You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
<strong>atw</strong> Vol. 62 (<strong>2017</strong>) | Issue 6 ı June<br />
RESEARCH AND INNOVATION 410<br />
| | Fig. 2.<br />
Nodalization for the primary system of AP1000.<br />
compared with the results of<br />
LOFTRAN, such as pressure, flow rate<br />
etc. The comparison results show that<br />
RELAP5 has the capability to predict<br />
the system parameters [11] correctly.<br />
Among above these parameters, the<br />
reactor coolant temperature is the<br />
most important parameter for the<br />
PRHRS loop, as shown in Figure 3.<br />
As depicted in Fig. 3, the results of<br />
RELAP5 and LOFTRAN exhibit the<br />
same trend. A similar value of maximum<br />
temperature has been observed<br />
in the two results. During the accident,<br />
however, a difference in the sequence<br />
of events and response of the reactor<br />
control system can lead to a slight<br />
difference in the temperature trend.<br />
This has no significant influence on<br />
the analysis.<br />
3 Calculation methods<br />
3.1 Latin hypercube sampling<br />
Latin Hypercube Sampling (LHS) [16,<br />
17] is an improvement over the traditional<br />
Monte Carlo sampling method.<br />
It can overcome its drawbacks, in that<br />
most of sampled results lie near the<br />
average value. In order to improve the<br />
accuracy of the parameters, therefore,<br />
the method covers the upper and<br />
lower limits of the distributions.<br />
Hence, this method has the ability to<br />
determine any value, as long as the<br />
parameters are known. The steps are<br />
as follows.<br />
(1) For each variable, the probability<br />
distribution is divided into N nonoverlapping<br />
equal probability<br />
interval [0, 1/N], [1/N, 2/N],…,<br />
[(N-1)/N, 1]. This ensures that<br />
the degree of the correlation of<br />
LHS is small.<br />
(2) The random standard normal<br />
sample matrix Z N×n is used to<br />
represent the order of sample<br />
points, and the integer matrix<br />
R N×n is used to record information<br />
regarding the ordering of the<br />
above sample points. Hence,<br />
R ij = k shows that the sequence<br />
of the j th variable in the i th sampling<br />
is k.<br />
(3) According to in each interval, the<br />
cumulative probability function<br />
of each sample point in the Latin<br />
hypercube can be obtained randomly,<br />
as shown in Eq. (1).<br />
(1)<br />
Here, i = 1,..., N and j = 1,..., n.<br />
The function r a n d (0,1) represents<br />
a random number, which is<br />
uniformly distributed in the [0,1]<br />
interval.<br />
(4) In the Latin hypercube, the sampling<br />
points are obtained by the<br />
method of equal probability<br />
change, as shown in Eq. (2).<br />
(2)<br />
Here, φ –1 (.) is the inverse normal<br />
distribution function.<br />
3.2 Grey Relation Method<br />
The Grey Relational Method [8] is a<br />
quantitative technique for comparative<br />
analysis. The basic idea is to determine<br />
the exponent of each factor in<br />
the correlation, according to their<br />
degree of similarity with the geometry<br />
of sequence curve. If the curve is close<br />
for a particular factor, it would have a<br />
high exponent in the correlation. X 0 is<br />
defined as the target parameter, with<br />
k referring to the sequence of the<br />
parameter, denoted as {X 0 (k)}. It is<br />
assumed that there are a total of m<br />
control parameters, and a parameter j<br />
in the same sequence k is called the<br />
comparison sequence, denoted as<br />
{X j (k)} (k = 1,…, N)(j = 1,…, m). In<br />
this correlation, the exponent of each<br />
factor can be obtained by comparing<br />
the tendency of development between<br />
the target and influence parameters.<br />
These steps are shown as follows [18].<br />
(1) The reference and comparison sequences<br />
are normalized, as shown<br />
in Eq. (3).<br />
(3)<br />
(2) The absolute value of difference<br />
between the reference and the<br />
comparison sequences is calculated<br />
as Eq. (4) based on above normalization<br />
results.<br />
(4)<br />
a) Results of RELAP5. b) Results of LOFTRAN.<br />
| | Fig. 3.<br />
Reactor coolant temperature.<br />
(3) Maximum and minimum absolute<br />
values are calculated shown as<br />
Eq. (5)-(6).<br />
(5)<br />
(6)<br />
Research and Innovation<br />
Reliability Analysis on Passive Residual Heat Removal of AP1000 Based on Grey Model ı Qi Shi, Zhou Tao, Muhammad Ali Shahzad, Li Yu and Jiang Guangming