Sunteți pe pagina 1din 42

MASTER'S THESIS

HIGH TEMPERATURE
MEASUREMENTS AT THE INTERNAL
NOZZLE WALL OF THE ZEPHYR

Manohar Karnal
2014

Master of Science (120 credits)


Space Engineering - Space Master

Luleå University of Technology


Department of Computer Science, Electrical and Space Engineering
HIGH TEMPERATURE MEASUREMENTS AT THE INTERNAL
NOZZLE WALL OF THE ZEPHYR
by

Manohar Karnal

A thesis submitted in partial fulfilment of the requirements for the degree of


Master of Science

Supervisor
Dr. Peter Rickmers
ZARM Institute, Germany

Examiners
Peter Von Ballmoos
Université Paul Sabatier Toulouse III, France

Mathias Milz
Luleå University of Technology, Sweden

September 2014
Université Paul Sabatier Toulouse III, France
A BSTRACT

In this thesis the heat transfer theory is studied and applied to develop a theoretical model
which can be used to estimate exhaust gas temperature of ZEpHyR main engine developed at
ZARM institute. The conjugate gradient method with adjoint problem for function estimation
iterative technique is used to solve the Inverse Heat Conduction Problem (IHCP) to estimate
heat flux and internal wall temperature of throat section of the nozzle. The convective heat
transfer coefficient is calculated using Bartz equation. The exhaust gas temperature is
determined using estimated heat flux, internal wall temperature and heat transfer coefficient.
The model is verified by comparing the estimated flux with the results from a published
research. The nozzle is made of Molybdenum material and the exhaust gas temperature is
determined to check material’s capability to withstand high gas temperature.

Ce stage avait pour but d’étudier la théorie des transferts de chaleur et de l’appliquer pour
développer un modèle théorique pouvant être utilisée dans le but d’estimer les températures
des gaz d’échappement du moteur principal ZEpHyR développé à l’institut ZARM. Afin de
la mettre en œuvre, on fait appel à une technique d’estimation itérative dite “ conjugate
gradient method with adjoint problem », cette est minimisation technique pour fonction
estimation. On l’utilise ici pour résoudre le problème de conduction de chaleur inverse pour
estimer les flux de chaleur et la température de la paroi interne au niveau de la section de
l’ouverture de la buse.

Le coefficient de transfert de chaleur convectif est calculé en utilisant l’équation de Bartz.


La température des gaz d’échappement est déterminée en mettant en relation une estimation
des flux de chaleur, la température interne des parois et le coefficient de transfert de chaleur.
Ce modèle est validé en comparant les flux estimés par cette méthode avec les résultats
renseignés dans une publication faisant suite à une recherche.

Il est intéressant de voir comment une buse faite en Molybdène peut supporter des
températures élevées de gaz, ce qui est possible après détermination des températures de gaz
d’échappement du moteur.

i
A CKNOWLEDGEMENTS

First of all, I would like to thank my parents for their support and prayers. My sincere
gratitude goes to my supervisor Dr. Peter Rickmers for providing the opportunity, guidance
and support during the thesis. I would also like to thank Prof. Peter von Ballmoos and Prof.
Mathias Milz for their support throughout the project.

Throughout my two years in the Space Masters program, I am thankful to Anette Snällfot
Brändström and Maria Winnebäck for taking care of the administrative issues. I would like
to thank all my friends especially Beenish Batul and Nader Samir for their support.

ii
C ONTENTS
I. Introduction ..................................................................................................................... 1

A. Motivation ............................................................................................................... 1

B. Objectives ................................................................................................................ 2

II. Theoretical Background .............................................................................................. 3

A. Nozzle Theory ......................................................................................................... 3

B. Theory of Inverse Heat Conduction Problem (IHCP) ............................................. 3

1) History.................................................................................................................. 4

C. Constraints of IHCP................................................................................................. 5

III. Theoretical Model ....................................................................................................... 6

A. Calculation of convective heat transfer coefficient ................................................. 8

IV. Solution of IHCP ......................................................................................................... 9

A. Conjugate Gradient Method (CGM) with adjoint problem for function estimation9

1) The direct problem ............................................................................................... 9

2) The inverse problem ............................................................................................ 9

3) The sensitivity problem...................................................................................... 10

4) The adjoint problem ........................................................................................... 11

5) The gradient equation ........................................................................................ 11

6) The iterative procedure ...................................................................................... 11

7) The stopping criteria .......................................................................................... 12

B. Computational algorithm ....................................................................................... 13

V. Model Verification .................................................................................................... 14

VI. Results and Discussion .............................................................................................. 19

A. Estimation of convective heat transfer coefficient and exhaust gas temperature .. 23

iii
VII. Conclusion and Future Work................................................................................. 26

A. Conclusion ............................................................................................................. 26

B. Future Work ........................................................................................................... 26

References ........................................................................................................................... 27

Appendix ............................................................................................................................. 29

iv
L IST OF F IGURES

Figure 1 Rocket nozzle [4] .................................................................................................... 3

Figure 2 The model geometry ............................................................................................... 6

Figure 3 Measured temperature vs time .............................................................................. 16

Figure 4 Comparison of estimated heat flux results............................................................ 16

Figure 5 Comparison of estimated heat flux for different time lags ................................... 17

Figure 6 Comparison of errors in heat flux for different time lags ..................................... 18

Figure 7 Comparison of test heat flux and 3 σ heat flux ..................................................... 20

Figure 8 Comparison of test heat flux and 5 σ heat flux ..................................................... 21

Figure 9 Comparison of test heat flux and 10 σ heat flux ................................................... 21

L IST OF T ABLES

Table I Material properties of Mild Steel [17] .................................................................... 14

Table II Time vs Measured temperature [16] ..................................................................... 15

Table III Properties of Molybdenum [3] ............................................................................. 19

Table IV Estimated heat flux and internal wall temperature .............................................. 22

Table V Properties and transport parameters of combustion gas ........................................ 23

Table VI Estimated Heat transfer coefficient and exhaust gas temperature ....................... 24

v
N OMENCLATURE

L Thickness [m]

x Space coordinate [m]

∆x Node thickness [m]

D Diameter [m]

r Throat radius of curvature [m]

A Area [m2]

t Time [s]

∆t Time difference [s]

T Temperature [K]

To Initial temperature [K]

Y Measured temperature [K]

q Heat flux [W m-2]

h Convective heat transfer coefficient [W m-2 K-1]

k Thermal conductivity [W m-1 K-1]

ρ Density [kg m-3]

Cp Specific heat at constant pressure [J kg-1 K-1]

μ Viscosity [Pa s]

vi
Pc Chamber pressure [bar]

c* Characteristic velocity [m s-1]

g Gravitational acceleration [m s-2]

Pr Prandtl number

γ Specific heat ratio

M Mach number

ε Tolerance

σ Standard deviation

ω Random variable

Subscripts

0,1.. Node identifier

w Wall surface

g Gas

s Stagnation

t Throat

f Final value

e Exact

vii
High temperature measurements at the internal nozzle wall of the ZEpHyR

I. INTRODUCTION
In past couple of decades, the development of the hybrid rockets has increased. The major
reason being hybrid rockets are much safer and easier to use than liquid and solid rockets in
terms of complexity and controllability [1]. Hybrid rockets generally have higher specific
impulse compared to solid rockets but lower than liquid rockets. In rocket theory, specific
impulse is the measure of rocket motor performance. It describes the thrust produced for the
amount of fuel burned. The stoichiometric oxidiser and fuel mixture ratio provides higher
specific impulse and much higher flame temperature is achieved in turn increasing the
exhaust gas temperature [2].

One of the factors significant to the success of the rocket engine depends on the nozzle.
The nozzle has to survive extreme temperature and pressure, where the combustion products
(exhaust gas temperature) entering the nozzle can reach up to 3500K and a pressure of up to
35 bar. Only a hand full of materials are present that can withstand these conditions without
active cooling. By using fuel rich mixture ratio, the flame temperature can be reduced but this
drags down the specific impulse of a rocket [2]. Hence it is important to know the exhaust gas
temperature which will help in finding the optimum oxidiser-fuel mixture ratio that will
deliver the highest engine performance and as well as maintain the exhaust gas temperature
within the melting point of material.

A. Motivation

At ZARM Institute – Univeristät Bremen, a hybrid rocket engine is being developed under
the framework of the DLR-STERN (Studentische Experimentalrakten) program by German
Aerospace Agency (DLR - Deutsches Zentrum für Luft und Raumfahrt). The ZEpHyR
(ZARM Experimental Hybrid Rocket) engine runs paraffin as solid fuel and liquid oxygen as
the oxidiser to power it to an altitude exceeding 12 km. Currently, the engineering model is
being integrated at the ZARM rocket engine test stand. The engine’s nozzle is made of the
Molybdenum material. The melting point of Molybdenum is 2896.15 K [3]. Hence it is
required to calculate the gas temperature and internal wall surface temperature in order to
maintain the temperature of the nozzle material within its melting point with active cooling if
required or use different material for future projects. In this research the oxidiser-fuel mixture
by thermochemistry with known gas temperature is not determined but it will be helpful for
further studies to improve the performance of the rocket engine.

1
High temperature measurements at the internal nozzle wall of the ZEpHyR

B. Objectives

The main objective of the thesis is to estimate the exhaust gas temperature on the inside of
the nozzle by using thermocouple data at the outer surface of the nozzle. The scope of the
thesis includes the study of heat transfer theory to establish one dimensional transient heat
transfer model which describes the processes involved. Using the above heat transfer model,
the internal wall surface temperature is calculated at throat section of nozzle which is then
used to approximate the exhaust gas temperature. The gas temperature obtained is used to
verify the suitability of Molybdenum material for the nozzle of ZEpHyR main engine.

2
High temperature measurements at the internal nozzle wall of the ZEpHyR

II. THEORETICAL BACKGROUND

A. Nozzle Theory

A nozzle is a convergent divergent tube used for propulsion in the rocket engine so that
the combustion gases exit at supersonic velocities. This is achieved by accelerating the
combustion gases that enter the convergent section of the rocket nozzle at subsonic velocities.
The accelerating gases enter the throat section at sonic velocities where the cross-sectional
area is the smallest. From the throat, the gases expand as they enter the divergent section of
the nozzle and exit at supersonic velocities [4]. The combustion gases enter the nozzle at very
high temperature. Therefore the material of the nozzle is chosen to withstand this high
temperature.

Figure 1 Rocket nozzle [4]

B. Theory of Inverse Heat Conduction Problem (IHCP)

Temperature distribution within a solid gives important information about its


characteristics and response to the external environment. This temperature distribution within
the solid can be calculated if the boundary conditions such as temperature or heat flux as
function of time are known. This type of problem is termed as direct problem. In most cases,
boundary condition is unknown and needs to be determined. This can be estimated by taking
temperature measurements at interior location of a solid. This type of problem is called
inverse problem or Inverse Heat Conduction Problem (IHCP).

3
High temperature measurements at the internal nozzle wall of the ZEpHyR

It is very difficult to determine the unknown boundary condition by direct measurements


in a rocket nozzle as it is subjected to very high temperatures. Many experimental problems
arise in implanting a thermocouple or heat flux sensor at the internal surface of the nozzle due
to extreme environmental conditions. Moreover implanting the probe at the surface disturbs
the boundary condition and the accuracy of the measurements. But it is easier to accurately
measure temperature at interior location or insulated surface of the nozzle material. Hence by
using inverse technique, unknown boundary condition can be estimated indirectly from the
measured temperature within the nozzle

1) History

Inverse heat problems have been the area of research for several decades. In 1950’s
methods to solve the IHCP were based more on pure intuition than on mathematical
formulation. One of the earliest publications dates back to 1960s where the heat transfer rates
were calculated during quenching of a body [8]. The inverse heat conduction problem gained
considerable importance at the start of the space age. The applications included design of
nose cones of missiles, rocket nozzles and other devices. James V Beck [6][7] developed a
basic concept on IHCP which allowed much smaller time steps than the Stolz method [5].
The testing of nuclear components was another research area that extensively required
solution of IHCP. Notable applications for IHCP included periodic heating in combustion
chambers in internal combustion engines, solidification of gases, indirect calorimetry for
laboratory use and transient boiling curve studies [8].

Different approaches have been used to solve inverse heat conduction problems.
Duhamel’s theorem can be used to solve linear IHC problems [8]. Burggraf published an
exact solution technique for IHCP in 1964 [9]. Similarly Imber and Khan, Langford and
others also proposed techniques for exact solutions. Limitation for the exact solution
techniques is that it does not consider variation in physical properties. To solve the nonlinear
inverse heat conduction problems many numerical methods have been employed such as
finite difference and finite element [10].

IHCP is mathematically ill-posed as the solution of the problem may become unstable due
to measurement errors. Numerous procedures have been developed and steady advancement
is being made to solve the ill-posed problems. A.N.Tikhonov, O.M.Alifanov, James.V.Beck
and Arsenin overcame the instabilities of the ill-posed problems. They devised methods
based on approximating the ill-posed inverse problem in terms of well-posed problem to
4
High temperature measurements at the internal nozzle wall of the ZEpHyR

solve the IHCP by utilizing regularization methods in order to reduce the sensitivity of ill-
posed problems to measurement errors. These methods are still used nowadays. Most of these
methods utilize the least squares sense to obtain the solution to IHCP. A.N.Tikhonov adds a
smoothing term to the least squares to reduce the instabilities due to measurement errors
[11][12]. In Alifanov’s iterative regularization methods, the number of iterations act as the
smoothing term as in Tikhonov’s method and the stopping criterion is chosen so that the
reasonable final solution is obtained [13].

C. Constraints of IHCP

In inverse heat conduction problems temperature measurements are assumed to be the


main source of error. This poses difficulty in solving the IHCP. The difficulties are
particularly prominent when large amount of information is retrieved from the available data.
Increasing the amount of information requires a reduction in time step at which calculations
are done that in return has an adverse effect on the stability of the solution unless some kind
of restrictions are used.

Time lag and damping are two significant physical constraints in transient heat
conduction problem. Time lag is the time taken for the thermal information to reach the
measurement site at certain distance. Damping, it is the decrease in responses at the
measurement sites relative to the rapid fluctuations at the surface. These two constraints
should be taken into account when solving inverse transient heat conduction problem [8].

5
High temperature measurements at the internal nozzle wall of the ZEpHyR

III. THEORETICAL MODEL


In general, the theoretical model represents the physical system and processes involved in
it and also it should be able to accurately predict the system behaviour. The theoretical model
is required in order to obtain the information that would be difficult to gather from the
physical system when subjected to certain conditions. As an example, when the space shuttle
vehicle’s surface is subjected to extreme high temperatures during the re-entry into earth’s
atmosphere it becomes impossible to directly measure the temperature or the heat flux on the
surface. However, it is much easier to measure the temperature at interior location of the
vehicle and input this partial information into the theoretical model to accurately predict the
temperatures or heat flux at the surface.

In this research heat transfer theory is applied to build the theoretical model of the nozzle
system. One dimensional transient heat transfer model is built considering the nozzle throat
as slab of finite thickness where it is convectively heated at one surface and perfectly
insulated at the other.

Consider a slab of thickness L and assume constant thermal properties. The adiabatic
condition is applied to x = L. At a specific time (t > 0), a heat flux q(t) is applied to x = 0. The
model geometry is given in Figure 2.

Figure 2 The model geometry

6
High temperature measurements at the internal nozzle wall of the ZEpHyR

The dimensional mathematical formulation of heat conduction problem is given:

𝜕 2𝑇 𝜕𝑇
𝑘 2 = 𝜌𝐶𝑝 0 ≤ 𝑥 ≤ 𝐿 𝑎𝑛𝑑 𝑡 > 0 (1.a)
𝜕𝑥 𝜕𝑡

𝑇(𝑥, 𝑡) = 𝑇𝑜 0 ≤ 𝑥 ≤ 𝐿 𝑎𝑛𝑑 𝑡 = 0 (1.b)

𝜕𝑇(𝑥, 𝑡)
=0 𝑥 = 𝐿 𝑎𝑛𝑑 𝑡 > 0 (1.c)
𝜕𝑥

𝜕𝑇(𝑥, 𝑡)
𝑘 = 𝑞(𝑥, 𝑡) 𝑥 = 0 𝑎𝑛𝑑 𝑡 > 0 (1.d)
𝜕𝑥

The above system is represented numerically in finite difference form by using the implicit
method (backward difference). Implicit method uses both the current and future state of the
system to solve the equation. The finite difference method of the transient heat transfer
equation for the model is written for ‘n’ number of nodes and current time ‘i’. The equations
are represented in tri-diagonal form as follows.

For the 1st node

1 1 2∆𝑥𝑞(𝑡)
�−2 − � 𝑇0𝑖+1 + 2𝑇1𝑖+1 = − 𝑇0𝑖 − (2.a)
𝜏 𝜏 𝑘

For (n-1) node

𝑖+1
1 𝑖+1 1 𝑖
𝑇𝑛−2 + �−2 − � 𝑇𝑛−1 + 𝑇𝑛𝑖+1 = − 𝑇𝑛−1 (2.b)
𝜏 𝜏

For nth node

𝑖+1
1 1
𝑇𝑛−1 + �−2 − � 𝑇𝑛𝑖+1 = − 𝑇𝑛𝑖 (2.c)
𝜏 𝜏

𝛼∆𝑡 𝑘
𝑊ℎ𝑒𝑟𝑒 𝜏= ; 𝛼=
∆𝑥 2 𝜌𝐶𝑝

7
High temperature measurements at the internal nozzle wall of the ZEpHyR

A. Calculation of convective heat transfer coefficient

In order to calculate the temperature of the exhaust gas it is required to calculate the heat
transfer coefficient. Here we use Bartz equation to determine convective heat transfer
coefficient. Bartz derived this equation from the non-dimensional form of Nusselt number for
a fully developed turbulent pipe flow. The equation calculates heat transfer coefficient as a
function of specific heat, area ratio, viscosity, correlation factor (ξ) etc. Bartz equation is as
follows [14].

41.8565 𝜇 0.2 𝐶𝑝 𝑃𝑐 𝑔 0.8 𝐷𝑡 0.1 𝐴𝑡 0.9


ℎ= � 0.6 � � ∗ � � � � � 𝜉 (3.a)
𝐷𝑡0.2 𝑃𝑟 𝑐 𝑟 𝐴

1
𝜉=
𝑇𝑤 𝛾−1 1 0.68 𝛾−1 0.12 (3.b)
�2𝑇 �1 + 2 𝑀2 � + 2� �1 + 2 𝑀2 �
𝑠

The values of Prandtl number, viscosity, specific heat and characteristic velocity are
obtained from NASA CEA (Chemical Equilibrium with Applications) program. This
program calculates chemical equilibrium product concentrations from any set of reactants and
determines thermodynamic and transport properties for the product mixture. It has
applications that can calculate chemical equilibrium products inside the rocket chamber and
the transport properties for a given oxidizer and fuel.

Using the relationship given below the gas temperature, Tg can be calculated.

𝑞(𝑥, 𝑡) = ℎ(𝑇𝑔 − 𝑇𝑤 ) (4)

The ‘q(x,t)’ and ‘Tw’ values in above equation are calculated from IHCP while the value
of heat transfer coefficient ‘h’ is the one calculated from Bartz equation.

8
High temperature measurements at the internal nozzle wall of the ZEpHyR

IV. SOLUTION OF IHCP

The problem at hand is to determine the temperature of exhaust gas temperature at the
throat section of the nozzle of ZEpHyR engine. For this first internal wall surface temperature
and heat flux is to be estimated. These values cannot be measured directly from a sensor as it
disturbs the boundary conditions and hence the measurements. So the transient temperature
is measured at x=L (see Figure 2) by using single thermocouple. This is a classic case of
inverse heat conduction problem. As discussed in section II.B there are many techniques to
solve the inverse problem. Here the conjugate gradient method is used that belongs to the
class of iterative regularization methods suggested by Alfanov [13].

A. Conjugate Gradient Method (CGM) with adjoint problem for


function estimation

The conjugate gradient method with adjoint problem for function estimation is used to
solve heat flux as no prior information about its functional form is available. This method is
applied to the theoretical model developed in section III to estimate the heat flux. The steps
for this method are as follows [15].

1) The direct problem

The direct problem involves the determination of temperatures within the slab with known
heat flux. The formulation of direct problem is given by equations (1.a)–(1.d).

2) The inverse problem

In inverse problem, the heat flux is unknown at x = 0 but temperature readings at x=L are
available from the sensor and is denoted by Y(L,t). Utilizing the measured data, unknown
heat flux is estimated by minimizing the following function.

𝑡𝑓
𝑆[𝑞(𝑡)] = � {𝑌(𝐿, 𝑡) − 𝑇(𝐿, 𝑡)}2 𝑑𝑡 (5)
𝑡=0

Here T(L,t) are the estimated temperatures at x=L and time t, tf is the final time. T(L,t) is
determined from the solution of the direct problem by using the estimated heat flux.

9
High temperature measurements at the internal nozzle wall of the ZEpHyR

3) The sensitivity problem

The sensitivity function ∆T(x,t), solution of the sensitivity, is defined as the directional
derivative of temperature T(x,t) in the direction of the perturbations of the unknown function
[13]. The sensitivity problem is obtained by assuming that when q(0,t) is perturbed by an
amount ∆q(0,t) the resulting temperature is perturbed by ∆T(x,t). By replacing in the direct
problem q(0,t) by [q(0,t) + Δq(0,t)] and T(x,t) by [T(x,t) + ΔT(x,t)] and then subtracting the
original direct problem we get the following sensitivity problem.

𝜕 2 ∆𝑇 𝜕∆𝑇
𝑘 2
= 𝜌𝐶𝑝 0 ≤ 𝑥 ≤ 𝐿 𝑎𝑛𝑑 𝑡 > 0 (6.a)
𝜕𝑥 𝜕𝑡

∆𝑇(𝑥, 𝑡) = 0 0 ≤ 𝑥 ≤ 𝐿 𝑎𝑛𝑑 𝑡 = 0 (6.b)

𝜕∆𝑇(𝑥, 𝑡)
=0 𝑥 = 𝐿 𝑎𝑛𝑑 𝑡 > 0 (6.c)
𝜕𝑥

𝜕∆𝑇(𝑥, 𝑡)
𝑘 = ∆𝑞(𝑥, 𝑡) 𝑥 = 0 𝑎𝑛𝑑 𝑡 > 0 (6.d)
𝜕𝑥

The above equations are solved by using the finite difference method. ∆q(0,t) is the only
quantity that is needed to solve the above sensitivity problem. The calculation of this term is
discussed later.

10
High temperature measurements at the internal nozzle wall of the ZEpHyR

4) The adjoint problem

The Lagrange multiplier is used to minimize the function S[q(t)]. The adjoint problem is
calculated to find the response of the system to small perturbations. The Lagrange multiplier
‘λ(0,t)’ is determined by the solution of the problem that is adjoint to the sensitivity problem
given by equation (6.a)-(6.d). After several mathematical steps, the following adjoint problem
is obtained.

𝜕 2𝜆 𝜕𝜆
𝑘 2
= −𝜌𝐶𝑝 0 ≤ 𝑥 ≤ 𝐿 𝑎𝑛𝑑 𝑡 > 0 (7.a)
𝜕𝑥 𝜕𝑡

𝜆(𝑥, 𝑡) = 0 0 ≤ 𝑥 ≤ 𝐿 𝑎𝑛𝑑 𝑡 = 0 (7.b)

𝜕𝜆(𝑥, 𝑡) −2[𝑇(𝐿, 𝑡) − 𝑌(𝐿, 𝑡)]


= 𝑥 = 𝐿 𝑎𝑛𝑑 𝑡 > 0 (7.c)
𝜕𝑥 𝑘

𝜕𝜆(𝑥, 𝑡)
=0 𝑥 = 0 𝑎𝑛𝑑 𝑡 > 0 (7.d)
𝜕𝑥

The Lagrange multiplier is calculated from the above equations by using the finite
difference technique.

5) The gradient equation

For the function S[q(t)], the gradient equation is given as follows;

∇𝑆[𝑞(𝑡)] = 𝜆(0, 𝑡) (8)

6) The iterative procedure

The unknown function q(0,t) is estimated by using the iterative procedure of conjugate
gradient method that is given by the following equation.

𝑞 𝑘+1 (𝑡) = 𝑞 𝑘 (𝑡) − 𝛽 𝑘 𝑑 𝑘 (𝑡) (9)

11
High temperature measurements at the internal nozzle wall of the ZEpHyR

In equation (9) ‘k’ is the number of iteration, ‘βk’ is the search step size and ‘dk(t)’ is the
direction of descent. The search step size βk can be calculated by minimizing S[q(k+1) (t)] with
respect to βk. The final expression obtained for the step size is given below.

𝑡
𝑓
∫𝑡=0[𝑇(𝐿, 𝑡) − 𝑌(𝐿, 𝑡)]∆𝑇(𝐿, 𝑡)𝑑𝑡
𝑘
𝛽 = 𝑡𝑓
(10)
∫𝑡=0[∆𝑇(𝐿, 𝑡)]2 𝑑𝑡

The direction of descent ‘dk(t)’ can be calculated from the below equation.

𝑑𝑘 (𝑡) = ∇𝑆[𝑞 𝑘 (𝑡)] + 𝛾 𝑘 𝑑𝑘−1 (𝑡) (11)

‘γk’ is the conjugate coefficient that is calculated from the following Fletcher-Reeves
expression.

𝑡
𝑓
∫𝑡=0{∇𝑆[𝑞 𝑘 (𝑡)]}2 𝑑𝑡
𝛾𝑘 = 𝑡𝑓
(12)
∫𝑡=0{∇𝑆[𝑞 𝑘−1 (𝑡)]}2 𝑑𝑡

With γ0 = 0 for first iteration

7) The stopping criteria

The iterations shall be continued until the functional value S[q(t)] reaches the prescribed
tolerance ‘ε’. The stopping criterion is given as follows.

𝑆[𝑞(𝑡)] < 𝜀 (13.a)

If there are no errors in the measured data then the tolerance can be chosen to be a small
value. However, there are always some measurement errors in the data. In this case, the
solution can be assumed accurate when

|𝑌(𝐿, 𝑡) − 𝑇(𝐿, 𝑡)| ≈ 𝜎 (13.b)

Where ‘σ’ is the standard deviation of the measurement errors and is assumed to be
constant.

12
High temperature measurements at the internal nozzle wall of the ZEpHyR

From equation (5), ‘ε’ is obtained as follows.

𝜀 = 𝜎 2 𝑡𝑓 (13.c)

The conjugate gradient method with adjoint problem for function estimation can be used
to solve the inverse heat conduction equation. A Matlab code is developed for the above
iteration which is used to estimate the gas temperature and is given in Appendix.A.

B. Computational algorithm

Take an initial guess q0 (t) for the function q(t). Set k = 0 and then follow the steps below:

1. Solve the direct problem (1.a) - (1.d) and compute T(x,t), based on qk (t).
2. Check the stopping criterion (13) and continue if not satisfied.
3. Knowing T(L,t) and measured temperature Y(L,t), compute λ(0,t) by solving the
adjoint problem (7.a) - (7.d).
4. Knowing λ(0,t), compute the gradient ∇𝑆[𝑞 𝑘 (𝑡)] from equation (8).
5. Knowing ∇𝑆[𝑞 𝑘 (𝑡)], compute γk from equation (12) and the direction of descent
dk(t) from equation (11).
6. Set Δqk(t) = dk(t) and obtain ΔT(L,t) by solving the sensitivity problem (6.a) - (6.d).
7. Knowing ΔT(L,t), compute the search step size βk from equation (10).
8. Knowing the search step size βk and the direction of descent dk(t), compute the
new estimate qk+1 (t) from equation (9) and return to step 1.

13
High temperature measurements at the internal nozzle wall of the ZEpHyR

V. MODEL VERIFICATION
The theoretical model developed in section III is solved for n=3 nodes to find the heat flux
in the nozzle. The conjugate gradient method from section IV.A is used to find the solution of
the inverse heat conduction problem.

R.C.Mehta published a research paper to estimate the heat flux in a rocket nozzle by
applying Newton-Raphson iteration technique to the inverse heat conduction problem [16]. In
his paper, R.C.Mehta approximates the calculated temperature as a linear function of heat
flux using the Taylor series method. The results presented in his research are used to verify
the model formulated in this thesis. The material properties used in his research paper are
incorporated in the model along with the measured temperature data. The properties and the
measured data are given in Table I and Table II respectively.

Table I Material properties of Mild Steel [17]

Property Value Units

Thermal conductivity 35 W m-1 K-1


Specific heat capacity 545 J kg-1K-1
Density 7900 kg m-3
Slab thickness 0.0211 m

14
High temperature measurements at the internal nozzle wall of the ZEpHyR

The measured data is given for time t=7s to t=16s for initial temperature of 300K.

Table II Time vs Measured temperature [16]

Time (s) Measured Temperature (K)

7 342

8 356

9 380

10 402

11 425

12 440

13 460

14 479

15 507

16 528

15
High temperature measurements at the internal nozzle wall of the ZEpHyR

Measured temperature
600

550

500
Temperature (K)

450

400

350

300
0 5 10 15 20
Time (s)

Figure 3 Measured temperature vs time

The values of Table I and Table II are used to estimate the heat flux by applying conjugate
gradient with adjoint problem for function estimation.

The results computed from the model are compared with those of R.C.Mehta and are
shown in Figure 4.

R.C Mehta Model


4.0E+06

3.5E+06

3.0E+06
Heat Flux (W m-2)

2.5E+06

2.0E+06

1.5E+06

1.0E+06

5.0E+05

0.0E+00
5 7 9 11 13 15 17
Time (s)

Figure 4 Comparison of estimated heat flux results

16
High temperature measurements at the internal nozzle wall of the ZEpHyR

As seen in the Figure 4, the estimated heat flux obtained from the theoretical model agrees
well with the heat flux values estimated by R.C.Mehta from 11 s to 16 s interval but there is
difference in heat flux before 11s.

The difference in results might be because different iteration techniques are used to solve
the inverse heat conduction problem. The calculated temperature is, in general, a nonlinear
function of heat flux. R.C.Mehta considered a linear approximation when solving the inverse
problem with Newton-Raphson method but in CGM such an approximation is not required.

Also the heat flux obtained by R.C.Metha has a decrease from 10s to 12s the reason
behind this dip in flux remains to be unknown. This behaviour might be due to time lag
considered by the R.C.Mehta. But time lag is not included in the above case when estimating
the heat flux. Since no information is available about the time lag values considered by
R.C.Mehta, constant values of time lag are used and resulting heat flux are estimated. The
Figure 5 and Figure 6 show the values of heat flux and the resulting percentage errors after
considering time lag of 1s, 1.5s and 1.8s in the model.

R.C Mehta Time lag = 1s


Time lag = 1.5s Time lag = 1.8s
4.0E+06

3.5E+06

3.0E+06
Heat Flux (W m-2)

2.5E+06

2.0E+06

1.5E+06

1.0E+06

5.0E+05

0.0E+00
5 7 9 11 13 15 17
Time (s)

Figure 5 Comparison of estimated heat flux for different time lags

17
High temperature measurements at the internal nozzle wall of the ZEpHyR

% error (time lag = 1s) % error (time lag = 1.5s)


% error (time lag = 1.8s)
40

35

30
Percentage Error

25

20

15

10

0
5 7 9 11 13 15 17
Time (s)

Figure 6 Comparison of errors in heat flux for different time lags

The plots of Figure 5 and Figure 6 show that the maximum error is for time lag of 1s
while minimum error is present when time lag of 1.8s is considered. Including time lag
values improves the results but the estimated heat flux obtained shows no resemblance to the
behaviour as seen in heat flux obtained by R.C.Mehta. As seen in Figure 3 the measured
temperature increases with time should expect an increase in heat flux which is the case in the
estimated heat flux obtained from the model. Therefore, the theoretical model developed and
the solution technique applied can be efficiently used to predict the heat flux in throat section
of the nozzle of ZEpHyR engine.

18
High temperature measurements at the internal nozzle wall of the ZEpHyR

VI. RESULTS AND DISCUSSION

The theoretical model for ZEpHyR is used to estimate the heat flux and nozzle internal
surface wall temperature for n = 3 nodes. The heat transfer coefficient is calculated with the
help of Bartz equation. Using all these values the exhaust gas temperature is computed. The
rocket nozzle is made of Molybdenum having a wall thickness of 0.0103m. The properties of
Molybdenum are summarized in Table III.

Table III Properties of Molybdenum [3]

Property Value Units

Thermal conductivity 138 W m-1 K-1

Specific heat capacity 250 J kg-1 K-1

Density 10220 kg m-3

Currently, the ZEpHyR main engine is being assembled on a test bench at ZARM
institute. As a result, the tests cannot be conducted to obtain temperatures from the sensor on
the nozzle. Hence, a triangular test flux is used to collect the exact temperatures (Te) data
every second by direct method given by equations (1.a)–(1.d) for the total duration of 18 s.

But the exact temperatures obtained by using the test flux do not include errors. Therefore,
random measurement errors are introduced in the exact temperatures obtained by direct
method. This is done by the following equation [15].

𝑌 = 𝑇𝑒 + 𝜔𝜎 (14)

Where ‘σ’ is the standard deviation and ‘ω’ is the random variable with normal
distribution, zero mean and unitary standard deviation and for the confidence bound of 98.9%
we have -2.576 < ω < +2.576 [15]. Y is the measured temperature with errors.

19
High temperature measurements at the internal nozzle wall of the ZEpHyR

The standard deviations ‘σ’ of 3 K, 5 K and 10 K are used. The stopping criterion is
calculated for each standard deviation from equation (13.c) and the initial guess of zero is
made for heat flux in the iteration. In order to verify the accuracy of CGM, the heat flux
estimated from the measured temperature is compared with heat flux obtained from exact
temperature values. The plots are shown in Figure 7, Figure 8 and Figure 9. Difference
between test and estimated flux can be seen in all three plots for the initial time duration and
difference becomes larger as standard deviation increases which is minimum for 3 K and
maximum for 10 K. This is because as the standard deviation increases the stopping criterion
also increases. Although maximum difference is observed for σ = 10 K the results are still in
good agreement. This shows that even with high measurement errors the heat flux estimated
from CGM is quiet accurate.

Test flux 3 sigma


3.00E+06

2.50E+06

2.00E+06
Heat Flux (W m-2)

1.50E+06

1.00E+06

5.00E+05

0.00E+00
0 5 10 15 20
Time (s)

Figure 7 Comparison of test heat flux and 3 σ heat flux

20
High temperature measurements at the internal nozzle wall of the ZEpHyR

Test flux 5 sigma


3.00E+06

2.50E+06

2.00E+06
Heat Flux (W m-2)

1.50E+06

1.00E+06

5.00E+05

0.00E+00
0 5 10 15 20
Time (s)

Figure 8 Comparison of test heat flux and 5 σ heat flux

Test flux 10 sigma


3.00E+06

2.50E+06

2.00E+06
Heat Flux (W m-2)

1.50E+06

1.00E+06

5.00E+05

0.00E+00
0 5 10 15 20
Time (s)

Figure 9 Comparison of test heat flux and 10 σ heat flux

21
High temperature measurements at the internal nozzle wall of the ZEpHyR

The heat flux and internal surface wall temperature are estimated using the conjugate
gradient iteration technique for initial temperature of 300 K and σ = 10 K as it presents the
maximum error in measurements (worst case). The initial guess of zero is made for heat flux
in the iteration. The results obtained are summarized in Table IV.

Table IV Estimated heat flux and internal wall temperature

Estimated
Exact Measured Estimated heat
internal wall
Time (s) Temperature temperature flux
temperature
Te (K) Y (K) q x 106 (W m-2)
Tw (K)
1 306.79 313.20 300 0

2 331.88 306.36 300 0

3 375.91 400.97 435.38 0.99

4 438.92 422.46 473.65 0.99

5 520.93 567.42 621.24 1.51

6 621.94 594.18 678.97 1.51

7 741.94 747.06 810.88 1.77

8 880.94 841.85 878.30 1.77

9 1038.93 1020.33 1100.01 2.19

10 1215.93 1200.68 1290.22 2.46

11 1331.09 1341.41 1435.42 2.58

12 1317.34 1341.55 1427.66 2.35

13 1284.60 1292.33 1368.00 2.07

14 1232.89 1223.30 1288.64 1.78

15 1162.19 1183.15 1241.45 1.59

16 1072.51 1093.04 1142.09 1.33

17 963.83 966.92 1005.73 1.05

18 836.17 850.85 881.11 0.82

22
High temperature measurements at the internal nozzle wall of the ZEpHyR

For t =1s and 2s the heat flux estimated is zero because the stopping criteria value
obtained for σ = 10 K is very high that limits the iterations performed in the CGM.

A. Estimation of convective heat transfer coefficient and exhaust


gas temperature

The heat transfer coefficient is estimated at the throat section of the nozzle by using Bartz
equation given in section III.A. The parameters required for its calculations are obtained from
NASA’s CEA program for oxidizer to fuel mixture ratio of 2.1 and chamber pressure of 35
bar [18]. The oxidizer and fuel used in ZEpHyR are liquid oxygen and paraffin respectively.
The properties and transport parameters obtained are listed in Table V.

Table V Properties and transport parameters of combustion gas

μ Cp Dt r c*
Property Pr A/At γ
(Pa s) (J kg-1 K-1) (m) (m) (m s-1)

Value 9.8873x10-5 2028 0.0206 0.0103 1768.1 0.6196 1 1.1572

Using the estimated heat flux and internal wall temperature from Table IV, heat transfer
coefficient and exhaust gas temperature are estimated. The exhaust gas temperature is
calculated from equation (4). The results obtained are given in Table VI.

23
High temperature measurements at the internal nozzle wall of the ZEpHyR

Table VI Estimated Heat transfer coefficient and exhaust gas temperature

Time Estimated heat transfer coefficient Estimated Exhaust gas temperature


-2 -1
(s) h (W m K ) Tg (K)

1 - -

2 - -

3 6201.77 596.20

4 5856.40 643.96

5 4869.98 931.90

6 4584.43 1008.97

7 4063.06 1246.55

8 3848.28 1338.29

9 3302.12 1764.21

10 2962.72 2119.96

11 2755.48 2370.11

12 2765.65 2279.09

13 2847.11 2093.67

14 2965.19 1889.44

15 3041.39 1763.47

16 3218.90 1556.69

17 3509.59 1306.30

18 3839.94 1095.15

24
High temperature measurements at the internal nozzle wall of the ZEpHyR

A triangular text flux is used to check the stability of the model. The model is stable to
estimate the heat flux and exhaust gas temperature with measurement errors. Furthermore,
this model can be used to estimate the exhaust gas temperature using actual test measured
data and help predict whether Molybdenum is able to withstand the high exhaust gas
temperature.

25
High temperature measurements at the internal nozzle wall of the ZEpHyR

VII. CONCLUSION AND FUTURE WORK

A. Conclusion

The theoretical model developed for nozzle is used to obtain the temperature of exhaust
gas by applying the inverse heat conduction technique and Bartz equation. Conjugate
gradient numerical iterative technique used to estimate the heat flux has the advantage that it
requires no prior knowledge about the functional form of heat flux and has high rate of
convergence. Also high accuracy of results is obtained from CGM that shows that it is not
sensitive to measurement errors. Thus model can be used to estimate the exhaust gas
temperature using actual test measured data. And help foresee whether the Molybdenum
material is suitable for the nozzle of the ZEpHyR engine without active cooling.

B. Future Work

The measured data for temperature is not available from the ZEpHyR nozzle so a test case
has been used for heat flux. Actual temperature data can be obtained from the thermocouple
that can be used to estimate heat flux and thus the actual exhaust gas temperature.

Time lag has a significant effect on the results and hence should be considered. Much of
the research is done in this area and techniques are available that can be used for its
estimation.

Constant thermal properties are assumed in this research. The thermal properties of a
material vary with varying temperature. In order to improve the results the variation in
properties can be considered. In addition modelling the nozzle as a hollow cylinder and
incorporating multiple thermocouples at different locations in the nozzle will also improve
the accuracy of results.

26
High temperature measurements at the internal nozzle wall of the ZEpHyR

REFERENCES
[1] "A Brief History of Hybrid Rocket Technology". Space Propulsion Group, Inc.

[2] George P. Sutton & Oscar Biblarz, “Rocket propulsion elements” - 7th Edition,
December 29, 2000.

[3] Lide, David R., ed. (1994). "Molybdenum". CRC Handbook of Chemistry and Physics 4.
Chemical Rubber Publishing Company. p. 18. ISBN 0-8493-0474-1.

[4] Anderson, John D., Jr., Fundamentals of Aerodynamics, 2nd Edition McGraw-Hill, New
York, 1991

[5] Stolz, G., "Numerical Solutions to an Inverse Problem of Heat Conduction for Simple
Shapes," Journal of Heat Transfer, Vol. 82,Feb. 1960, pp. 20-26.

[6] Beck, J.V., "Surface Heat Flux Determination Using an Integral Method," Nuclear
Engineering Design, Vol. 7, 1968, pp. 170-178

[7] Beck, J.V., "Nonlinear Estimation Applied to the Nonlinear Inverse Heat Conduction
Problem," International Journal of Heat and Mass Transfer, Vol. 13, April 1970, pp. 703-
716.

[8] J.V.Beck , B.Blackwell & C.St.Clair, Inverse Heat Conduction: Ill-posed Problems,
Wiley, N.Y.,1985.

[9] Burggraf, O.R., "An Exact Solution of the Inverse Problem in Heat Conduction Theory
and Applications," Journal of Heat Transfer, Vol. 86, Aug. 1964, pp. 373-382.

[10] M. Raynaud & J. Bransier, "NEW FINITE-DIFFERENCE METHOD FOR THE


NONLINEAR INVERSE HEAT CONDUCTION PROBLEM" Published online: 27 Feb
2007

[11] Tikhonov, A. N., "Solution of Incorrectly Formulated Problems and the


Regularization Method", Soviet Math. Dokl., 4(4), 1035- 1038, 19637

[12] Tikhonov, A. N., "Regularization of Incorrectly Posed Problems", Soviet Math. Dokl.,
4(6), 1624- 1627, 1963

[13] Alifanov, 0. M., "Solution of an Inverse Problem of Heat-Conduction by Iterative


Methods", J. Eng. Phys., 26(4), 47 1-476, 1974.

27
High temperature measurements at the internal nozzle wall of the ZEpHyR

[14] Bartz, D. R., "A Simple Equation for Rapid Estimation of Rocket Nozzle Convective
Heat Transfer Coefficient," Jet Propulsion, Vol. 27, Jan. 1957, pp. 49-51

[15] M. N. Ozisik and H. R. B Orlande, Inverse Heat Transfer, Fundamentals and


Applications, Taylor & Francis, New York, 2000.

[16] Mehta, R.C., "Estimation of Heat-Transfer Coefficient in a Rocket Nozzle," AIAA


Journal, Vol. 19, Aug. 1981, pp. 1085-1086.

[17] Mehta, R. C., "Solution of the Inverse Conduction Problem," AIAA Journal, Vol. 9,
Sept. 1977, pp. 1355-1356.

[18] STERN_ZEpHyR_SPDv1-0_WORKINGSTAGE PDR.

28
High temperature measurements at the internal nozzle wall of the ZEpHyR

APPENDIX

A. Matlab code

%File: Estimation of Exhaust gas temperature

%Author: Manohar Karnal

%Date: 10−August −2014

clc
clear all
format long

n = 10; % no of nodes.
k = 138; % Thermal conductivity of the material. W/m K
Rho = 10220; % Density of the material. kg/m^-3
Cp = 250; % Specific heat of the material.J/kg K
delta_x = 0.0103/(n-1); % Distance between the nodes.
burntime= 18; % Total burn time.
q = 0; % Intial gueses of heat flux.
T_in = 300; % Intial temperature.
Sd = 10; % Standard deviation.
Sc = Sd^2*burntime; % Stopping criteria.
% lambda = Lagrange multiplier.
% yk = Conjugate coefficient.
% Tm = Measured temperature.
alpha = k/(Rho*Cp);
Tm = obtained from thermocouple data (test case)

%% -----------------Solution of inverse problem--------------------------%%

for t=1:burntime

tau=(alpha*t)/((delta_x)^2);

%% --------------Direct Problem with initial gusess 'q'------------------%%

for u = 1:200

for i = 1:n;
T(i,1) = T_in;
if i == 1;
B(i,1) = ((-1/tau)*T(i:i))-
((2*delta_x*q)/k);
else
B(i,1) = ((-1/tau)*T(i:i));
end

for j = 1:n;

if j == i;

A(i,j) = (-2-1/tau);

29
High temperature measurements at the internal nozzle wall of the ZEpHyR

elseif i == 1 && j == i+1;

A(i,j)=2;

elseif i == n && j == n-1;

A(i,j)=2;

elseif i>1 && i<=n-1;

for a = i-1:2:i+1;

A(i,a) = 1;

end

end

end

end

Tc = A\B;

if (-(Tc(n:n))+Tm(t:t))^2 <= Sc;

Tw_Tc_Tq_h_Tg(t,1) = Tc(1:1);
Tw_Tc_Tq_h_Tg(t,2) = Tc(n:n);
Tw_Tc_Tq_h_Tg(t,3) = q;

break;
end

%% ----------------------------Adjoint problem---------------------------%%

if u == 1;
lambda0= zeros(n); % intial value
else
lambda0 = lambda;
end

for l = 1:n;
A1(n,n)= -1;
if l == n;
B1(l,1) = ((-2*delta_x*(Tc(n:n)-
Tm(t:t)))/k);
else
B1(l,1) = ((-1/tau)*lambda0(l:l));
end

for m = 1:n-1;

if m == l;

A1(l,m) = (-2-1/tau);

elseif l == 1 && m == l+1;

30
High temperature measurements at the internal nozzle wall of the ZEpHyR

A1(l,m)=2;

elseif l == n && m == n-1;

A1(l,m)=1;

elseif l>1 && l<=n-1;

for a = l-1:2:l+1

A1(l,a) = 1;

end

end

end

end

lambda = A1\B1;

delta_S = lambda(1:1);
if u == 1;
yk = 0;
else
yk = (delta_S)^2/(delta_S0)^2;
end
%% --------------------------direction of descent------------------------%%
if u ==1
dk0 = 1;
else
dk0 = dk
end
dk = delta_S + yk * dk0;

%% ---------------------------Sensitivityproblem-------------------------%%
delta_q = dk;
delta_T = zeros(n);

for y = 1:n;

if y == 1;
B2(y,1) = ((-1/tau)*delta_T(y:y))-
((2*delta_x*delta_q)/k);
else
B2(y,1) = ((-1/tau)*delta_T(y:y));
end

for z = 1:n;

if z == y;

A2(y,z) = (-2-1/tau);

31
High temperature measurements at the internal nozzle wall of the ZEpHyR

elseif y == 1 && z == y+1;

A2(y,z)=2;

elseif y == n && z == n-1;

A2(y,z)=2;

elseif y>1 && y<=n-1;

for a = y-1:2:y+1;

A2(y,a) = 1;

end

end

end

end
delta_Tc = A2\B2;

%% --------------------------Search step size----------------------------%%

beta = ( Tc(n:n) - Tm(t:t) )/delta_Tc(n:n);

%% -------------------------Estimated heat flux--------------------------%%

qnew = q - (beta*dk);

%% --------------------------Next iteration------------------------------%%

delta_S0 = delta_S;
q = qnew;

end

%% --------------------------------------------------------------------- %%

%------Convective heat transfer coefficient using Bartz Equation------%

Dt = 20.60E-3; % Throat diameter. m


Mu = 0.098873E-3 ; % Viscosity. Pa s
Cp = 2.0284E3; % Specific heat at const pressure. J/kg-K
Pc = 35E5; % Chamber pressure. bar
g = 9.81; % Gravitational constant. m/s^2
c_star = 1768.1; % Characteristic velocity. m/s
r = Dt/2 ; % Throat radius of curvature. m
At = 1; % Throat area. m^2
A = 1; % Area along chamber axis. m^2
Pr = 0.6196; % Prandtl number.
gamma = 1.1572; % Ratio of specific heat.
M = 1; % Local Mach number.
Ts =3203.81; % Static temperature along the stagnation
streamline at throat section.

32
High temperature measurements at the internal nozzle wall of the ZEpHyR

% sigma = corelation factor.


% h = convective heat transfer coefficient.

%% --------------------Staganation temperature---------------------------%%

T0 = Ts*(1+((gamma-1)/2)*M^2);

%% -----------------------heat transfer coefficient----------------------%%

sigma = 1/(((((Tw_Tc_Tq_h_Tg(t,1)/2*T0)*(1+((gamma-1)/2)*M^2) +
1/2))^0.68) * (1+(((gamma-1)/2)*M^2)^0.12));

h = (41.8565/Dt^0.2)* ( ( ((Mu^0.2)*Cp)/Pr^0.6
)*((Pc*g/c_star)^0.8)*((Dt/r)^0.1)*(At/A)^0.9)*sigma ;

Tw_Tc_Tq_h_Tg(t,4) = h;

%% --------------Calculation of exhaust gas temperature------------------%%

Tg = (Tw_Tc_Tq_h_Tg(t,3)/Tw_Tc_Tq_h_Tg(t,4))+ Tw_Tc_Tq_h_Tg(t,1);

Tw_Tc_Tq_h_Tg(t,5) = Tg;

end

Tw_Tc_Tq_h_Tg

33

S-ar putea să vă placă și