into operator-based models and differential-based
models. Preisach model [5], which is a popular model that belongs to the class
of operator based models, is widely used to model the hysteresis behavior in elec-
tromagnetic systems, because of its good performance in the characterization at
narrow-band of frequency with no-load condition. Krasnosel’skii-Pokrovskii (KP)
model, obtained by introduction the Preisach hysteresis model into a formal-
ized mathematic form, is able to characterize the hysteretic behavior of a smart
Experimental Characterization of Metal-Mesh Isolators Damping Capacity 231
actuator, especially SMA actuator. Prandtl-Ishlinskii (PI) model has been pro-
posed, as a subclass of the Preisach model, to characterize and describe the
hysteresis nonlinearities of smarts materials actuators [6–8]. Unlike the previous
models (Preisach, KP and PI) that fall in the class of operator-based models,
other types of hysteresis models in the category of differential based are intro-
duced in literature, such as the Duhem model and Bouc-Wen model. The Duhem
model [9] is a differential equation-based hysteresis model that was proposed by
P. Duhem in 1897 to simulate a dynamic model. Although the Duhem model
can describe a large class of hysteresis, it is presented in more complex way than
the Bouc-Wen model. The Bouc-Wen model is widely adopted model in mechan-
ical applications due to its versatility and simplicity in the characterization of
the hysteretic behavior and thereby has been frequently employed to describe
hysteresis nonlinearity. The Bouc-Wen model was firstly introduced by Bouc in
1971 [10] as a theoretic analysis of a favorable hysteresis model and later used
by Wen in 1976 [11]. Zhu [12] proposed a non-symmetrical Bouc-Wen model to
describe the asymmetrical hysteresis in the piezoelectric ceramic actuators by
introducing a non-symmetrical formula into the Bouc-Wen hysteresis operator.
A modified Bouc-Wen (MBW) model is presented in [13] to describe the asym-
metric hysteresis of piezoelectric actuator, and to identify the model parameters
using a modified differential evolution (MDE) algorithm.
Based on the constitutive mechanical model of the metallic-wires material,
a damping capacity measurement method, presented in [14] and based on the
decomposition of the hysteresis loop, is brought forward in this paper to mea-
sure the equivalent loss factor of nonlinear material. The constitutive mechanical
model, involving the asymmetric non-linear elastic force, viscous damping and
hysteretic coulomb friction, are taken into consideration to describe the nonlinear
general restoring force. The measurement of metallic-wires material’s damping
capacity at different loading frequencies and amplitudes requires the identifica-
tion of the mechanical model’ parameters. In this paper, an experimental iden-
tification methodology is presented to determine the unknown parameters of
the constitutive mechanical model. The Bouc-Wen model was implemented to
identify the unknown parameters of the hysteretic damping force.
2 Experiment
2.1 Experimental Setup
As shown in the schematic diagram (Fig. 1), the test rig essentially consists
on hydraulic actuator to apply the vertical load to the wire-mesh specimen, lower
and upper clamps to assure the preloading, a force transducer mounted on the
upper clamp to measure the applied force and a displacement-sensor fixed to the
lower clamps to evaluate the compression-motion along the vertical direction.
The displacement and force signals of the specimen are recorded by the data
acquisition system. The measured data are affected by the noise due to the
sensor sensitivity, as depicted in Fig. 2. To remedy this error, a filter technique
is implemented in the recorded signal by removing the sub-harmonics above the
third harmonic.
10-3
1.5 150
Recorded Signal Recorded Signal
Filtred Signal Filtred Signal
1 100
0.5 50
Dispalcement [m]
Force [N]
0 10-4 0
10
50
-0.5 -50 0
9.5
-50
-1 -100 -100
9
0.05 0.055 0.06 0.05 0.1 0.15
-1.5 -150
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
Time Time
(a) (b)
Fig. 2. the recorded and filtered signals of (a) the displacement and (b) the force with
fixed amplitude of 1.0 m and frequency of 12 Hz
Experimental Characterization of Metal-Mesh Isolators Damping Capacity 233
where
fK (t): the asymmetric nonlinear elastic force.