Mass Estimation Using a Loose Parts Monitoring System

Progress in Nuclear Energy, Vol. 36, No. 2, pp. 109-122, 2000

© 2000 Publishedby Elsevier ScienceLtd All rights reserved. Printed in Great Britain 0149-1970/00/$ - see front matter

Pergamon www.elsevier.com/locate/pnucene P I h S0149-1970(00)00008-1

Mass Estimation Using a Loose Parts Monitoring System

Jung-Soo Kim° and Joon Lyou* ° Korea Atomic Energy Research Institute MMIS Team P.O.Box. 105 Daeduk danji Taejon 305-606, Korea

Email: ki s(a3.nanum.kaeri.re.kr -

v

*Dept. of Electronics Engineering, Chungnam National Univ. 220, Gung-dong, Yusung-ku, Taejon 305-764, Korea

Abstract - This paper is concerned with estimating the mass of a loose part in the steam

generator of nuclear power plant. Although there is the basic principle known as "Hertz Theory" for estimating mass and energy of a spherical part impacted on an infinite fiat plate, the theory is not directly applicable because real plants do not comply with the underlying ideal assumptions. In this work, a practical method is developed based on the basic theory and considering amplitude and energy attenuation effects. Actually, the impact waves propagating along the plate to the sensor locations become modified in shape and frequency spectrum from the original waveform due to the plate and surrounding conditions, distance attenuation and damping loss. To show the validity of the present mass estimation algorithm, it has been applied to the mock-up impact test data and also to the real plant data. The results show better performance comparing to the conventional Hertz schemes. © 2000 Published by Elsevier Science Ltd.

All rights reserved.

K E Y W O R D S : LPMS, Diagnosis, Mass Estimation, Impact Theory

109

J.-S. Kim and J. Lyou

II0

1.1ntroduction LPMS (Loose Parts Monitoring System)

is

a diagnostic system that monitors the

integrity of Nuclear Steam Supply System (NSSS) and analyzes the impact event caused by moving or loose parts. This system provides the necessary information for the operator's proper decision to maintain a reliable and safe nuclear power plant. The loose pans (metal pieces) are produced by being parted from the structure of the reactor coolant system (RCS) due to corrosion, fatigue, and friction of components and also by coming into RCS from the outside during the period of reactor test operation, refueling, and overhaul. These loose parts ,are mixed with reactor coolant fluid, move with high velocity along RCS circuits, and generate collisions with RCS components. When a loose pan strikes against the components within the pressure boundary, an acoustic impact wave is generated and propagates along the pressure boundary. In order to detect the impact signal, conventional LPMS uses the accelerometer sensor installed on the outer surface of the pressure boundary of RCS and announces an alarm when the detected impact signal exceeds a pre-determined level. The sensors are usually installed on the probable places where loose pans may collect or exist such as the upper head of the reactor pressure vessel, hot chamber of the Steam Generator (Mayo,1988). See Fig. 1, where the sensor locations are marked with small black rectangles. In the existing LPMS, if the alarm is triggered, the detected signal is recorded on magnetic tape. Later, experienced operators analyze the recorded data and determine whether the detected signal is an impact signal by a loose pan or a noise signal. If it is decided that loose pans caused the signal, they evaluate the characteristic parameters such as impact location, energy, and mass. After the diagnosis process mentioned above is completed, the proper procedure required for maintaining the safe and reliable operation is performed. For use of the conventional diagnostic method, the operators should have expert knowledge and moreover, it takes a long time to analyze the detected signal data. Possibly, fatal damage of components may occur during the analysis procedure. Therefore, it is very desirable that if the alarm is triggered by a loose part's impact, the detected signal is stored in the computer memory, the automatic diagnosis procedure is activated immediately, and the diagnostic results (such as location, mass, and energy of loose parts) are displayed on the operator's monitors.

Loose parts monitoring system

111

enerator/ ~il

i i i I1~

p~

Sensor

Fig. 1. The typical sensor locations Generally, two methods for mass estimation have been known: one is time series analysis and the other is frequency analysis. In the time series analysis method, Hertz theory is used to determine the loose parts impact signal model and the equations for plate wave propagation are derived (Thie,1981 & Mayo,1985 & O1ma,1985). But the theory is not directly applicable to real plants because of violations of the basic assumptions. For instance, the structure of a steam generator consists of two parts; the side is of cylindrical shape and upper & lower parts of hemisphere shape. The impact source is not the solid sphere. On the other hand, for the frequency analysis method (Tsunoda,1985 & Olma and Schutz,1987 & Mayo, 1999), we should first change the impact (time) data to frequency data using Fast Fourier Transformations (FFT). From the frequency spectrum, we then find the characteristic frequency and estimate the mass referring to the look-up table. Fig. 2 shows a sample of converted frequency spectrum. In the figure, we can not exactly distinguish the characteristic frequency from those of sensor resonance and background noise. In this study, an automatic estimation algorithm of the mass of a loose part is developed using time domain data and modifying the basic theory in a practical way. The present scheme is applied to the impact data of mock-up facility and real steam generator, and the experimental results demonstrate that it performs better than the conventional one.

112

J.-S. Kim and J. Lyou Auto

°'"5'sl o.,ooo

!I!

03000 I"

i

s#ectrum

,~ 0. 2000 .:4

1

J

0.1000--

--0.0110 0

.

.

. 4000. Frequenc~ ¢Hz~000 ! :755X+ 755X+ !

12800

Fig.2 Example of the frequency spectrum of an impact signal

2. Hertz Impact Theory The Hertz theory describes the impact of a solid sphere on metal infinite plate. "['his theory is based on the assumption that the principal response of the plate is a bending wave of half period equal to the duration of the impact. Experimental results support this theory when the diameter of the sphere is not large comparing to the thickness of the plate and when the impact velocity is sufficiently small to avoid plastic deformation (Mayo, 1985). Fig. 3 displays the sinusoidal waveform of the impact signal, when the clash test is performed. D(t)

Steel

Ball

E~,w,m,l~

{

~r

Steel Plate

L:i,~ Fig.3. Clash test and the impact waveform.

Loose parts monitoring system

113

Under the assumption of sinusoidal waveform, the maxhnum amplitude (Dmax) of the displacemer~t and the contacting time (Td) of the solid Sphere are given by: Dm..~ = K h (mVo2)°4 R-°2 T d

-=

D ,n~x V o

2 .94

K~ =

(1)

(2)

,_...i]

~

~"~=-

o.,

(1.a)

where m is the mass of the sphere, R is the radius of the sphere', v) and Elate Poisson's ratio and Young's Modulus for the plate, v 2 and E2 are Poisson's ratio and Young's Modulus for the sphere: and V0 is the initial velocity of the sphere. In (2), it is noted that the Hertz theory defines the relationship between V0 and Td. Elastic or neat-elastic impacts between metal objects and a metal plate are generally characterized by the contact force trajectory which is very close to the half-sine function (Mayo, 1985). Also, it is well known that the acceleration of the impacting object during contact is proportional to the applied force through the Newton's second law of motion. This relationship and the maximum displacement predicted by Hertz theory, in turn, lead to the resultant equations of motion of the object, where D(t), V(t) and A(t) denote the displacement, the velocity and the acceleration, respectively.

D(t)=

r(t)=

A(t)=-

x sin

"~-'d Dmax cos

=~a

Dm~, sin

(3)

~

t

(4)

()

(5)

-~--at

Then, using (5), the contact force is obtained as F(t) = m A(t) 0< t < Td From (4), the maximum velocity would be V m a x -~- V ( 0 )

= V 0=

7~/Td D . . . .

(6) (7)

That is, the contact time Td is given by Ta= ~ Dmax/ V0 (8) Comparing (2) with (8), (2) differs from (8) by a factor of n/2.94(M.6). Hence, in

114

J.-S. Kim and J. Lyou

accordance with the half-sine

interpretation,the Hertz

theory based calculation needs to

be adjusted by the factor. 3. C o n s i d e r a t i o n o f the a t t e n u a t i o n e f f e c t s ,

3. ! Wave propagation along the plate The solution for two dimensional waves propagating from a locafized force we/e given by Lamb(Lamb, 1920 & Mayo, 1999) and after amplitiid6 n6rrnalization;:.it c a n b e used to calculate the plate surface acce!eratiofi w~t~,e Sh~tpe as a function of impacting object mass and velocity. However, it is very difficult to do due to nonlinear nature o f the equation. An alternative easy way is to assume that the acceleration wave"consists primarily of a few half periods at the frequency 1.6-times mo~e than that of the observed actual signals. A n d the magnitude o f this signal is set to the plate acceleration waveforrn. The force and the acceleration at the sphere's point are related by Aball=Fmax/m=

khlm

-0.4V01.3 R 0.2

(9)

Redefining into the plate's point, (9) becomes Apla,o = Fmax/Mefr (10) where Fmaxis the maximum impact force and Morris an effective mass of the plate. Note that the same force term Fm~ is used in (9) and (10), while the mass term, m, in (9) is replaced by Mcff in (10). The effective mass of the plate volume responding during the contact time is given by Me~ = n (CbTd) 2 h Pstcd Cb = Cu ( 1.8hfa / CH + 4.5hf~)°5

(10,a) (10.b)

where Cb is the phase velocity of a bending wave in steel plate, f~ is the frequency of interest in Hz, CL~ is 5,270 rn/sec, Td is contact time, h is the plate thickness and p steelis the density o f the plate. See Fig. 4.

? Fig.4. The effective mass of the plate

Loose parts monitoring system

115

3.2 Distance attenuation and Damping loss In (10), we calculate the mass estimation for the case that the impact point is equal to the installed sensor position. But, in general, since t h e sensor position does not coincide with the impact point, it is necessary to compensate for damping and distance attenuation. A transverse impact against a plate excites a cylindrical wave whose radius increases with time ms the wave propagates away from the' impact point. The amplitude o f this wave decreases as a function o f the distance due to the i ficreasing area covered by the wave and energy losses, referred to as damping. I'he distance attenuation occurs when the impact signal i s propagating from the impact point to the sensor location. The characteristic o f the impact "signal can be expressed as many values o f kr (k is the wave ntunber and ~r is the distance) with the asymptotic approximations with the Hankel function..Eqn.

(11) expresses the

relationship between the distance and the amplitude. The decrease in amplitude o f a plate's bending wave is proportional to an increase in distance from the impact source.

O(r)--

,kr)]

(11)

where D(r) is the wave amplitude at the distance r, Do is the amplitude at the point of impact, k is the wave number 2•f/Do, H0 is the Hankel function o f the 2 nd kind, Ho(kr) = 2j/n In (kr) {Ikrl 1 }. Impact wave attenuation during propagation also occurs due to the internal energy dissipation along the plate and the radiation energy to the surrounding fluids. Eqn. (12) expresses this type of loss for bending waves.

D

(r)

=

D

0e

(12)

where Do is the initial amplitude, r I is the loss factor defined in (12.a), f is the frequency o f interest, 7 is the distance traveled and Cg is the bending wave group velocity given by (12.b)

rl =

C

tO oCo 2zfpsh

=

x

M ~A M ~[- 1

(12.a)

3 • 6 C t , 2Ih [

c b ,(c L, +

9 h

)

(12.b)

116

J.-S. Kim and J. Lyou

4. E x p e r i m e n t a l

Results and Discussions

4.1 M o c k - u p e x p e r i m e n t For the laboratory level test purpose, we first gathered the impact signals from the mock-up system. Fig. 5 shows the reactor mockup made of stainless steel whose size is scaled down to 1/7 and thickness is 10mm. Main components installed are structure supports,

feed-water and

drain

components,

and

accelerometers.

18

sensors

(accelerometers) are attached on the side at intervals of 60 °; 14 sensors are attached on the upper head, and 14 sensors on the lower head. Fig. 6 display the input parameter values and estimated results., The input parameters are the distance between sensor and impact position, the initial half period of impact wave and the maximum amplitude at initial period.

! )t •I

I

Fig.5. Reactor Mockup

Fig. 6. Display Window

The developed mass estimation algorithm is summarized in Fig. 7. From the flowchart, we know that the basic algorithm of Hertz theory is modified by revising it experimentally. In Fig. 7, the basic algorithm is only one way at the modification part. That is, the basic algorithm only assumes that the plate is infinite and the impact signal is sinusoidal waveform. So the basic algorithm only calculate eqn (8). But, practically the real plant structure is finite and the structure shape is cylindrical and hemisphere. Through the many tests, three half period branches are set to adjust the weighting factor according to a scope o f the initial half period. Table 1 demonstrates the mass estimation results after running the developed algorithm in the cases where there are impacts at 30cm, 70 cm, 110 cm away from the sensor using some impact ball at the height of 6

Loose parts monitoring system

117

cm. Tests were conducted three times for each case, and the average estimation error rate was about 16 %.

Table 1. Test results (actual mass : 30.5g and diameter : 1.9cm) Distance-

Amplitude

Half

Mass

frequency

(acceleration)

period(~ts)

(g)

Diameter (cm)

30-1

0.4718

40

36.75

2.0746

30-2

0.4172

40

33.24

2.0065

30-3

0.5105

39

39.85

2.1314

70-1

0.3182

37

34.11

2.0238

70-2

0.3417

37

36.15

2.0634

70-3

0.3480

37

36.70

2.0737

110-1

0.2742

34

36.00

2.0605

110-2

0.2307

34

31.30

1.9666

110-3

0.2684

34

35.38

2.0485

4.2 Application to the real impact data In the figure 8 and 9, the Nuclear Power Plant (NPP) composed of three Steam Generators(S/G) are schematically diagramed, and the sensor numbers are denoted. Two accelerometers are installed in S/G; one at the primary side and the other at the secondary side. The radius of steam generator is 2.115m and the normal flow velocity was 7.8 m/see and the flow velocity at hot chamber was 1.23 m/see. The distance between the primary and the secondary is 1.5m apart. Figs. 10 & 11 show the recorded impact signals from the S/G accelerometers. We observe that the impact bursts show up on the two points 755 and 757, and 754 is affected by the impact signal of 755. Fig. 11 enlarges a portion of Fig. 10 on the extended time scale. And Figs. 12 emphasizes the burst signals of Fig. 11. Based on these data, we can estimate the impact location (Kim, 1998 and Kim, 1999 and Rhee, 1997), and hence the distance between the sensor and impact position. The other input parameters are obtained from the acceleration waveform as shown in Fig. 13. The estimated values are as follows: the distance is about 1.58m, the maximum amplitude about 139. 763g (g = 9.8m/sec2), and the initial half period about 75.37~tsec. Now, using the input parameters, the developed algorithm guesses the mass about 500 - 600 gram. The impact events occurred were three. Table 2 shows the results of our analysis are nearly the same as those of the expert signal

118

J.-S. Kim and J. Lyou

analysis-company. Even if the estimated results are more or less different from the actual value, our results are thought to be good enough taking the actual environment in account (different background noise level, the impact source is not a sphere and so on.). For the first events, it was found the impact objects were bolts and nuts located at two different places. Note that the estimation results seem less accurate than the latter two cases because of multiple impact sources.

(

Start

Modification part

~

Input parameter

/

-- A m p l i t u d e ]-- h a l f p e r i o d L- D i s t a n c e

i............................

.....................

..................

...............

0.00005=< hal f~vperiod O.O ~ ~

-0.1 -0.2 -0.3

0.2527

-0.4 -0.5

Time(sec)

Fig. 13 The initial half period and maximum amplitude

(g gram) Table 2. Comparison o f the estimated and actual masses Event DatePresent Foreign company Actual Number o f Mass Frequency estimation estimation result impact object result 1997 - 1 A: 1.5 Pound 755 • about 200g A:500-600 755:3 B "1 pound 757 : about 100g 757:2 B:300-400g 1999-1 1999- 2

39 -52 g 75 - 82 g

N/A N/A

N/A 76.6 g

5. Concluding Remarks In the conventional LPMS, the operators should have had much knowledge for analyzing the impact signal and mass estimation. In this work, we have developed an automatic mass estimation algorithm and applied it to the reactor mockup and the actual plant impact signal. In the case o f the mockup tests, the average error rate between the estimated value and actual value was relatively small due to noise-free environment. While, the error rates for the real signal case increased mainly due to the distance attenuation and damping loss effects. In applications point of view, the developed algorithm is expected provide the operators with more accurate estimation results contributing to the safety and the prevention o f accidents in a NPP. Further studies of immediate interest are to introduce the artificial intelligent techniques such as neural network or wavelet transformation.

122

J.-S. Kim and J. Lyou

References [1] Mayo C. W. et al (1988), Loose Parts Monitoring System Improvements, EPRI NP5743, Final Report. [2] Joseph A. Thie (1981), Power Reactor Noise, American Nuclear Society(ANS). [3] Charles W. Mayo (1983), "Loose part signal theory", Progress in Nuclear Energy, Vol. 15, pp. 535-543. [4] B. J. Olma (1985)," Source Location and mass Estimation in Loose Parts Monitoring of mass estimation algorithm LWRs." Progress in Nuclear Energy, Vol. 15. [5] T. Tsunoda, et al (1985), " Studies cn the Loose part Evaluation Technique", Progress in Nuclear Energy, Vol. 15. [6] B. J. Olma and B. Schutz (1987), " Advanced Burst Processing Methods in Loose Parts Monitoring", SMORN V, Progress in Nuclear Energy, Vol. 21, pp. 525-535. [7] Mayo (1999), "Loose Part Mass and Energy Estimation", Progress in Nuclear Energy, Vol.34, pp. 263-282. [7] H.Lamb (1924),"On Wave-propagation in Two Dimensions", Proceedings of the London Mathematical Society, Vol. 35,1920,pp.141-161. [8] J.S.Kim and J.Lyou (1998), "An Automatic Diagnosis Method for Impact Location Estimation". Proceeding of KACC, International Session, pp. 295-300. [9] I.K.Rhee, et a1(1997), Development of Loose Part Signal Location Estimating Technique in High Pressured Structure, KAERI/CM- 146/96,1997. [10] J.S.Kim, et a1(1999), "A mass estimation algorithm for Loose Parts using Hertz Theory", Proceeding of the Korean Nuclear Society, Abstract pp. 75, Pohang, May 1999.