Sunteți pe pagina 1din 7

Real Time Processing of Velocity & Displacement

Parameters from Accelerometer Signals in Machinery


Protection System
Mohit Kalra1, D.A.Roy2 and U.W.Vaidya3
Reactor Control Division 1, 2, 3
Bhabha Atomic Research Centre, Trombay, Mumbai 400085
1kalra@barc.gov.in, 2daroy@barc.gov.in, 3uday@barc.gov.in

ABSTRACT
For vibration monitoring and protection of large rotating machines like turbines, large motors,
etc, a DSP based Machinery Protection System (MPS) has been designed and developed at RCnD. The
system monitors vibration parameters of interest from sensors mounted on the machines. For an
elastic structure, the displacement is proportional to the stress, which is required to analyze the
structural integrity. The displacement and velocity of movement are important parameters that need
to be measured. This paper presents implementation of discrete signal processing algorithm from
which velocity and displacement data can be found from acceleration data measured using
accelerometers via discrete integration.
Key Words: Vibration, MPS.
1. INTRODUCTION
Machinery Protection System (MPS) consists of an embedded system installed in 19" bin and a
PC-based Engineering Console (EC), which are connected through Ethernet. The embedded system is
of modular construction and is user configurable through the Engineering Console. The embedded
system acquires various dynamic signals (vibration signals) and if any of these signals crosses the
ALARM (alert) limit, it generates a digital alarm output and if the measured value crosses the TRIP
(danger) limit then it generates a trip contact output which is used to protect/shutdown the
machinery. The block diagram of Machinery Protection System is shown in fig. 1.
Embedded System
VME Bus

Engineering
Console
S
C
M

R
O
M

M
P
M

Ethernet Network
Fig. 1 Block Diagram of MPS

The Embedded System consists of following modules.


DSP based Machinery Protection Module (MPM)
Input Output Module (IOM)
Relay Output Module (ROM)

I
O
M

Power Supply Module (PSM)


System Controller Module (SCM)
Accelerometers are widely used to measure the absolute vibrations of the various components
of the large rotating machineries and its supported structures. Accelerometers are of small size and
lightweight and can be easily attached to the body. Accelerometers work well over a very wide range
of frequencies & are the most popular sensor in the study of vibrations because of its ability to pick up
high frequency content, higher sensitivity and good thermal stability. Thus it is advantageous to use
accelerometers for measuring vibration parameters. The process of converting an acceleration
measurement into a displacement measurement is accomplished using double integration. The first
discrete integration is performed on an acceleration signal to get a velocity signal. Subsequently,
second discrete integration is performed on the velocity signal to get the displacement.
For performing discrete integration on sampled data the standard method is to calculate the
area under the curve of the appropriate trace. For measured or digitized data an approximate solution
can be found by using numerical integration techniques such as rectangular or trapezoidal integration
method. Certain challenges need to be addressed when performing a double integration. Integration
requires a known initial condition, whether it is initial velocity or displacement. Secondly, if the data
contains even a small amount of low frequency or DC offset components (due to drift in
accelerometer) then these can often lead to misleading (although numerically correct) results. The
problem is not caused by loss of information inherent in the digitization process nor due to the effects
of amplitude or time quantization. It is in fact a characteristic of integrated trigonometric functions
that their amplitudes increase with decreasing frequency. Both the above problems of initial condition
and dc offset can lead to serious integration errors and present major challenges in the design of
algorithm.
2. SIGNAL PROCESSING ALGORITHM
To get the displacement parameter the acceleration signal has to be double integrated. The first
integration gives the velocity signal as given below.
t

v(t ) = v(t 0 ) + a( )d

(1)

t0

Where a( ) is the acceleration signal, v(t ) is the velocity signal, t 0 is the initial time and v (t 0 ) is
the initial velocity, which is constant. The second integration of the velocity signal gives the
displacement as given below.
t

x(t ) = x(t 0 ) + v( )d

(2)

t0

Where x(t ) is the displacement signal, v( ) is the velocity signal, t 0 is the initial time and x(t 0 )
is the initial displacement, which is constant.
Therefore, for double integration to be performed on acceleration signal, two initial conditions
namely velocity v (t 0 ) and displacement x(t 0 ) must be known. However it is impractical to get the
initial condition of the rotating machines.
3. DIGITAL INTEGRATION ALGORITHMS
A number of discrete integration algorithms are available to perform integration. The
acceleration signal is sampled at the sampling rate, fs, making it a discrete function of time. The
simplest way to perform numerical integration is to use the rectangular integration method. This

method uses an accumulator to sum all past sampled inputs and the current input sample and divide
by the sampling rate fs. Rectangular integration is represented by the following difference equation:

y (n ) =

1
fs

x(n k ) = y(n 1) +
k =0

1
x(n )
fs

(3)

Where x(n) is the integrand, y(n) is the output of the integrator, and fs is the sampling frequency.
Another well known numerical integration method uses the trapezoidal rule. The results are
more accurate with this method than with the rectangular method. The difference equation for
trapezoidal integration is:

y (n ) = y (n 1) +

1
[x(n 1) + x(n )], n > 0
2 fs

(4)

Another numerical integration method uses Simpsons rule. It is defined by the following
difference equation:

y (n ) = y (n 1) +

1 x (n 1) + 4 x (n ) + x (n + 1)

6
fs

(5)

Unlike the other methods, the Simpsons method requires a future sample of the integrand, x, to
get the current sample of the integrated signal, y, so it cannot be performed in real time. After
analyzing the different methods it was decided to use the trapezoidal method of integration to perform
integration.
4. DESIGN OF ALGORITHM
For designing an integration algorithm it is necessary to know the initial condition of velocity
and displacement. After the Acceleration signal is integrated, the velocity signal having DC component
is achieved. If initial velocity is not added to this signal and integrated further to get the displacement,
the displacement signal will increase linearly and give misleading results. It is impractical to get the
initial velocity and initial displacement for machine vibrations or structural vibrations. So a high pass
filter is used to remove that DC component of the velocity signal and after the velocity signal is
integrated to get displacement, the displacement signal is also subjected to high-pass filtering to
remove DC component.
Another design issue is with the accelerometer drift (DC offset in the signal). Ideally, there
should be no DC offset from the accelerometer for the measurement of a vibration. The presence of
drift can lead to large integration errors. If the acceleration signal from a real accelerometer was
integrated without any filtering performed, the output could become unbounded over time. To solve
the problem of accelerometer drift, a high-pass filter is used to remove the DC component of the
acceleration signal. The frequency response of the filter must have a very low cutoff frequency
compared to the bandwidth of the signal. By filtering before integrating, drift errors (dc offset errors)
are eliminated.
The following algorithm has been developed to double integrate the acceleration signal to get
the displacement signal. The algorithm contains following five steps for double integrating the
acceleration signal to get the displacement signal.
1. High pass filter to remove the DC (drift) errors.
2. Integration to get the velocity
3. High pass filter to remove DC component of the velocity to eliminate the need for initial
velocity condition.
4. Integration to get the displacement.

5. High pass filter to remo


remove
ve DC component of the displacement to eliminate the need for
initial displacement condition.
The block diagram of the above algorithm is shown in fig. 2.

Fig. 2 Block Diagram of Algorithm

5. IMPLEMENTATION ON TARGET
RGET HARDWARE
The above algorithm was implemented on Digital Signal Processor (DSP)) based Machinery
protection module of MPS. The MPM uses a h
high performance 32-bit DSP,, which runs at 100 MHz. The
MPM acquires analog signals, processes the acquired inputs using the DSP and if any parameter
exceeds the set alarm limit, it generates alarm. The system configuration from the Engineering Console
can be downloaded to MPM through onboard communication (Ethernet) controller. The MPM module
is shown in fig. 3.

Fig. 3 MPM Module

The choice of high pass filter was decided based on performance and real time operation. The
advantage of FIR filter is that it has liner phase response & cut
cut-off
off frequency can be set very low
around 5-10
10 Hz. Its disadvantage is that filter order can be very high which will require large time for
computations. IIR filter has an advantage over FIR filter with respect to filter order. An IIR filter that
meets the same magnitude response specifications will have a much lower order than its FIR
counterpart.
part. Therefore, computations can be done faster with an IIR filter. However, its phase
response isnt linear like the FIRs response.
In MPS system the machine vibrations acceleration signals in the frequency band of 50Hz to
1000Hz are acquired. A sampling
ng frequency of 7000Hz is used to digitize the acquired analog
acceleration signal. For filtering, a lower sampling rate would be useful to make the filter specifications
easier to meet. However, if the sampling rate gets too low, the accuracy of the integration
inte
decreases.
To use the FIR filter having cut-off
off frequency of around 40Hz with sampling rate of 7000Hz will
require very high filter order & very large computation time. Because of timing limitation IIR filter of
4th order having cut-off
off frequency 4
40 Hz was implemented.

6. RESULTS
A test setup comprising of Embedded System (MPM module, IOM module), function generators
and PC based console were used to test the above algorithm as shown in fig. 4.
Machine Protection System
Function Generator
Sinusoidal
Signal

Embedded
System

PC-Based
Console

Ethernet Link
Fig. 4 Test Setup

A signal from the function generator was fed to the MPM module. After processing on MPM
module the real time processed value is sent to PC-Based OC over Ethernet. MPM module also
computes the RMS of the integrated data and sends to the PC-based OC. Sinusoidal signal of various
frequencies were fed to the MPS system from the function generator and the first integration
(Velocity) & second integration (Displacement) results were obtained. The following figures show
various signals fed from function generators and their first integration and second integration
waveforms.

Fig. 5 600Hz Sinusoidal Signal as Acceleration

Fig. 6 First Integration (Velocity)

Fig. 7 Second Integration (Displacement)

Fig. 8 Two Sinusoidal Signals

Fig. 9 First Integration (Velocity)

Fig. 10 Second Integration (Displacement)

RMS values were also computed for 600 samples of integration of above signals. The above
algorithm was also implemented on simulation environment and the data from the real-time
processing in MPM module were compared with the simulation results. The results are given in table 1
below.
Table 1: Comparison of the results with MATLAB
Sr.
No.

1.

Signal

600
Hz
4V Peak
Amp
200Hz 1V
Peak Amp
+ 600 Hz
2V Peak
Amp

Typical
Accelerome
ter
Sensitivity

RMS of
First
Integration
(mm/S)

RMS of
Second
Integration
(m)

Simulation
Result of
RMS of
Second
Integration
( m)
18.5

Error for
First
Integration
(%)

Error for
Second
Integration
(%)

17.7

Simulation
result of
RMS of
First
Integration
(mm/S)
71.7

100mV/g

68.7

4.2

4.2

62.0

41.7

65.6

44.7

5.5

6.7

7. CONCLUSION
Signal processing algorithm to get the displacement signal from the acceleration signal was
designed, implemented and successfully tested on MPS prototype. The RMS results were also
compared with the simulation results and error in results includes all errors due to digitization,
filtering, integration, rounding etc. The above algorithm will be further tested with actual signal from
accelerometers at field. The effort will also be made to increase the sampling frequency to reduce the
error due to integration.
REFERENCES
[1] Ribeiro J.G., Freire J.L. and de Castro J.T.; Some Comments on Digital Integration to Measure Displacements
using Accelerometers, Shock and Vibration Digest, v 32 n1, p.52,Jan. 2000.
[2] Oppenheim Alan V., Schafer Ronald W.; Discrete-Time Signal Processing, 2nd Ed., Prentice Hall.
[3] Sanjay K. Jain, M. Kalra, Gaurav and D. A. Roy; Machinery Protection System for Large Rotating Machines,
BARC Newsletter, p.23,Jan-Feb 2013.

S-ar putea să vă placă și