ASSIGNMENT 2
RESPONSE SENSITIVITY ANALYSIS
DIKSHA TIWARI (S5210086)
RASHID KHAN (S5152104)
MALAV PATEL (S5672581)
For this exercise we will apply Sensitivity Response Analysis (RSA), estimating the effect of several input
parameters on the heat exchanger, and compare the result with a Monte Carlo (MC) method by using
MATLAB.
Assignment ADDMS [Link]
Stochastic parameters:
Group ṁhot [kg/s] ṁcold [kg/s] Thot [K] Tcold [K] Heat exchanger type
6 (μ=590, σ=2) 135 (μ=525, σ=2) (μ=310, σ=1) One shell pass
Parameter Unit Value
2
U [W/m K] 20 · 103
A [m2] 10
cp,hot [J/kg K] 1.005 · 103
cp,cold [J/kg K] 4.180 · 103
Task 1
In the assignment, we should estimate the mean value, standard deviation, and PDF of the input data pertaining
to a heat exchanger with 10^5 samples.
To get the output data, we have written the code of the output function (heat exchanger), that consists of the
input data mass flow of cold and hot fluid and temperature of hot flow as the output. In our case we are
looking at a One shell Pass.
First input parameter is mass flow rate of the HOT liquid, which has mean value =590, sigma =2. Second
input HOT fluid inlet temperatures value is mean= 525 and sigma=2 and third output parameters are COLD
fluid inlet temperatures, which have mean value =310, sigma=1.
Run a MC simulation with 105 samples to estimate the transferred heat Q. Evaluate mean μQ and
standard deviation σQ of Q and plot the corresponding PDF.
The PDF below shows the average value (µ) and standard deviation (σ) related to the heat exchanger,
derived from 10^5 samples:
Figure 1: The PDF of the heat exchanger for 10^5
samples MC simulation
Consistent with our program output, the output parameter, the heat exchanger, which has a mean value =
1.7808e+04, sigma = 185.4443, Q Exchanger = 1.754942418595840e+04 was taken from the normal
distribution.
From Figure 1, we can see that the heat distribution is a normal distribution.
Task 2
Find a function that fits the PDF well and report its characteristic parameters.
The different function that fits well the PDF is clearly shown from the Figure 2 are:-
Normal
Burr
Log_Logistic
Figure 2: The PDF of the heat exchanger for 10^5
samples with three different functions
Comparison Table of Normal, Log-Logistic and Burr Distribution on the basis of their
characteristics
Distribution Normal Log-Logistic Burr
-664169 -665097
Log Likelihood -2565.96
-Inf < y < Inf 0 < y < Inf
Domain 0 < y < Inf
17807.7 17806.4
Mean 47.5852
34389.6 36867
Variance 10.5693
17834.1 2.49562
Mu 17807.7 0.586426 3.8602 0.0020757 Alpha
Paramete
r Estimate c 159.138 0.826557
Sigm
Std. Err. 185.444 0.414669 0.037562 0.000984494
a k 1.17549 0.0167304
mu sigma mu sigma Alpha c c
A
Estimated l
mu 0.343896 -3.26656e-16 mu 4.30855e-06 -1.99885e-08 p 6.22813 -1.74965 0.0405755
covariance of h
parameter a
estimates c -1.74965 0.683197 -0.0119768
sig
sigma -3.26656e-16 0.171951 -1.99885e-08 9.69228e-07 0.000279906
ma k 0.0405755 -0.0119768
Figure 3: The PDF of the heat exchanger for 10^5
samples with Normal Function that fits the PDF well
Task 3
The goal of the second task is to RSA with the following steps [0.01, 0.1, 1, 10].
Heat exchanger sensitivity for 0.01 step.
sigma_Q_exchanger 0.01 185.200340261369
m_Q_exchanger 0.01 17807.0845886259
Heat exchanger sensitivity for 0.1 step
sigma_Q_exchanger 0.1 185.200344092673
m_Q_exchanger 0.1 1.780708459155763e+04
Heat exchanger sensitivity for 1 step
sigma_Q_exchanger for 1 NaN + NaNi
m_Q_exchanger for 1 NaN + NaNi
Heat exchanger sensitivity for 10 step
sigma_Q_exchanger 1.846465157144084e+02 - 2.779654550936633e+00i
for 10
m_Q_exchanger for 10 1.780711014683805e+04 - 9.619723057998159e-03i
Since for step 1 the mean and sigma of heat exchanger is NaN and for step 10 the values are in complex so
both this step compromises the consistency of the results and are replaced.
Since 0.01 and 0.1 steps have no such values that cause inconsistency or compromises the result.
Task 4
In the fourth task you must calculate the mean and standard deviation for each step, plot the PDF, evaluate
the result, and compare with the MC which steps give more accurate results. Method RSA is based its
calculation algorithm on the knowledge of the properties of the main probability input variables (mean and
standard deviation).
This study evaluated the output heat exchanger variables using a first order and second order formula and
sensitivity formula
As we have already seen the mean and standard deviation for each step, in Task 3.
Now plotting the PDF of each and every step:-
1. For step 0.01
Figure 4: PDF for Step 0.01
2. For step 0.1
Figure 5: PDF for Step 0.1
3. For step 10
Figure 4: PDF for Step 10
We can see from the results of the mean and standard deviation of the output variable (heat exchanger) that the
results are similar for each step in the RSA method, except for step 1 in which there is no result of the mean and
standard deviation of the heat exchanger and there is no PDF. Step 10 shows complex results of mean and
standard deviation, therefore step 10 is not acceptable. whereas 0.01 step would be too small to compromise the
consistency of the result and therefore is replaced by step 0.1.
0.1 steps produce more accurate results compared to MC. As mean and standard deviation is very close when
compared with Monte Carlo.
Task 5
Referring only to the optimal value of δ0 found in #4, perform a sensitivity analysis to understand how
different inputs affect the output of the system (i.e., the transferred heat Q). Plot the non-dimensional
sensitivity for each input and comment on the results.
Task is to perform a sensitivity analysis to understand which parameters have the strongest impact on the
performance of the heat exchanger
However, MCs doesn’t allow to evaluate which input has the strongest impact on the monitored outputs
Therefore, Sensitivity analysis should be performed.
The goal of the fourth task is to analyze the results of the sensitivity analysis by plotting them on an
appropriately plotted graph.
1. For step 0.01
Sensitivity of q exchanger for 0.01
-0.00925049552210422. 2.44186046511634 -1.44186046511633
nominal_out_cost
1.780710272356003e+04
2. For step 0.1
sensitivity_Q_exchanger for 0.1
-0.00927152304964224 2.44186046511628 -1.44186046511629
nominal_out_cost
1.780710272356003e+04
As clearly seen from the figure above of Step 0.1 (which has been preferred
over other steps from task 4)
As we can see from the figure above that Parameter m HOT and T_hot_in has a negative correlation whereas
T_cold_in has a positive correlation with the heat exchanger.
T_COLD_IN has the highest impact on the heat exchanger.
Pressure drop on hot has a very small impact on heat exchanger Pressure drop on Temperature hot has higher
impact on heat exchanger.
3. For step 1:
sensitivity_Q_exchanger for 1
NaN + NaNi 2.44186046511628 + 0.00000000000000i -1.44186046511628 + 0.00000000000000i
nominal_out_cost
1.780710272356003e+04
4. For step 10:
sensitivity_Q_exchanger for 10
-0.338696218972996 + 0.415890594300286i 2.44186046511628+0.00000000000000i 1.44186046511628 + 0.00000000000000i
nominal_out_cost
1.780710272356003e+04
Comparison Table of Sensitivity result for each input parameter and step
Steps Mass Flow Rate of hot Fluid Mass Flow Rate of cold inlet Temperature Hot fluid inlet temperature
0.01 -0.00925049552210422 2.44186046511634 -1.44186046511633
0.1 2.44186046511628 -1.44186046511629
-0.00927152304964224
-1.44186046511628 +
1 NaN + NaNi 2.44186046511628 + 0.00000000000000i
0.00000000000000i
-0.338696218972996 + -1.44186046511628 +
10 2.44186046511628 + 0.00000000000000i
0.415890594300286i 0.00000000000000i
The table shows that the sensitivity of the heat exchanger is not so much different when compared the
RSA method 0.1, and the step 0.01 data are relatively like each other.
In each step there is some relation between the parameters and the heat exchanger but in step 1, parameter
m_hot shows no relation with the heat exchanger. Whereas step 10 in this case shows the clear and strongest
relation when compared with the other steps.
But in order to avoid any inconsistency of the result we have to replace step 10 and step 1 with 0.1 step which
has same sensitivity as of step 0.01.