Professional Documents
Culture Documents
net/publication/316174400
CITATIONS READS
17 5,138
5 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Zeping Wu on 19 October 2017.
Solid-Rocket-Motor Performance-Matching
Design Framework
DOI: 10.2514/1.A33655
An effective performance-matching design framework for solid rocket motor tailored toward satisfying various
thrust-performance requirements is presented in this research paper through an innovative and specialized general-
design approach developed to evaluate the general-design parameters. During the general-design stage, a
combination of grain web and area ratio is selected as the design variables to be adjusted to obtain the general
parameters. Based on the general parameters obtained, a grain-design stage incorporates the level-set method and
simulates solid-propellant evolution and internal ballistic analysis, thereby obtaining the thrust performance. Grain-
design effectiveness is determined by how closely the designed solid-rocket-motor performance matches and
compares to a prespecified thrust curve. An efficient sequential-field-approximate-optimization algorithm is
proposed and used to minimize the average rms error between the desired and designed thrusts. Validation of the
proposed design framework is carried out by evaluating motor cases possessing different thrust requirements, and
results obtained highlight the proposed framework as a practical and efficient strategy for solid-rocket-motor designs.
obtaining extended details of thrust performance prior to the detailed presents the general-parameter evaluation method, which determines
design of the SRM, which implies that a prespecified thrust-time all general parameters of an SRM scheme. Section IV describes the
history can be determined by the conceptual design of a vehicle, thus grain-design method based on the SFAO. Section V presents several
ensuring engineers possess the capability of designing SRMs that test cases for validation of the proposed design framework, followed
would match a specified thrust-time performance. Recently, more by concluding remarks in Sec. VI.
efforts have been made toward taking the thrust-time history into
account [10,11]. Albarado et al. [10] developed a performance- II. General Framework of SRM Design
matching design method, driven by a pattern search/particle swarm
optimization (PSO) algorithm. A similar approach was employed in In SRM design, even with the aim of locating optimum values for
[11], which applies a modified ant-colony optimization (ACO). In design variables, there still exist an infinite number of possibilities
these evolutionary algorithms, PSO and ACO were integrated with a that might render such design impossible to evaluate. Therefore, to
local search approach to perform the optimization. Simple star grain perform practical SRM design, there is a need to model and solve the
design problem by coupling the design variables to the trajectory and
was specified as the grain configuration for its burn-back analysis,
target of the design. Hence, there exists an aim to design an SRM that
which is implemented analytically, and a desirable thrust curve
would match a prespecified performance characteristic as much as
obtained after thousands of simulations or more. However, when a
possible. In this present research, the effectiveness of the design
complex three-dimensional grain and its consequent time-consuming
optimization is determined by how closely the SRM can match and
burn-back analysis are taken into account, this optimization approach compare with the prespecified thrust curve with respect to time.
Downloaded by NATIONAL UNIVERSITY OF DEFENSE on June 1, 2017 | http://arc.aiaa.org | DOI: 10.2514/1.A33655
mentioned so far becomes practically infeasible due to innumerous The proposed SRM design framework is presented in Fig. 1. Initial
iterations and its expensive associated computational costs. inputs to the entire design process include the design requirements
Therefore, for modern-day technological applications, require- and SRM materials. The design requirements specify the thrust with
ments are not only limited to designing a system that fits engineering respect to time and the design constraints, including the volumetric
needs, but also to attain an optimized system requiring the lowest loading efficiency, minimal chamber pressure, and permissible
possible cost, yet still possessing the greatest efficiency. Therefore, temperature. SRM materials determine the choice and properties of
this research proposes an efficient performance-matching design the chamber casing, propellant, and insulations. Based on these initial
framework for SRM. The design framework comprises two stages: inputs, the general-design stage is performed so as to feed in key
the general-design stage and the grain-design stage. The general- design parameters of nozzle-area ratio and grain web into the next
design stage aims to evaluate the general parameters of an SRM stage of grain design. During general design, the grain burn-back
scheme, and these are subsequently fed into the grain-design stage. analysis and internal ballistic analysis are employed as simulation
During the grain-design stage, a sequential field approximate models to generate the thrust curve. Critical grain-configuration
optimization (SFAO) is proposed to implement the grain design, due parameters are designated as design variables, which are adjusted by
to the efficacy of SFAO. the SFAO algorithm to search for the best-matching thrust
The remaining part of this paper is organized thus; Sec. II performance. Because of the implementation of the SFAO, the design
elucidates the general framework of the proposed method. Section III efficiency is highly increased, as will be shown in Sec. V.
Grain design
System design variables(X)
Iterations
x1: Nozzle area ratio End run out?
General design
Performance
End
• Propellant mass estimate matched?
SFAO
• Chamber design
- Insulation mass estimate
- Casing mass estimate Internal ballistic analysis
• Nozzle design
- Nozzle mass estimate Grain burn-back analysis
• Overall mass Mtot estimate
Find: Mtot(X) Grain design variables
Satisfy: Constraints
Initial Sampling
Field Approximation
Discretizing the thrust curve by N points on the time axis
Sequential Sampling
Constructing the surrogate model of thrust performance at every time node by the RBF
Downloaded by NATIONAL UNIVERSITY OF DEFENSE on June 1, 2017 | http://arc.aiaa.org | DOI: 10.2514/1.A33655
Evaluating the thrust performance of the optimal result by invoking the level set
burn-back analysis and internal ballistic analysis
Yes
End
The level-set method is widely used in different engineering in which v is the interface evolving velocity. By projecting the
applications for computing interface evolution [1,13], such as crystal velocity v onto the normal direction of the interface, then
growth, two-phase flows, flame propagation, solid–fluid–gas
interactions, etc. It naturally represents the complex boundaries φt vn ⋅ j∇φj 0 (15)
that form holes, split into multiple pieces, or merge with other
q
boundaries [13,14].
in which j∇φj φ2x φ2y φ2z .
In this paper, the level-set method is adapted and used to simulate
the solid-propellant regression. The most notable advantage of the The grain-burning surface evolving with a speed of vn toward its
level-set method is that a general code can be developed to perform normal is simulated using the level-set equation (15). Based on the laws
the burn-back analysis for different grains with high precision [13]. of propellant burning in parallel layers, vn is equal to the burning rate of
Furthermore, it imposes no limitations on either two-dimensional or propellant r. The burning surface is regarded as the boundary between
three-dimensional configuration. the phases of solid grain and gas in the SRM chamber. According to
The main idea of the level set is to represent a moving interface Γt Eq. (15), φX; t is less than 0 for X in the solid-grain phase, whereas
bounding a region of Ωt in Rn by a Lipschitz continuous function φX; t is greater than 0 for X in the gas phase within the chamber at
φX; t, which has the following properties:
8 Table 1 Design parameters of neutral thrust case
>
< φX; t −dX in Ωt
Parameters Values Units
φX; t 0 on Γt (12)
>
: φX; t dX Insulation Thermal-conductivity insulation 0.84 W∕mK
in Ωt
material λ0
Initial temperature T 0 300 K
in which dX min jX − XI j, XI ∈ Γt. Density of case material ρi 1150 kg∕m3
Because the interface Γt is represented as the zero level set of Chamber Density of case material ρc 7800 kg∕m3
function φX; t, the motion of Γt can be regarded as an evolution case
Strength of extension of case σ 1160 MPa
equation for φX; t by taking the time derivative of φX; t 0 to
Permissible casing wall temperature T 600 K
yield Propellant Burning-rate coefficient a 0.0765 mm∕Pan
Pressure index n 0.34 — —
∂φ ∂x ∂φ ∂y ∂φ ∂z Characteristic velocity C 1550 m∕s
φt 0 (13)
∂x ∂t ∂y ∂t ∂z ∂t Propellant density ρP 1700 kg∕m3
Specific-heat ratio λ 1.12 — —
which can also be described as Diameter of SRM DSRM 300 mm
Minimal pressure p0 2.5 MPa
Nozzle Nozzle efficiency ηn 0.95 — —
φt v ⋅ ∇φ 0 (14)
702 ZEPING ET AL.
X
k−1
Δe
tk (18)
each time step. The geometric parameters, such as burning area, k0
rk
burning perimeter, free volume, propellant volume and breathing area
for each burning step are captured by the level set method, which are
subsequently fed into the internal ballistic analysis procedure to obtain p
At;k π At;0 ∕π r0 tk 2 (19)
the performance of the SRM. The efficacy of the level-set method
applied in the grain burn-back analysis was demonstrated in [1], and
full details will be avoided here. in which r0 is the ablation rate of the nozzle throat.
After the pressure-time history is obtained, the thrust-time
2. Thrust-Performance Prediction relationship may be formulated as follows:
Pressure and thrust performances are predicted by the internal
ballistics analysis. In this present research, the internal ballistics Ft ηn CF pc tAt t (20)
analysis is performed by making an assumption of constant pressure
and nonerosion along the motor. Steady-state pressure is computed The thrust coefficient CF is given by Eq. (5).
by equating the mass generated within the chamber to the mass
ejected through the nozzle throat. When web burnt is ek , the pressure 80
pc;k is solved as follows:
Requirement
characteristic velocity of the propellant, and Ab;k is the burning area 40
SFAO
SAO
20 PSO
0
0 1 2 3 4 5
Time (s)
Fig. 4 Configuration of the finocyl grain. Fig. 6 Contrast between thrust results.
Table 5 General parameters of neutral thrust case 2 Table 6 Optimal results of neutral thrust case 2
Parameters Values Units L1, L2, R0, R1, θ, Computation costs,
Grain web e0 97.9 mm Parameters mm mm mm mm deg min
Area ratio ε 9.6 — — PSO 2012 650 35 129 59 4800
Diameter of nozzle throat Dt 90.1 mm SAO 1941 795 32 136 68 1788
Outer diameter of propellant Dp 292.7 mm SFAO 1932 765 35 134 67 252
Maximum pressure pmax 11.1 MPa
Mass of propellant 185.5 Kg
Mass of chamber 34.3 kg
Mass of nozzle 7.7 kg 2. Field Approximation
N points on the time axis t1 ; t2 ; : : : ; tN are selected to discretize
the thrust curves, which are converted into the following form:
B. General Procedures of Grain Design Based on SFAO
2 3
To commence the grain-design stage, configuration of the grain is X1 ; f1 t1 f2 t1 ··· f2m t1
initially determined based on the characteristics of the prerequisite thrust 6 X2 ; f1 t2 f2 t2 ··· f2m t2 7
6 7
profile. The properties of the propellant are preset in the general-design 6 .. .. 7 (21)
4 . . 5
stage and followed here. Also, the critical configuration parameters are
f1 tN f2 tN f2m tN
Downloaded by NATIONAL UNIVERSITY OF DEFENSE on June 1, 2017 | http://arc.aiaa.org | DOI: 10.2514/1.A33655
screened out and designated as design variables. SFAO is now employed X 2m ; ···
to adjust the design variables, and invokes the burn-back analysis and
thrust-performance-prediction modules iteratively so as to obtain the Based on the sample sets Xj ; fj ti j 1; 2; : : : ; 2m,
desirable thrust performance. Figure 3 presents a detailed procedure for approximate models si X are constructed using the radial basis
SFAO-based grain design, which is roughly divided into four stages: function (RBF) [1] at every time node. The thrust vs time
initial sampling, field approximation, termination evaluation, and performance field is surrogated and represented by N approximate
sequential sampling. models, which is the core reason this algorithm is called an SFAO.
Based on the N approximate models, the following optimization
1. Initial Sampling
problem is derived to approach the specified performance curve:
In grain design, the critical configuration parameters of the solid
grain are usually designated as design variables X. Suppose the
dimensionality of X is m, then the optimal Latin hypercube design 1X N
min s X − F0 ti 2 s:t:J ≤ J0 η ≥ η0 (22)
method is employed to sample 2m points in the feasible domain of X. N i0 i
Finally, the level-set burn-back analysis and internal ballistic analysis
are invoked to obtain 2m thrust-performance curves as follows:
100
2 3
X1 ; f1 t
6 X2 ; f2 t 7
6 7
6 .. .. 7
6 . . 7
6 7
4 5
X2 m ; f2 m t Requirement
Thrust (kN)
50 SFAO
SAO
PSO
0
0 1 2 3 4 5
Time (s)
Fig. 7 Configuration of the star grain. Fig. 9 Contrast between thrust results.
Table 7 General parameters of dual-thrust case Table 8 Optimal results of dual-thrust case
3. Termination Evaluation history of the objective is illustrated in Fig. 5a. It is obvious that this
optimization converges after 21 true function evaluations (10 initial
Based on the prespecified search precision eps and the maximum
sampling points plus the 11 sequential sampling points in 11
search iterations Kmax , termination of the optimization is
optimization iterations). The optimal grain-configuration parameters
implemented using the following criterion:
are listed in Table 4. The comparison between the prespecified thrust
First, define the following indicator at iteration k:
curve and the optimized result is illustrated in Fig. 6, and it is obvious
v that these are in agreement.
u N k−1
u1 X s ti − ski ti 2 To confirm the efficacy of this proposed method, various
errork t i
(23) optimization techniques can be used to minimize the rms error
N i0 F0 ti
(RMSE) to reach an optimal solution. The powerful sequential
approximate optimization (SAO) technique [15,16] and PSO [10] are
If errork < eps or k K max, the optimization is terminated and tested and used for verification within this research paper. Figure 5b
the optimized result is the output. Otherwise, the proposed algorithm presents the convergence history for the SAO technique. As clearly
proceeds to the sampling stage. seen, the SAO-based grain design converges after 149 true function
evaluations. A complete execution of the level-set burn-back analysis
4. Sequential Sampling
In this stage, the optimal solution of problem (22) together with the 200
thrust performance obtained by the level-set burn-back analysis and
internal ballistic analysis is added to the sample set so as to update the Requirement
surrogate models at the next iteration. SFAO
150
SAO
Thrust (kN)
Parameters L1, mm L12, mm L13, mm R6, mm bw mm L5, mm L4, mm R7, mm fw mm Computation costs, min
PSO 4042.6 640.6 674.4 148.0 26.2 2021.3 1036.2 328.4 44.4 4800
SAO 4240.2 1334.8 1731.0 197.8 36.9 840.3 798.3 328.4 96.9 3060
SFAO 4295.9 1200.8 1709.5 196.3 38.6 941.9 894.8 328.4 96.9 780
Downloaded by NATIONAL UNIVERSITY OF DEFENSE on June 1, 2017 | http://arc.aiaa.org | DOI: 10.2514/1.A33655
The optimized results obtained by the SAO and PSO are also listed The optimized results obtained by the SAO and PSO are also listed
in Table 8 for comparison. The convergence histories of the SAO and in Table 10 for comparison. The convergence histories of the SAO
PSO are presented in Figs. 10b and 10c, respectively, whereas the and PSO are presented in Figs. 12b and 12c, respectively, whereas the
optimized thrust curves obtained by the SAO and PSO are presented optimized thrust curves obtained by the SAO and PSO are presented
in Fig. 11 for comparison also. in Fig. 13.
References [10] Albarado, K. M., Hartfield, R. J., Hurston, B. W., and Jenkins, R. M.,
“Solid Rocket Motor Performance Matching Using Pattern Search/
[1] Wang, D., Fei, Y., Hu, F., and Zhang, W., “An Integrated Framework for Particle Swarm Optimization,” 47th AIAA/ASME/SAE/ASEE Joint
SRM Grain Design Optimization,” Journal of Aerospace Engineering, Propulsion Conference, AIAA, Reston, VA, 2011, pp. 1–14.
Vol. 228, No. 7, 2014, pp. 1156–1170. [11] Kiyak, Z. J., and Hartfield, R. J., “Solid Rocket Motor Design Using a
[2] Nisar, K., Guozhu, L., and Zeeshan, Q., “A Hybrid Optimization Modified Ant Colony Optimization Metaheuristic with Local Search
Approach for SRM FINOCYL Grain Design,” Chinese Journal of Capability,” 49th AIAA/ASME/SAE/ASEE Joint Propulsion Conference,
Aeronautics, Vol. 21, No. 6, 2008, pp. 481–487.
AIAA, Reston, VA, 2013, pp. 1–8.
doi:10.1016/S1000-9361(08)60164-8 [12] Wolpert, D. H., and Macready, W. G., “No Free Lunch Theorems for
[3] Kamran, A., and Guozhu, L., “Design and Optimization of 3D Radial Optimization,” IEEE Transactions on Evolutionary Computation,
Slot Grain Configuration,” Chinese Journal of Aeronautics, Vol. 23, Vol. 1, No. 1, 1997, pp. 67–82.
No. 4, 2010, pp. 409–414. doi:10.1109/4235.585893
doi:10.1016/S1000-9361(09)60235-1 [13] Osher, S., and Fedkiw, R., Level Set Methods and Dynamic Implicit
[4] Rousseau, C. W., Steyn, S. F., Sullwald, W., De Kock, E. R., Smit, G. J.
Surfaces, Springer–Verlag, New York, 2003, pp. 25–90.
F., and Knoetze, J. H., “Rapid Solid Rocket Motor Design,” 49th AIAA/
[14] Wang, M. Y., Wang, X. M., and Guo, D. M., “A Level Set Method for
ASME/SAE/ASEE Joint Propulsion Conference, AIAA, Reston, VA, Structural Topology Optimization,” Computer Methods in Applied
2013, pp. 1–12. Mechanics and Engineering, Vol. 192, Nos. 1–2, 2003, pp. 227–246.
[5] Peterson, J., and Garfield, J., “The Automated Design of Multi-Stage doi:10.1016/S0045-7825(02)00559-5
Solid Rocket Vehicles,” AIAA Paper 1976-744, July 1976. [15] Wang, D., Wu, Z., Fei, Y., and Zhang, W., “Structural Design Employing
[6] Kamran, A., and Liang, G., “An Integrated Approach for Optimization a Sequential Approximation Optimization Approach,” Computers &
Downloaded by NATIONAL UNIVERSITY OF DEFENSE on June 1, 2017 | http://arc.aiaa.org | DOI: 10.2514/1.A33655