ACOUSTIC OPTIMIZATION OF AN UNDERWATER VEHICLE

A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Engineering

By RAHUL KHAMBASWADKAR B.E., University of Mumbai, India 2001

2005 Wright State University

WRIGHT STATE UNIVERSITY

SCHOOL OF GRADUATE STUDIES

June 6, 2005 I HEREBY RECOMMEND THAT THE THESIS PREPARED UNDER MY SUPERVISION BY Rahul Khambaswadkar ENTITLED Acoustic Optimization of an Underwater Vehicle BE ACCEPTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF Master of Science in Engineering.

Ravi C. Penmetsa, Ph.D. Thesis Director

Richard J. Bethke, Ph.D. Department Chair Committee on Final Examination

Ravi C. Penmetsa, Ph. D.

Ramana V. Grandhi, Ph.D.

Nathan W. Klingbeil, Ph.D.

Joseph F. Thomas, Jr., Ph.D. Dean, School of Graduate Studies

ABSTRACT Khambaswadkar, Rahul, M. S. Engineering., Department of Mechanical and Materials Engineering, Wright State University, 2005. Acoustic Optimization of an Underwater Vehicle Involving Fluid-Structure Interaction A torpedo is a guided missile that travels underwater and detonates when it comes in proximity of the target. Its speed and accuracy make it one of the most lethal weapons in navy munitions. The torpedo is a complex system comprising various subsystems: propulsion, weapon, guidance and control, and many other complicated auxiliary equipment important for proper operation of the torpedo. The structural design and optimization of a lightweight torpedo involves multiple disciplines, such as structures, fluids, and controls, of which acoustic analysis is a critical part. In underwater warfare, sophisticated active and passive SONAR (SOund NAvigation and Ranging) techniques are used by the enemy submarines to detect approaching torpedoes. Therefore, it is very important for a torpedo to be acoustically silent in order to increase its effectiveness. Each torpedo emits a specific acoustic signature depending on its propulsion, hydrodynamics, and other auxiliary noise-producing sources. In this thesis, experimental data available for the gear noise is simulated using computational sound sources that are then used to determine the acoustic signature of a torpedo. Furthermore, the Finite Element Method (FEM) is used to quantify acoustic behavior for the computational model of a lightweight torpedo. A framework for computational modeling of experimental data from various sources, incorporation of this information into the acoustic analysis, and multidisciplinary optimization of a lightweight torpedo are the main focal points of this thesis.

iii

TABLE OF CONTENTS
1. Introduction.... 1

1.1 Literature Review.. 1.2 Project Approach... 2. Finite Element Modeling of Torpedo, Fluid, and Noise Source... 2.1 Modeling of Lightweight Torpedo 2.2 Modeling of Fluid. 2.3 Fluid-Structure Interaction 2.4 Normal Mode Analysis Results 3. Torpedo Noise Modeling...... 3.1 Sources of Noise Generation. 3.2 Experimental Setup and Noise Profile 3.3 Optimization Formulation for Noise Source Modeling...... 3.4 Results and Discussion.. 4. Multidisciplinary Design Optimization (MDO) of Lightweight Torpedo 4.1 Optimization Formulation......... 4.2 Optimization in NASTRAN. 4.3 Optimization Results and Discussion 5. Concluding Remarks ... Appendix........ Appendix A: C++ Program to generate torpedo model input file.... Appendix B: MATLAB file to read NASTRAN output and calculate error Appendix C: NASTRAN input deck for Optimization Run.

2 4 8 8 11 19 23 28 29 31 33 37 41 46 47 49 54 56 56 63 65

iv

References

70

LIST OF FIGURES Figure 1.1 Figure 2.1 Figure 2.2 Figure 2.3 Figure 2.4 Figure 2.5 Figure 2.6 Figure 2.7 Figure 3.1 Figure 3.2 Figure 3.3 Flowchart of Research Approach Dimensions of Radial and Longitudinal Stiffeners Finite Element Model of the Lightweight Torpedo Fluid and Structural Finite Element Models Tolerances for Fluid-Structure Interaction First Bending Mode Second Bending Mode Breathing Mode Important Sources of Noise Generation in a Torpedo Experimental Setup Used for Gear Noise Noise Levels on the Meridian of Hemisphere about the MK-40 Torpedo Figure 3.4 Figure 3.5 Figure 3.6 Figure 3.7 Figure 3.8 Figure 3.9 Figure 4.1 Figure 4.2 Figure 4.3 Figure 4.4 Finite Element Model of Air Chamber Noise Recovery Points in the Air Chamber Optimization Algorithm to Determine Source Strength Intensity of a Pulsating Point Source Results for a Constant Profile Case Variation Between Experimental and NASTRAN Results Air Mesh Inside Torpedos Transmission Section Torpedo and Node Locations Design Variables for the Problem Method of feasible direction 32 34 34 36 38 39 40 43 43 46 48 5 9 10 15 16 24 25 25 29 31

vi

Figure 4.5 Figure 4.6 Figure 4.7 Figure 4.8 Figure 4.9

Pareto Optimization Curve Iteration History for Sound Iteration History for Frequency Iteration History for Mass Changes in Design Variables with Iterations

49 50 50 51 51

vii

LIST OF TABLES Table 2.1 Table 2.2 Table 2.3 Table 2.4 Table 4.1 Table 4.2 Table 4.3 Dimensions of MK-48 Lightweight Torpedo Structural Frequencies Coupled Structural Frequencies Frequencies of the Fluid Model Noise Levels with Infinite Boundary Condition Noise Levels without Infinite Boundary Condition Torpedo Optimal Configuration 9 24 26 26 44 45 52

viii

Acknowledgement

I would like to thank my advisor Dr. Ravi Penmetsa for his guidance and support and for giving me opportunity to work on this project. He was instrumental in directing me towards successful completion of this thesis. My sincere thanks go to Dr. Ramana Grandhi for his constructive comments and encouragement. Also, a special mention has to be made about efforts taken by Dr. Vipperla Venkayya throughout this research work. His eagerness to guide me in proper direction is highly appreciated. I would like to express my gratitude to Dr Klingbeil for being a part of thesis committee. I was well supported by all the members at Computational Design and Optimization Center (CDOC), Wright State University. These are wonderful people to work with. Apart from that, my friends, especially, Mayur, Nikhil, Ajay, Savio, Arun, Justin, Milind and Prithvi made my stay at Dayton enjoyable and deserve an appreciation. I would like to thank Brandy Foster for her help in making this document grammatically correct and readable. Finally, I would like to take this opportunity to thank my parents and my elder sister without whose efforts it would have been impossible for me to come to USA for studies.

ix

I would like to dedicate this thesis to, My parents Rekha and Bhagwan Khambaswadkar

1. Introduction
A torpedo is an underwater missile that can be launched from a submarine, a ship, or an aircraft. It is a highly sophisticated weapon whose optimal design requires satisfying multiple conflicting criteria that are equally important. Since every torpedo has numerous subsystems that produce easily detectable noise, the acoustic signature of a torpedo becomes one of the critical design criteria. The early detection of a torpedo gives the target time to take the necessary countermeasures to avoid the assault and reduces the effectiveness of the torpedo as a weapon. Therefore, when performing design optimization of a torpedo, it is important to ensure that the torpedos acoustic signatures are below the detectable range of certain SONAR systems. The self-generated noise typically increases with the speed of the torpedo and is extremely undesirable. Noise produced by the torpedo can have other detrimental implications as well: It may damage or interfere with the smooth operation of different electronic sensors inside the torpedo itself, which in turn will have an effect on the guidance and control of the torpedo. Furthermore, the amplification of structural vibrations due to this sound might result in fatigue of the panels. Finally, noise produced by torpedoes that is within a specific frequency can be of concern to the sea life. Due to all of these reasons, structural design of a torpedo subject to acoustic constraints is required for improved stealth characteristics.

In this research, a computational finite element model of a lightweight torpedo is developed that has longitudinal and radial stiffeners to provide additional strength to the shell. Since these components are absent in most of the conventional torpedoes, the current model needs to be analyzed and optimized to meet various design requirements. Due to these structural modifications, it is also important to investigate how structureborn noise is transmitted to the fluid, which involves solving a fluid-structure interaction problem.

1.1 Literature Review: The problem involving the interaction of an elastic structure with fluid has been of primary interest to many researchers due to its wide applicability. The problems that can be associated with this phenomenon can be categorized into exterior and interior applications. The exterior problems are those in which the sound propagation is exterior to the structure, such as the sound produced by a vibrating cylinder placed in fluid, which involves the determination of radiated and scattered noise. The interior problems are associated with acoustic cavities, piping systems, and other applications in which the sound is propagated within the structure. These problems have applications in ship noise reduction, acoustic analysis of a car interior, vibration response of underwater structures, blast analysis, etc. Many different formulations were proposed to solve these problems and studies have been conducted to see the relative trade-offs between these formulations by many researchers. The methods available in the literature are: boundary element [1-2], finite element [3-14], coupled boundary-finite elements [2], energy finite elements [15-16], and

various decoupling approximations [17-18], to name a few. The numerical modeling schemes used to model the fluid differentiate these methods from one another. The usefulness of each of these formulations is highly problem-dependent, and their availability or the users experience with a particular kind of tool become significant factors. Methods involving boundary elements generally use the dynamic response of the structure as input to a boundary element code, which is used to obtain the far field acoustic response in the fluid domain [1]. The decoupling approximation methods decouple the structural response from the fluid response and can reduce computational complications involved with solving coupled equations [18]. Many researchers, including Zienkiewicz and Newton [3], were instrumental in initiating efforts towards the successful use of finite elements to solve structural acoustic problems. Everstine, Marcus et al, [4, 6], continued on the same research and formulated methods that use the capabilities of the finite element code NASTRAN to solve the fluid-structure interaction problems. Everstine summarized different finite element-based formulations to solve structural acoustics problems [7]. Finite element-based methods have the advantage that they can use the matrix capabilities of sophisticated commercial codes, which are easily accessible and have sophisticated visualization capabilities. In the current version of NASTRAN, the pressure analog method developed by Everstine [4, 9] is implemented in the acoustic module. This method uses solid finite elements to represent scalar fluid fields by modifying the material properties so that they represent fluid. This method uses an analogy between equations of elasticity for structure and acoustic wave equations, which will be discussed in detail in later chapters. In this thesis, the finite element method is used to solve an

acoustic radiation problem for a lightweight torpedo structural model. This method is also used to model the noise source that would result in gear noise characteristics similar to the experimental data.

1.2 Project Approach: The main objective of this study is to minimize this structure-born noise in a lightweight torpedo through the modification of structural parameters. In order to study the sound radiation from the lightweight torpedo structure, a noise source needs to be modeled that has the same characteristics as the experimental data available for the gear machinery.

Fluid
FEA Modeling

Experimental Data

Structure

Optimization Based Source Modeling

Modal Analysis

Noise Modeling

Acoustic Analysis Modified Structure

Minimize: Mass Subject to: Constraints

Frequency

Sound Is Design Optimum? Optimum structure

Figure 1.1 Flowchart of Research Approach

The design methodology is divided into different units that are identified in Figure 1.1. The first important step in the process is the modeling phase. The proposed computational model of the torpedo has a shell structure that is supported with ring and longitudinal stiffeners. These stiffeners provide additional stiffness to the structure with minimal increase in weight. The fluid surrounding the structure is also modeled using

finite elements by using an analogy between equations of elasticity and the acoustic wave equation. In the fluid-structure interaction, structural displacements cause variations in fluid pressure and these variations in turn affect the structural behavior. The coupling effect becomes more significant when modal frequencies for the structure and fluid are similar. Frequency analysis is performed to ensure that the structure-borne noise is not amplified due to matching of the fluid and structural frequencies. The results of this eigenvalue analysis provide information about which structural frequencies to avoid while redesigning the torpedo. The next task is the modeling of the source that produces structural vibrations, which result in pressure variations in the fluid. This pressure distribution around the torpedo is its acoustic signature subject to the noise that is modeled. There are many sources of noise that excite the torpedo; however, the current research effort is targeted towards modeling noise due to speed reduction machinery, such as gears, in the torpedo. Even though this is not the most critical noise source, it is selected because only its experimental data is available in the public literature [22]. The goal of this research is to develop a methodology to obtain computational noise sources that produce similar characteristics as the experimental data. An optimization-based problem formulation is used to generate a computational model of the experimental noise data for gears. The fluid model represents the sea, which is modeled as an infinite domain. To represent the infinite nature of the water model, the finite element model of the fluid is truncated at a certain distance from the structure and a doubly asymptotic approximations-based radiation boundary condition is applied on the

outer surface of the fluid to ensure that there are no reflections back into the fluid from the boundaries. Once the modeling of the structure, fluid, and source is completed, the frequency response analysis, with the modeled source as excitation is performed. Finally, a multidisciplinary design optimization problem is formulated to reduce structural mass with the frequency and acoustics response as constraints. A constraint that reduces noise generated by the torpedo will result in increased weight. And the amount of this increment depends on the desired reduction of sound. Therefore, a multi-objective optimization problem is solved to obtain the Pareto frontier that clearly shows the tradeoff between the weight and sound produced. The following chapters will discuss these individual tasks in detail. Chapter 2 discusses the finite element models used to model the fluid and structure that was used in this research. Chapter 3 discusses the noise sources and the methodology developed to determine the computational source that would simulate experimental data. Chapter 4 discusses the multidisciplinary design optimization problem formulation and the Pareto frontier results obtained.

2. Finite Element Modeling of Torpedo, Fluid, and Noise Source

2.1 Modeling of Lightweight Torpedo A torpedo structure is essentially a cylindrical shell. The dimensions for the torpedo configuration used in this research are based on the data available in the public literature about the lightweight torpedo. The diameter of 0.32 m and a length of 2.42 m are selected for the structure. The nose and tail sections of the torpedo are tapered cones with lengths of 0.12 m and 0.35 m, respectively. Longitudinal and ring stiffeners are placed along the torpedo shell to improve structural rigidity. The width and thickness of these stiffeners is taken as 0.015 m and 0.01 m, respectively. The thickness for the torpedo shell is assumed to be 0.00635 m. Table 2.1 shows the dimensional parameters of the torpedo structure. Figure 2.1 shows the solid model of the torpedo with the longitudinal and radial stiffeners. The stiffeners in this research are modeled without specifying any offset from the nodes that are used to model the shell elements. The fundamental frequency of the torpedo structure has a one percent variation from the offset stiffeners model, at the configuration shown in Table 2.1. This variation is considered to be negligible compared to the complexities that will be introduced in the optimization problem in which the offsets need to be adjusted in each iteration, based on the shell thickness.

Overall Length Body Diameter Nose Length Tail Length Shell Thickness Stiffener Width Stiffener Thickness

2.42 m 0.32 m 0.12 m 0.35 m 0.00635 m 0.015 m 0.01 m

Table 2.1: Dimensions of MK-48 Lightweight Torpedo

Longitudinal Radial

Figure 2.1 Dimensions of Radial and Longitudinal Stiffeners The shell surface is modeled using quadrilateral and triangular plate elements and the circular rings and longitudinal stiffeners are modeled with bar/beam elements. The surface plate elements are called QUAD4 or TRIA3 elements for quadrilateral or triangular elements. The torpedos auxiliary equipments, warhead, and fuel contribute to

its total mass. These non-structural components are modeled using mass elements distributed along the torpedo body at nodal locations. The overall mass of the torpedo structure is assumed to be 254 Kg. The material chosen for the torpedo structure is aluminum-2024. Initial analysis suggests that the finite element model of the nose was introducing local modes that restricted the determination of the actual axial, bending, and breathing modes of the torpedo structure. These local modes were clearly dominated by the nose model rather than the torpedo structure. Therefore, in order to capture the behavior of the entire structure, rigid elements were introduced in the torpedo nose to connect the tip of the torpedo to the first section of the main body of the torpedo. These rigid elements provide enough strength to the otherwise flimsy nose structure to capture the global vibration characteristics of the torpedo.

Longitudinal Stiffeners

Ring Stiffeners

Figure 2.2 Finite Element Model of Lightweight Torpedo

10

Figure 2.2 shows the finite element model of the torpedo structure used in this research. The box-like structures distributed all along the model are the mass elements that represent the non-structural components of the torpedo. This non-structural mass becomes critical when optimizing the torpedo structure for frequency and frequencydependent responses.

2.2 Modeling of Fluid The three-dimensional fluid is modeled using the conventional solid finite elements available in NASTRAN. These solid finite elements are used to represent the fluid using an analogy between the equations of elasticity, which are generally used for solving structural problems, and the wave equation, which represents fluid acoustics [21, 24, 26]. The pressure analog method [4, 9] developed by Everstine is used in the current version on NASTRAN. The structural-acoustic analogy is discussed in detail here.

Equation of state: The equation of state for a fluid relates the internal restoring forces to the corresponding deformations. Therefore, in fluids terminology it relates pressure to condensation as:
p = Bs

(2.1)

where,
p = acoustic pressure or excess pressure at any point = p i - p 0

p i = instantaneous pressure at any point p 0 = constant equilibrium pressure in the field

11

B = adiabatic bulk modulus = 0 (pi / i )

0 = constant equilibrium density of the fluid i = instantaneous density at any point

s = condensation at any point = ( i - 0 )/ 0
Continuity Equation: This equation gives the functional relationship between particle velocity and the instantaneous density. The equation is as follows:

r + ( u ) = 0 t

(2.2)

Eulers Equation: This is a simple nonlinear inviscid force equation that is obtained using Newtons second law:
r r r u + (u )u = pi t

(2.3)

This equation becomes linear if it is assumed that changes in momentum are negligible, meaning that the second term inside the bracket vanishes.
r u 0 = pi t

(2.4)

The Homogeneous Linearized Wave Equation: The above three equations are combined to obtain a single differential equation with one dependent variable [21, 24]. Using equations (2.1), (2.2), (2.3) we get:

12

1 2 pi pi = 2 c t 2
2

(2.5)

Where,

c = phase speed for acoustics waves in the fluid =

Equation 2.5 represents a compressible, inviscid fluid whose pressure satisfies the above equation. In Cartesian co-ordinates the above equation can be written as:
x 1 p x + y 1 p y + z 1 p z 1 d2p = B dt 2

(2.6)

Elastic-Acoustic Analogy: In classical elasticity, the equation for equilibrium of stresses in the x direction is given as follows [27].

xx xy xz d 2u x + + = s x y z dt 2
Where, u x = structural displacement in x-direction

(2.7)

xx , xy , xz = stress components

s = structural mass density

Now stress-strain relation by Hooks law is given by:

xx G11 xy = G14 G xz 16

G14 G44 G46

G16 xx G46 xy G66 xz

(2.8)

where G ij is an element of a 6 by 6 anisotrpic elastic material matrix.

13

Since displacement components in y and z direction are zero the strain-displacement relationships becomes,

xx = xy xz

u x x u = x y u = x z

Since u x represents scalar pressure the strains are analogues to pressure gradients. Comparing equations (2.6) and (2.7) and taking, ux = p 1 B 1 p xx = x 1 p xy = y 1 p xz = z

s =

G11 = G44 = G66 =

G14 = G16 = G46 = 0

It can be shown that, with the above assumptions, Equation 2.6 and 2.7 are analogous. It is clear that the equations of elasticity represent the acoustic fluid wave equation with certain modifications in the properties of the elements. In NASTRAN, fluid properties entered on a MAT10 card are internally modified to obtain corresponding analogous structural properties that are used in matrix manipulations.

14

Fluid is modeled as a cylinder surrounding the torpedo structure. The sphere diameter of the cylinder is taken as 2 m and the overall length of the fluid domain is taken as 4m. Solid elements CTETRA are used to model the fluid with an element edge length of 0.004 m. Figure 2.3 shows the fluid domain modeled surrounding the torpedo structure.

Tetrahedral Solid Elements

2.42

4m Torpedo Structure

Figure 2.3 Fluid and Structural Finite Element Models

The interface between the fluid and structure may be modeled so that the grid points of the fluid are coincident with those of the structure. This is called a matching grid. But, because of the large size of the fluid domain and the irregular shape of the structure, it is difficult to obtain the matching grid. Therefore, free meshing is used to create the nonmatching fluid mesh around the structure. NASTRAN uses a method called Body in

15

White (BW) to determine wetted elements for fluid-structure interaction coupling. According to this method, a fluid node is coupled to a structural node for the load transfer based on the tolerances specified by the user. The method used for determining wetted nodes is called boxing technique. First step in the BW algorithm is to determine related structural and fluid nodes. Figure shows the fluid face and a box around in based on the tolerances provided on ACMODL card.

Figure 2.4 Tolerances for Fluid-Structure Interaction Here, L - The smallest edge length for the particular fluid element D - The distance from centre of the fluid face to the fluid node NORMAL, SKNEPS, INTOL The tolerances specified on ACMODL card. The structural nodes in the above mentioned box will be coupled to the fluid nodes of that specific fluid element. Once the fluid and structural faces have been determined, a face co-ordinate system is established for each fluid face. The resultant pressure force on each node on the fluid element is determined by the relation:

16

Ri = N f dS {pi = 1; p j = 0}i = 1, N Grid/Elem

s

A virtual work expression used to resolve this resultant pressure force for a unit grid pressure to grids of the fluid element is as follows:

{Fi } = {N f } N f dS {pi }
T s

The centre of pressure for the fluid face can be obtained using relations:
X pi =
Grids j

R (X
Fi
i

X0)

Y pi =

Grids j

R (Y
Fi
i

Y0 )

The resulting load distribution at the grids of the each of the structural element is calculated using rigid relations such that there is unit motion normal to fluid face with appropriate moment relationships. The area of each of the structural element is projected normal to fluid element face is used as weighing factor. The expression for load is as follows:

{F } = [W ][ R]([ R]
j

[W ][ R])

T 1

Ri 0 0

Here,

F j - Vector of resulting load distribution

W - Diagonal weighting matrix

R - Rigid transformation matrix

17

The forces at the structural grids are looped over at each of the fluid element grids. The same procedure is repeated for all the wetted elements. It was important to ensure that the fluid modeled is good enough to capture the fluidstructure interaction effect based on frequency of excitation. NASTRAN provides guidelines on how far the fluid should be modeled in order to be able to capture all the required effects. The rule states that six elements per wavelength are required in all directions from the structure for approximately 10% accuracy. So for 99% accuracy, approximately 60 elements are required per wavelength. In the current problem the excitation frequencies for the load is less than 100 Hz. So, the wavelength for this frequency is given by:

wavelength =

c f

where c = speed of sound in water, f = Frequency of excitation in Hz.

The wavelength is obtained as 14.5 m, which requires that 60 elements should be used in 14.5 m of fluid extent. For 2 m extent, approximately 8 elements are required in all the directions from the structure. Since the element edge length used is 0.04 m, the used mesh density is sufficient to capture the effect of the fluid-structure interaction.

18

2.3 Fluid-Structure Interaction (FSI): In general engineering problems, the effect of fluid-structure coupling is investigated under the following assumptions of the fluid state: 1. Compressible 2. Inviscid 3. Irrotational The compressibility assumption is required because the acoustic wave equation involves the speed of sound, which directly depends upon the bulk modulus of the fluid. For incompressible fluids, the bulk modulus is infinite, so the assumption of compressibility is required in the current analysis. Moreover, when deriving the boundary conditions for the wave equation, the viscous terms are neglected, so the assumption of inviscous fluid is justified. Since the small motion theory is assumed for the fluid particles, the rotational effects have very little chance of playing a role in the analysis, so irrotational assumption is required. NASTRAN uses a potential-based formulation to represent fluid finite elements. In these formulations, displacement remains the primary unknown for structural grid points whereas pressure is the primary unknown for fluid grid points. The standard equation of motion for the structure is given by:

&& [ M s ]{u s } + [ K s ]{u s } = Fs

(2.9)

19

where

M s and K s are the structural mass and stiffness matrices.

Since fluid pressure is applied to the structure as load, the force term on the right hand side can be divided into external forces on the structure and the fluid pressure on the interface nodes

Fs = Ps [ A]{ p} .

(2.10)

Here, the second term represents the load applied on the structure due to fluid pressure. The matrix [A] is called a coupling matrix, which is defined as follows: [ AT ] = {N f }N s dS .
S

(2.11)

Where,

N f and N s are fluid and structural shape functions, respectively,

and {p} is a vector of fluid pressure values at the interface grid points.

So the equation of motion for the structure becomes

&& [ M s ]{u s } + [ K s ]{u s } = {Ps } [ A]{ p} .

(2.12)

Also, the equation of motion for the fluid can be written as

[ M f ]{ &&} + [ K f ]{ p} = F f . p

(2.13)

20

Again, the force term on the right hand side can be divided into a fluid force, or acoustic disturbance in the fluid, and the force applied on the fluid grid points due to structural displacement.

&& F f = {Pf } + [ AT ]{u s }

(2.14)

Here, the second term represents the effect of structural displacement on the fluid grids at
&& the interface. And the vector {u s } is the vector of accelerations of the structural grid

points at the interface. So, the fluid equation of motion can be written as

&& [ M f ]{ &&} + [ K f ]{ p} = Pf + [ AT ]{u s } . p

(2.15)

By combining Equations (2.12) and (2.15), the coupled system of equations is obtained as
Ms AT 0 u s K s && && + 0 M f p A u s Ps = . K f p Pf

(2.16)

The above equations are unsymmetric and difficult to solve. But, Everstine developed a symmetric version of these equations by using the potential formulation as follows [23]. Let the velocity potential be defined as

& p =q.

21

And by substituting this in the structure Equation (2.12),

&& & [ M s ]{u s } + [ K s ]{u s } = {Ps } [ A]{q} .

(2.17)

By integrating the fluid equation with time and multiplying it by -1,

&& & [ M f ]{q} [ K f ]{q} = Pf dt [ AT ]{u s }

(2.18)

Now let G = Pf dt .

Now by combing Equations (2.17) and (2.18) the symmetric formulation of the fluidstructure coupled system is obtained as

M s 0

0 u s K s && && + 0 M f q

0 u s Ps = . K f q G

(2.19)

NASTRAN uses this equation by default to solve a coupled system. The output from this equation is expressed in terms of structural displacement and fluid pressure. It is to be noted that in all the above equations the damping terms are not included and they can be added to the equations, if there is damping present in the system. Therefore, the current approach will be a conservative model that can be improved by incorporating the structural damping.

22

2.4 Normal Mode Analysis Results:

One of the most significant analyses required in designing a torpedo is the modal analysis for determining the natural frequency of vibration. This is important not only to maintain the structural integrity of the torpedo, but also to ensure that the torpedo does not operate near certain critical frequencies. These frequencies could be those that would interfere with the onboard electronics, or those that would match with the fluid frequencies and amplify the noise generated, or others that are undesirable for safe operation of the torpedo. A modal complex eigenvalue analysis is performed to obtain structural frequencies, fluid frequencies, and coupled structural frequencies. Table 2.2 shows the first eight structural frequencies. Some of the frequencies are grouped due to the two axis of symmetry present in the current model. This table shows the structural frequencies with no added mass effects from the surrounding fluid. Table 2.3 shows the coupled structural frequencies where the added mass effect from the surrounding fluid can clearly be noticed. The frequencies are reduced compared to the structural model that was analyzed without the surrounding fluid.

23

Mode number 1 2 3 4 5 6 7 8

Frequency (Hz) 22.114 22.114 127.24 127.97 136.64 136.64 142.47 178.38

Table 2.2 Structural Frequencies

Figure 2.5 First Bending Mode

24

Figure 2.6 Second Bending Mode

Figure 2.7 Breathing Mode

25

Mode number 1 2 3 4 5 6 7 8

Frequency (Hz) 21.52 21.62 126.92 127.71 133.91 134.31 142.47 177.97

Table 2.3 Coupled Structural Frequencies

Mode number 1 2 3 4 5 6 7 8

Frequency (Hz) 26.39 52.84 58.74 63.00 63.00 67.73 68.31 79.42

Table 2.4 Frequencies of the Fluid Model

26

Table 2.4 shows the fluid frequencies that represent the vibration characteristics of the fluid model surrounding the torpedo. Since this model has two symmetric planes XY and XZ, the modes 4 and 5 are repeating modes. Similar trend was observed in NASTRAN verification problems [28]. Since the fluid elements have only pressure as degree of freedom, the mode shape information models the pressure distribution characteristics in the fluid. It can clearly be seen that the fluid and structural frequencies do not match in the current configuration. However, the first fluid frequency is close to the structural frequency and must be avoided during structural re-design.

27

3. Torpedo Noise Modeling

In order to determine the acoustic signature of a torpedo, a frequency response analysis is performed. A frequency-dependent force is used to excite the structure, which interacts with the fluid surrounding it to produce noise. In the literature, one can find numerous acoustic simulations in which the noise is modeled as a simple pulsating force with a wide range of frequencies. By using this forcing function, the sound produced by the torpedo is analyzed and minimized using structural sizing algorithms. The drawback of all of these techniques is the failure to realize that the response is entirely dependent on the spatial distribution of the forcing function and the frequency of excitation that is determined by the noise source used. Therefore, this chapter is dedicated to the accurate modeling of the noise source based on actual experimental data available. In the literature, no emphasis is placed on the modeling of the noise sources in this manner. In this research, an optimization-based formulation is used to model the noise source that will mimic the experimental data available through the literature for the lightweight torpedoes. The noise source thus modeled will be placed inside the torpedo structure for the multidisciplinary optimization of the torpedo.

28

3.1 Sources of Noise Generation:

Figure 3.1 lists some of the major noise sources in a torpedo based upon the information available in literature. These noises induce structural vibrations at different frequencies that are then transmitted to the fluid to generate noise.

Exhaust-wake interaction

Fans, Converters, Pumps etc. Guidance & Control

Propulsor Engine Assembly Noise Boundary Layer

Figure 3.1 Important Sources of Noise Generation in a Torpedo

Among these noise sources, the propulsor is the most critical source that is of interest to the U.S. Navy. However, due to a lack of experimental data available in the public literature, a less significant but important noise source - engine assembly noise is selected in this research. Experimental data for this noise source is available along with the details of the experimental setup. This comprehensive information about the data enabled the modeling of a computational setup that would mimic the experimental setup. In a torpedo, the transmission gears or the engine assembly is used in the speed reduction machinery to control the propeller angular velocity. These gears produce significant noise, despite their high precision manufacturing [22]. Gear noise comes from a variety of sources, as follows:

29

Roughness of tooth surface Deflection of gear tooth Eccentricity of tooth rotor Misalignment of tooth rotor Unbalance of tooth rotor Noise of rotor bearing Noise of flexible couplings Seal rub or squeal Vibrations transmitted by driving or driven equipments Most of these noise sources can be avoided, but during severe operating conditions some of these might result in noise generation. Moreover, in some instances the air trapped between the gear tooth will result in an amplified noise that would then be propagated through the structure. A program of experimental research was undertaken at the U. S. Naval Ordnance Test Station (NOTS) to improve the basic knowledge of gear noise transmission in torpedoes. The gears were considered as non-uniform point sources radiating into a sphere, and the total noise output was calculated by integrating data obtained at numerous locations. The results of these experiments give the noise profile generated by the transmission gears of a MK-40 lightweight torpedo. This MK-40 torpedo is the precursor to the currently operating lightweight torpedoes. In this research, an acoustic source that will emit a noise pattern similar to that of an experimental setup described below will be determined. An optimization-based problem formulation is used for designing a computational noise source model that will represent the experimental data.

30

3.2 Experimental Setup and Noise Profile:

The experimental test setup involves a gear assembly placed in an acoustically quiet chamber, with sensitive microphones placed at fixed radial distances from the transmission [22]. Figure 3.2 shows the details of the experimental setup.

Noise sensors

Transmission

Turbine

Dynamometer

Acoustic cavity Acoustically silent room Figure 3.2 Experimental Setup Used for Gear Noise

A steam turbine is used to rotate the transmission gears and is connected by long shafts so that the turbine noise does not influence the experiment results. The dynamometer is used to absorb the load and to measure the torque and speed. Here, the concept of spherical measurement is used to measure noise. Sensitive microphones placed at a distances of 0.32 m from the transmission are used as measuring points to collect information about the sound produced by the gear mechanism. The transmission-noise profile generated by the MK-40 lightweight torpedo captured by the above mentioned experimental setup is shown in Figure 3.3 [22].

31

116 dB 113 dB 110 dB 109 dB 110 dB 106 dB 103 dB 110 dB 110 dB

114 dB

Figure 3.3 Noise Levels on the Meridian of Hemisphere about the MK-40 Torpedo

Figure 3.3 clearly shows the nonlinear nature of the sound generated by the machinery noise. The spatial nonlinearity exhibited by the experimental data indicates that the previous attempts by the researchers to model the noise as a pulsating force at varying frequencies is inaccurate. This data is used to model a source on the axis of the torpedo model that can result in a pressure distribution similar to the experimental acoustic data. The pressure distribution obtained from NASTRAN is converted into the appropriate decibel level by using the analytical equations shown in the following section.

32

3.3 Optimization Formulation for Noise Source Modeling:

A finite element model of the air representing the hemisphere on which sensors are placed is modeled using solid finite elements. The use of these solid finite elements to represent air is possible because of an acoustic-elastic analogy discussed in the previous chapter. An acoustic load is applied at the center of the cavity to act as a simple noise generating source. This simple noise source can be imagined to generate a pulsating sphere in an infinite space. This source will emit noise in a spherical direction, the magnitude of which will depend on the strength and frequency of the source. This noise source is used as excitation in a frequency response analysis. The noise emission at certain key locations (Figure 3.5) in the air model that match the sensor locations in the experimental setup is monitored. Therefore, this analysis identifies the source strength that will give the exact same sound profile as the experimental result. The following figure shows the air model with the source placed at its center. The current air model has a diameter of 0.64 m whereas the torpedo has a diameter of 0.32 m. This difference occurs because the sensor locations in the experimental are at 0.64 m. Therefore, once the source strength and the frequency that would match the sound levels of the experimental setup is determined, this source would be placed in a similar air model within the torpedo internal cavity.

33

Figure 3.4 Finite Element Model of Air Chamber

Recovery points Noise Source

Y 5 6 3 2 1 Z X 4 7 8

Figure 3.5 Noise Recovery Points in the Air Chamber

34

The objective of the optimization problem is to minimize the squared error between the noise levels at particular locations in the air model and the experimental values. The design variables in the problem are the source strength of the acoustic load and the frequency of the source. The upper and lower bounds on the source strength and frequency are also applied as side bound constraints in the optimization problem. The optimization problem can be summarized as follows: Minimize

( Ai Bi )2
where

(3.1)

Ai = Noise value obtained from acoustic analysis at location i and Bi = Noise value from the literature at location i, subject to:

0.001 S 100.0
10 f 1000 ,

and

(3.2) (3.3)

where S is the source strength that is the design variable for the problem and f is the frequency of the source. The flowchart in Figure 3.6 explains the flow of the optimization algorithm. The Design Optimization Tool is used for the optimization [25].

35

Nastraninput.cpp Generate the NASTRAN input

Modified Source Strength and Frequency

NASTRAN

MATLAB Read NASTRAN Output and Calculates Error

DOT

Is Error Minimum ?

No

Yes
Final Source

Figure 3.6 Optimization Algorithm to Determine Source Strength

Since gradient-based search methods were used in the optimization iterations, the sensitivity of the objective to the source strength and frequency were required in this algorithm [25]. These sensitivities are calculated using the finite difference method executed using a series of function calls between MATLAB and a C++ program that is used to generate the NASTRAN model.

36

3.4 Results and Discussion Analytical Verification:

Before optimizing the torpedo for the minimum acoustic signature, it was important to verify the accuracy of the finite element simulation results for the noise levels in the air model. For the noise emitted at a certain distance by a simple noise source such as a pulsating sphere, approximate equations are available that give the source strength needed for a particular decibel level at a specified location [21]. Since this analytical equation is applicable for a constant sound profile at a distance from the source, the profile that is considered is constant strength at all the key locations. Once the NASTRAN and analytical results are verified, the experimental data can be matched using a similar approach. The NASTRAN acoustic source is a simple monopole source; therefore, the acoustic intensity radiated from this simple point source is given by the following equation:
I = W 4 r 2 , (3.4)

where r is the radius of the sphere in which the source radiates energy and W is source strength in watts. From the available experimental profile, it can be seen that 116 dB is the maximum sound emitted by the gear assembly. These decibels can be converted into intensity by using the relation
I = I o (10)

(dB10 )

(3.5)

37

I =

W 4r 2

Figure 3.7 Intensity of a Pulsating Point Source

And, by substituting this intensity into the above equation, the analytical source strength needed to produce 116 dB at 0.32 m from the source is obtained as follows
Watts = I 4r 2

(3.6)

For 116 dB, I = 3.981E 5 W/cm 2 Watts = 0.5122 W This source strength is given as input to NASTRAN and the noise generated by this source measured at 0.32 m from the center is analyzed. The finite element model results were compared to these approximate equations, and the deviation was 4% from the expected values. This validated the finite element setup to within the required accuracy. In this case, the optimization problem is solved such that the noise levels at all the desired locations are expected to be 116 dB. Figure 3.8 shows the difference between the

38

analytical and the NASTRAN results at various key locations. These key locations are the same as the sensor locations in the experimental setup.

4% Deviation
125 120
N AST AN R
Analytical

Sound in dB

115 110 105 100 95 90 1 2 3 4 5 6 7 8

Recovery Points Figure 3.8 Results for a Constant Profile Case

Matching the Exact Sound Profile:

The experimental noise profile that was shown earlier is highly nonlinear. In this case, the goal is to match the nonlinear profile as close as possible and then to use the obtained sources to determine the acoustic signature of the torpedo. Initial attempts to match the experimental data showed that it is not possible to match the nonlinear profile with only one design variable; i. e., only one acoustic source. Therefore, in order to match the profile exactly, more sources are distributed in the transmission section, which increases the number of design variables for the problem. By using many different combinations of source distributions in the transmission area and varying the frequency of the sources, the best fit for the data is obtained. From the optimization results it was clear that we needed two sources with source strengths 0.9 watts and 0.15 watts at 77.85 Hz. Figure 3.9 shows

39

the deviation between NASTRAN and the experimental results. The maximum deviation at a given key location is 4%. The current noise source model with a 4% deviation from the experimental data is more realistic than the traditional approaches that use pulsating forces to model the noise source. The optimization formulation that is used in this research is generic, and can be used for any experimental data that is available in the future. The general idea behind this whole effort is to use the source obtained from the optimization problem as a load in the proposed computational model of the lightweight torpedo for acoustic analysis. This will ensure that realistic data is used to model the source instead of applying random forces to excite the structure.
125 120 115

NAST RAN

Experimental

4% Deviation

Sound in dB

110 105 100 95 90 1 2 3 4 5 6 7 8

Recovery Points Figure 3.9 Variation Between Experimental and NASTRAN Results

40

4. Multidisciplinary Design Optimization (MDO) of Lightweight Torpedo

The torpedo body can be broadly divided in three sections: transmission, fuel and warhead, and guidance and control. The modeled noise source will be placed in the transmission section of the torpedo body, and will act as an excitation force in determining the frequency response of the torpedo. Figure 4.1 shows the source placed in the transmission section. An air chamber is modeled inside the transmission section using solid models and material properties that reflect air density and bulk modulus. This model has half the diameter as the air model used in the source modeling section, because of the available cavity size inside the torpedo. The source determined from the earlier analysis is placed in the appropriate location. The boundaries of the air model transverse to the axis of the torpedo are left free. This condition assumes no transmission of noise along the torpedo length through the rest of the cavity. The only transmission is through the structure. Therefore, the fluid-structure interaction conditions are very significant in an air-torpedo interface and a torpedo-water interface. The source inside the air cavity produces a pressure variation in the transmission section of the torpedo that will result in the displacement of the torpedo structure. This displacement will be transmitted into the water model resulting in a pressure distribution, which is the acoustic response of the torpedo. The fluid-structure model is analyzed to

41

verify the effect of the fluid-structure interaction on the results obtained. If the fluidstructure interaction effect is turned off in the analysis, then the sound intensity in the water is found to be zero, which indicates that the structural displacements are not transferred to the fluid model. As discussed in the modeling section of this document, the outer surface of the fluid has a radiation boundary condition that simulates the infinite nature of the fluid. In order to verify the validity of this boundary condition, two analyses, one with the radiation boundary condition and one without, were performed. The results from the two analyses can be seen in Figure 4.2 and Tables 4.1 and 4.2. It can be seen from the tables that the sound is reflecting back from the surface in the case in which there is no absorbing boundary condition. Also, with an increase in the distance from the source, the noise should reduce. This is clearly seen from the radiation boundary condition case, but this trend not very evident in the case without a radiation boundary condition. Therefore, the infinite boundary condition is a critical component of any underwater acoustic analysis. When one has to simulate reflections from the ocean floor, one side of the fluid can be modeled without the infinite boundary conditions.

42

Air Mesh

Transmission

Fuel, warhead, and guidance and control Acoustic Source

Figure 4.1 Air Mesh inside Torpedos Transmission Section.

14 13 12 1 2 3 11

Vibrating Torpedo 4 7 8 9 10 5 6

Water Medium

Figure 4.2 Torpedo and Node Locations

43

Node Location 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Node Number 1514 5158 4592 5095 4716 1501 5771 5018 4642 1808 5786 5101 4741 1763

Node Co-ordinates in m Sound in dB X -1.09 -0.82 -0.37 2.23 2.77 2.91 0.32 0.32 0.32 0.32 0.32 0.32 0.32 0.32 Y 0 0.09 0 0 0 0 -0.621 -0.735 -0.86 -0.995 0.621 0.735 0.86 0.995 Z 0 0 0 0 0 0 0 0 0 0 0 0 0 0 38.47 49.39 59.38 74.81 64.15 53.02 66.56 65.05 61.44 50.77 66.94 64.54 60.25 49.36

Table 4.1 Noise Levels with Infinite Boundary Condition

44

Node Location 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Node Number 1514 5158 4592 5095 4716 1501 5771 5018 4642 1808 5786 5101 4741 1763 X

Node Co-ordinates Sound in dB Y 0 0.09 0 0 0 0 -0.621 -0.735 -0.86 -0.995 0.621 0.735 0.86 0.995 Z 0 0 0 0 0 0 0 0 0 0 0 0 0 0 88.91 88.99 89.23 81.41 81.37 81.38 87.36 87.22 87.13 87.09 88.941 88.944 88.945 88.945 -1.09 -0.82 -0.37 2.23 2.77 2.91 0.32 0.32 0.32 0.32 0.32 0.32 0.32 0.32

Table 4.2 Noise Levels without Infinite Boundary Condition

45

4.1 Optimization Formulation:

The final objective of this project is to determine the optimum configuration of the torpedo that would have minimum noise propagated into the surrounding water. To achieve this, an optimization problem is formulated as follows:

Minimize: Mass of the structure Subject to: Sound level at certain location < 70 dB Natural frequency of the torpedo >= 23 Hz The structural parameters, such as the thickness of different sections of the shell, the cross-sectional width, and the height of the ring and longitudinal stiffeners, were used as design variables for the optimization problem. Figure 4.3 shows these design variables.

L T LW

R H

Radial Stiffeners

RW

Torpedo Shell

Longitudinal Stiffeners

Figure 4.3 Design Variables for the Problem The optimum design is one that has minimum mass; however, this would mean that the sound signatures from the structure are increased to meet the requirements. This is

46

obvious because as the mass is reduced, the shell thickness and dimensions of the stiffeners decreases, which results in increased noise. Therefore, a realistic solution for this problem will provide a trade-off analysis between the weight and sound levels produced by the source.

4.2 Optimization in NASTRAN:

NASTRAN has in built design optimization capabilities and most of it comes from Design Optimization Tools (DOT) which is customized to run with NASTRAN. The optimization algorithms that NASTRAN uses are gradient-based methods. The finite difference method is used for gradient calculation here. In this thesis, for the acoustic optimization problem method of feasible directions, which is a default method for NASTRAN, is used. Next section briefly explains this method.
Feasible Directions Method:

Design Optimization Tools (DOT) uses method of feasible directions which is very popular method used for constrained optimization problems. The basic idea of the method is to move from one feasible design to another improved feasible design by taking small steps. The method tries to keep the design away from boundaries as much as possible.
s
T

> 0

gj critical constraints

sT f = sTg < 0

f Objective function

47

g2=0

g1=0

g1 g2
s

x -f

Figure 4.4 Method of feasible direction The first condition is the feasible direction condition. According to this condition, the search direction S shown in figure 4.6 should be such a way that a small step in along it should not make the design infeasible. The second condition is called usability condition. According to this condition the search direction should be such way that the objective should be reduced if it a minimization problem. Based on the above two conditions a compromise is defined by following maximization problem: Maximize Subject to - s T g j + j 0 s T f + 0 si 1 Where, j - Push-off factors Once the feasible usable search direction is obtained from above optimization problem, a line search is performed to determine how far to proceed along the obtained search j 0

48

direction. This leads to another feasible design and the process is repeated till convergence.
4.3 Optimization Results and Discussion:
360

Optimized Mass in Kg

340 320 300 280 260 240 62 64 66 68 70 72 74 76

Sound in dB

Figure 4.5 Pareto Optimization Curve This trade-off analysis can be seen in the Pareto frontier shown in Figure 4.4. This figure shows how the reduction in sound level increases the weight of the structure. Based on the weight requirements of the torpedo, an appropriate sound level can be determined from this plot along with the corresponding configuration for the thickness and crosssection of the stiffeners, which are available from previous optimization solutions.

49

73.50 73.00 72.50 72.00

Sound in dB

71.50 71.00 70.50 70.00 69.50 69.00 68.50 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Iteration Number Figure 4.6 Iteration History for Sound

27.00 26.00

Frequency in Hz

25.00 24.00 23.00 22.00 21.00 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Iteration Number Figure 4.7 Iteration History for Frequency

50

280.00 275.00

Mass in Kg

270.00 265.00 260.00 255.00 250.00 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Iteration Number

Figure 4.8 Iteration History for Mass

Ring Width
Shell Thickness
Long. Height

Ring Height
Long. width

Design Variables in mm

25 20 15 10 5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Iteration Number

Figure 4.9 Changes in Design Variables with Iterations

51

Longitudinal Shell Thickness (m) 0.0091 Mass (Kg) 274.47 Ring Width (m) 0.005 Ring Height Stiffener (m) Width (m) 0.005 0.022

Longitudinal Stiffener Height (m) 0.013 Frequency (Hz) 25.76

Sound (dB) 70.00

Table 4.3 Torpedo Optimal Configuration Figures 4.5, 4.6, and 4.7 show the histories of the objective and constraints as the optimization iteration progresses. The initial values for ring and longitudinal stiffener widths are given as 0.015 m, and their thicknesses are taken as 0.01 m, respectively. The initial shell thickness is taken as 0.0635 m. An inverse relation between the sound and mass of the structure is evident from these plots. Figure 4.8 shows the variation in all the design variations with optimization iterations. From this figure, it is clear that the optimizer is pushing the ring stiffener dimensions to the lower limits and the shell thickness and longitudinal stiffener dimensions are increased to reduce noise. From Figure 4.8, it can be observed that shell thickness is the most important design variable here. Table 4.3 shows the optimal configuration of the torpedo structure from one of the several optimization runs required to get the Pareto frontier. The table also shows the weight of the structure and the corresponding sound level at a critical location. This critical location was determined for one particular configuration, and kept constant in order to have a continuous function definition for all of the iterations in the optimization

52

problem. In reality, as the structural model is changed, the location of maximum sound intensity changes. However, it is assumed in this research that if the intensity at the fixed critical location is reduced, then the intensity at other locations is also reduced which was verified to be true.

53

5. Concluding Remarks
In this research, an acoustic optimization methodology is presented for a computational model of a lightweight torpedo using the finite element method to model both the fluid and the structure. Fluid and structural models are coupled to incorporate the effect of the fluid-structure interaction. As it can be seen from the numerical results, the fluid-structure interaction and the infinite boundary conditions are critical for the acoustic analysis of underwater structures. This research has shown that the noise profile generated by the gear machinery demonstrates spatial nonlinearity, which cannot be represented by the pulsating force models used by many researchers. Therefore, experimental results and the corresponding computational noise source models are very important for determining the acoustic signature of torpedo structures. The optimization problem solved in this work gives the relative trade-off between the mass of the structure and the sound emitted by it due to gear noise. Future work in this area can be directed towards acoustic analysis and optimization of composite structures based on the source model developed in this research. With the introduction of composite models, active and passive damping techniques can be explored through embedded systems in the torpedo shell. However, before the development of damping technologies is possible, all other sources need to be modeled and incorporated into the torpedo model. This can only be done through experimental

54

data collection for noise from the various sources, as discussed earlier. Finally, this thesis outlines the steps involved in the acoustic design of an underwater vehicle with a realistically modeled noise source.

55

APPENDIX
Appendix A

This appendix has C++ code to generate structural model of lightweight torpedo in the NASTRAN format.
#include<iostream.h> #include<stdlib.h> #include<stdio.h> #include<math.h> void main() { char outfile[]="Torpedowithoffset.dat"; int j=0,k=0; FILE *stream2; stream2=fopen(outfile,"w"); //Executive Control fprintf(stream2,"%s\n","ID"); fprintf(stream2,"%s\n","SOL 103"); fprintf(stream2,"%s\n","CEND"); fprintf(stream2,"%s\n","SPC=20"); fprintf(stream2,"%s\n","METHOD = 15"); fprintf(stream2,"%s\n","BEGIN BULK"); fprintf(stream2,"%s%11d%24d\n","EIGRL",15,20); int n_x, n_y,n_t, n_r, x,y,r, N1; double Dia, Len, Nose, th, xc,ti,h,ofs; Dia=0.32; //Diameter of body Len=2.0; //Length of body Nose=0.16; //Length of nose n_x=40; //Number of rings(node rings) on body n_y=3; //Number of rings(node rings) on nose n_t=7; //Number of rings(node rings) on tail n_r=24; //Number of nodes on each ring ti=0.00635; //Thickness of shell h=0.010; //Height of ring stiffeners ofs=(ti/2)+(h/2); //offset for ring stiffners double GridCoordX[10000],GridCoordY[10000],GridCoordZ[10000], theta, X_coord; theta=360.0/24.0; N1=0; X_coord=2.00/40.0; double p = 3.1416; char Grid[]="GRID"; int GridNum=1; xc=1.0; // rear tip node GridCoordX[N1]=-0.30; GridCoordY[N1]=0.0; GridCoordZ[N1]=0.0; fprintf(stream2,"%s%12d%16.2f%8.2f%8.2f\n",Grid,GridNum,GridCoordX[N1],GridCoordY[N1],GridCoordZ[N1]); GridNum=GridNum+1; N1=N1+1; // Tail section //-30 th=1.0; for(r=1;r<=n_r;r++) { GridCoordX[N1]=-0.30; GridCoordY[N1]=0.5*0.18*cos(th*theta*p/180.0);

56

GridCoordZ[N1]=0.5*0.18*sin(th*theta*p/180.0); fprintf(stream2,"%s%12d%16.4f%8.4f%8.4f\n",Grid,GridNum,GridCoordX[N1],GridCoordY[N1],GridCoordZ[N1]); GridNum=GridNum+1; N1=N1+1; th=th+1.0; } //-25 th=1.0; for(r=1;r<=n_r;r++) { GridCoordX[N1]=-0.25; GridCoordY[N1]=0.5*0.20*cos(th*theta*p/180.0); GridCoordZ[N1]=0.5*0.20*sin(th*theta*p/180.0); fprintf(stream2,"%s%12d%16.4f%8.4f%8.4f\n",Grid,GridNum,GridCoordX[N1],GridCoordY[N1],GridCoordZ[N1]); GridNum=GridNum+1; N1=N1+1; th=th+1.0; } //-20 th=1.0; for(r=1;r<=n_r;r++) { GridCoordX[N1]=-0.20; GridCoordY[N1]=0.5*0.22*cos(th*theta*p/180.0); GridCoordZ[N1]=0.5*0.22*sin(th*theta*p/180.0); fprintf(stream2,"%s%12d%16.4f%8.4f%8.4f\n",Grid,GridNum,GridCoordX[N1],GridCoordY[N1],GridCoordZ[N1]); GridNum=GridNum+1; N1=N1+1; th=th+1.0; } //-15 th=1.0; for(r=1;r<=n_r;r++) { GridCoordX[N1]=-0.15; GridCoordY[N1]=0.5*0.24*cos(th*theta*p/180.0); GridCoordZ[N1]=0.5*0.24*sin(th*theta*p/180.0); fprintf(stream2,"%s%12d%16.4f%8.4f%8.4f\n",Grid,GridNum,GridCoordX[N1],GridCoordY[N1],GridCoordZ[N1]); GridNum=GridNum+1; N1=N1+1; th=th+1.0; } //-10 th=1.0; for(r=1;r<=n_r;r++) { GridCoordX[N1]=-0.10; GridCoordY[N1]=0.5*0.26*cos(th*theta*p/180.0); GridCoordZ[N1]=0.5*0.26*sin(th*theta*p/180.0); fprintf(stream2,"%s%12d%16.4f%8.4f%8.4f\n",Grid,GridNum,GridCoordX[N1],GridCoordY[N1],GridCoordZ[N1]); GridNum=GridNum+1; N1=N1+1; th=th+1.0; } //-5 th=1.0; for(r=1;r<=n_r;r++) { GridCoordX[N1]=-0.05; GridCoordY[N1]=0.5*0.28*cos(th*theta*p/180.0);

57

GridCoordZ[N1]=0.5*0.28*sin(th*theta*p/180.0); fprintf(stream2,"%s%12d%16.4f%8.4f%8.4f\n",Grid,GridNum,GridCoordX[N1],GridCoordY[N1],GridCoordZ[N1]); GridNum=GridNum+1; N1=N1+1; th=th+1.0; } //0 th=1.0; for(r=1;r<=n_r;r++) { GridCoordX[N1]=0; GridCoordY[N1]=0.5*0.30*cos(th*theta*p/180.0); GridCoordZ[N1]=0.5*0.30*sin(th*theta*p/180.0); fprintf(stream2,"%s%12d%16.4f%8.4f%8.4f\n",Grid,GridNum,GridCoordX[N1],GridCoordY[N1],GridCoordZ[N1]); GridNum=GridNum+1; N1=N1+1; th=th+1.0; } // Main body for(x=1;x<=n_x;x++) { th=1.0; for(r=1;r<=n_r;r++) { GridCoordX[N1]=xc*X_coord; GridCoordY[N1]=0.5*Dia*cos(th*theta*p/180.0); GridCoordZ[N1]=0.5*Dia*sin(th*theta*p/180.0); fprintf(stream2,"%s%12d%16.4f%8.4f%8.4f\n",Grid,GridNum,GridCoordX[N1],GridCoordY[N1],GridCoordZ[N1]); GridNum=GridNum+1; N1=N1+1; th=th+1.0; } xc=xc+1.0; } // Nose section X_coord=0.16/4.0; xc=1.0; for(y=1;y<=n_y;y++) { th=1.0; for(r=1;r<=n_r;r++) { GridCoordX[N1]=xc*X_coord+2.0; GridCoordY[N1]=0.5*Dia*0.25*(4.0-xc)*cos(th*theta*p/180.0); GridCoordZ[N1]=0.5*Dia*0.25*(4.0-xc)*sin(th*theta*p/180.0); fprintf(stream2,"%s%12d%16.4f%8.4f%8.4f\n",Grid,GridNum,GridCoordX[N1],GridCoordY[N1],GridCoordZ[N1]); GridNum=GridNum+1; N1=N1+1; th=th+1.0; } xc=xc+1.0; } // Tip Node GridCoordX[N1]=Len+0.12; GridCoordY[N1]=0.0; GridCoordZ[N1]=0.0; fprintf(stream2,"%s%12d%16.4f%8.4f%8.4f\n",Grid,GridNum,GridCoordX[N1],GridCoordY[N1],GridCoordZ[N1]); GridNum=GridNum+1; N1=N1+1;

58

// Nodes for the orientation of the rings(GO entry in the nastran card) int nor; double R; R=0.1; for (nor=1;nor<=13;nor++) { GridCoordX[N1]=R; GridCoordY[N1]=0; GridCoordZ[N1]=0; fprintf(stream2,"%s%12d%16.4f%8.4f%8.4f\n",Grid,GridNum,GridCoordX[N1],GridCoordY[N1],GridCoordZ[N1]); GridNum=GridNum+1; N1=N1+1; R=R+0.15; } int ElmNum=1; char ELEM[]="CQUAD4"; char PBAR[]="PBAR"; char plus[]="+"; char MAT1[]="MAT1"; char PSHELL[]="PSHELL"; j=2; // Rear tria3 elements char ELEM1[]="CTRIA3"; for(r=1;r<n_r;r++) { fprintf(stream2,"%s%10d%8d%8d%8d%8d\n",ELEM1,ElmNum,1,j,j+1,1); ElmNum=ElmNum+1; j=j+1; } fprintf(stream2,"%s%10d%8d%8d%8d%8d\n",ELEM1,ElmNum,1,j,j-n_r+1,1); ElmNum=ElmNum+1; int pq,pr; pr=n_x+n_y+n_t; pq=(n_x+n_y+n_t)/3; j=2; for(x=1;x<pq;x++) { for(r=1;r<n_r;r++) { fprintf(stream2,"%s%10d%8d%8d%8d%8d%8d\n",ELEM,ElmNum,1,j,j+1,j+n_r+1,j+n_r); ElmNum=ElmNum+1; j=j+1; } fprintf(stream2,"%s%10d%8d%8d%8d%8d%8d\n",ELEM,ElmNum,1,j,j-n_r+1,j+1,j+n_r); ElmNum=ElmNum+1; j=j+1; } for(x=pq;x<2*pq;x++) { for(r=1;r<n_r;r++) { fprintf(stream2,"%s%10d%8d%8d%8d%8d%8d\n",ELEM,ElmNum,4,j,j+1,j+n_r+1,j+n_r); ElmNum=ElmNum+1; j=j+1; } fprintf(stream2,"%s%10d%8d%8d%8d%8d%8d\n",ELEM,ElmNum,4,j,j-n_r+1,j+1,j+n_r); ElmNum=ElmNum+1; j=j+1; } for(x=2*pq;x<pr;x++) { for(r=1;r<n_r;r++) {

59

fprintf(stream2,"%s%10d%8d%8d%8d%8d%8d\n",ELEM,ElmNum,3,j,j+1,j+n_r+1,j+n_r); ElmNum=ElmNum+1; j=j+1; } fprintf(stream2,"%s%10d%8d%8d%8d%8d%8d\n",ELEM,ElmNum,3,j,j-n_r+1,j+1,j+n_r); ElmNum=ElmNum+1; j=j+1; } //Nose tip elements for(r=1;r<n_r;r++) { fprintf(stream2,"%s%10d%8d%8d%8d%8d\n",ELEM1,ElmNum,3,j,j+1,1202); ElmNum=ElmNum+1; j=j+1; } fprintf(stream2,"%s%10d%8d%8d%8d%8d\n",ELEM1,ElmNum,3,j,j-n_r+1,1202); ElmNum=ElmNum+1; //Ribs char ELEM2[]="CBAR"; char BAR[]="PBAR"; j=194;//281 int a1=1203; float a,b,c,m,d,e,f; float p1,q,r1,m1,s,t,u; float X=0.10; float Y=0; float Z=0; for(x=1;x<n_x;) { for(r=1;r<n_r;r++) { a=GridCoordX[j-1]-X; b=GridCoordY[j-1]-Y; c=GridCoordZ[j-1]-Z; m=sqrt(a*a+b*b+c*c); d=-ofs*a/m; e=-ofs*b/m; f=-ofs*c/m; p1=GridCoordX[j]-X; q=GridCoordY[j]-Y; r1=GridCoordZ[j]-Z; m1=sqrt(p1*p1+q*q+r1*r1); s=-ofs*p1/m1; t=-ofs*q/m1; u=-ofs*r1/m1; fprintf(stream2,"%s%12d%8d%8d%8d%8d\n%32.5f%8.5f%8.5f%8.5f%8.5f%8.5f\n",ELEM2,ElmNum,1,j,j+1,a1,d,e,f,s,t,u); ElmNum=ElmNum+1; j=j+1; } a=GridCoordX[j-1]-X; b=GridCoordY[j-1]-Y; c=GridCoordZ[j-1]-Z; m=sqrt(a*a+b*b+c*c); d=-ofs*a/m; e=-ofs*b/m; f=-ofs*c/m; p1=GridCoordX[j-n_r]-X; q=GridCoordY[j-n_r]-Y; r1=GridCoordZ[j-n_r]-Z; m1=sqrt(p1*p1+q*q+r1*r1); s=-ofs*p1/m1; t=-ofs*q/m1; u=-ofs*r1/m1; fprintf(stream2,"%s%12d%8d%8d%8d%8d\n%32.5f%8.5f%8.5f%8.5f%8.5f%8.5f\n",ELEM2,ElmNum,1,j,j-n_r+1,a1,d,e,f,s,t,u); ElmNum=ElmNum+1; j=j+49;//49;

60

x=x+3.0;//3 X=X+0.15;//15 a1=a1+1; } //longitudinal bars double J[30]; int i; J[0]=196.0; J[1]=199.0; J[2]=202.0; J[3]=205.0; J[4]=208.0; J[5]=211.0; J[6]=214.0; J[7]=217.0; Y=0; Z=0; for (i=0;i<8;i++) { j=J[i]; X=0.10; for (x=1;x<(n_x)-1;x++) { k=j+24; a=GridCoordX[j-1]-X; b=GridCoordY[j-1]-Y; c=GridCoordZ[j-1]-Z; m=sqrt(a*a+b*b+c*c); d=-ofs*a/m; e=-ofs*b/m; f=-ofs*c/m; X=X+0.05; p1=GridCoordX[k-1]-X; q=GridCoordY[k-1]-Y; r1=GridCoordZ[k-1]-Z; m1=sqrt(p1*p1+q*q+r1*r1); s=-ofs*p1/m1; t=-ofs*q/m1; u=-ofs*r1/m1; fprintf(stream2,"%s%12d%8d%8d%8d%8d\n%32.5f%8.5f%8.5f%8.5f%8.5f%8.5f\n",ELEM2,ElmNum,2,j,k,1209,d,e,f,s,t,u); ElmNum=ElmNum+1; j=j+24; } } double C[10]; C[0]=0.45; C[1]=1.25; C[2]=1.95; int g,G,pid,lp,N; double NR; NR=0.10; G=196; N=1; char CONM[]="CONM3"; for (g=0;g<3;g++) { while(NR<C[g]) { for(lp=1;lp<=8;lp++) { fprintf(stream2,"%s%11d%8d%16.4f\n",CONM,ElmNum,G,1.9321); ElmNum=ElmNum+1; G=G+3; } NR=NR+0.15; G=G+48; } printf("%f\n",NR); }

61

int el,mp; int intergrid[200]; el=0; j=196; for (x=1;x<=49;x++) { for (lp=1;lp<=8;lp++) { intergrid[el]=j; j=j+3; el=el+1; } j=j+48; el=el+1; x=x+3; } for (mp=0;mp<130;mp++) { printf("%d\n",intergrid[mp]); } // Property definition fprintf(stream2,"%s%10d%8d%8.5f%8d\n","PSHELL",1,101,ti,101); // Material properties fprintf(stream2,"%s%12d%8.1E%16.2f%8.2f\n","MAT1",101,7E10,0.33,2823.0); // Boundary Conditions fprintf(stream2,"%s%12d%8d%8d%s%8d\n","SPC1",20,123456,1," THRU",25); fprintf(stream2,"%s%12d%8d%8d%s%8d\n","SPC1",20,456,25," THRU",1202); char s3[]="ENDDATA"; fprintf(stream2,"%s\n",s3); fclose(stream2); }

62

Appendix B:

This appendix gives the MATLAB file to read the NASTRAN output file and to calculate error.
%This code reads a .pch file and stores the decibel values at ten %different locations and calcualtes error in the sound %levels relative to actual value. clc clear all %opens air_shell_nastran.pch in read format fid=fopen('noisefile1.pch','r'); i=1; for i=1:90 tline=fgets(fid); end aa=sscanf(tline,'%*60c %12c %*8c'); tline=fgets(fid); tline=fgets(fid); tline=fgets(fid); tline=fgets(fid); bb=sscanf(tline,'%*60c %12c %*8c'); tline=fgets(fid); tline=fgets(fid); tline=fgets(fid); tline=fgets(fid); cc=sscanf(tline,'%*60c %12c %*8c'); tline=fgets(fid); tline=fgets(fid); tline=fgets(fid); tline=fgets(fid); dd=sscanf(tline,'%*60c %12c %*8c'); tline=fgets(fid); tline=fgets(fid); tline=fgets(fid); tline=fgets(fid); ee=sscanf(tline,'%*60c %12c %*8c'); tline=fgets(fid); tline=fgets(fid); tline=fgets(fid); tline=fgets(fid); ff=sscanf(tline,'%*60c %12c %*8c'); tline=fgets(fid); tline=fgets(fid); tline=fgets(fid); tline=fgets(fid); gg=sscanf(tline,'%*60c %12c %*8c'); tline=fgets(fid); tline=fgets(fid); tline=fgets(fid); tline=fgets(fid); hh=sscanf(tline,'%*60c %12c %*8c'); a=str2num(aa) b=str2num(bb)

63

c=str2num(cc) d=str2num(dd) e=str2num(ee) f=str2num(ff) g=str2num(gg) h=str2num(hh) S1=[a b c d e f g h] fclose(fid); no_exp=116; % % This line calculates the error value. Errorsquare=sqrt((a-103)^2+(b-106)^2+(c-109)^2+(d-110)^2+(e-110)^2+(f110)^2+(g-114)^2+(h-116)^2); fid1=fopen('ErrorSquare.dat','w'); fprintf(fid1,'%10.6f\n',Errorsquare); fclose(fid1);

64

Appendix C:

This appendix gives the input file for the NASTRAN Optimization Run.
NASTRAN REAL = 63000000 NASTRAN SYSTEM(151)=1 INIT MASTER(NORAM) INIT DBALL LOGICAL=(DBALL(500000)) INIT SCRATCH(NOMEM) LOGICAL=(SCRATCH(1800000)),SCR300=(SCR300(1800000)) ASSIGN OUTPUT2='final_optimization.op2',UNIT=12 ID NoiseModelling SOL 200 DIAG 8,12 ECHOOFF CEND $$$$$$$$$$$$$$$$$$$ CASE CONTROL BEGINS HERE ECHO=NONE set 44 = 5951 TITLE=NOISE MODELLING ANALYSIS SPC=10 $DESGLB = 5 DESOBJ(MIN) = 33 subcase 1 ANALYSIS = MODES METHOD(STRUCTURE)=10 DESSUB = 55 subcase 2 ANALYSIS = MFREQ METHOD(STRUCTURE)=20 METHOD(FLUID)=20 DLOAD=70 FREQUENCY=15 DISPLACEMENT(SORT1,PRINT,PUNCH)=44 FORCE(SORT1,PRINT,PUNCH)=44 DESSUB=117 $ DESOBJ(MIN) = 100 $$$$$$$$$$$$$$$$$$$ BULK DATA BEGINS HERE BEGIN BULK ECHO=NONE EIGRL,10,,,50 EIGRL,20,,,50 $ Define the design variables DESVAR,1,ringwid,0.015,0.005,0.03 DESVAR,2,ringthic,0.01,0.005,0.03 DESVAR,3,shelthic,0.00635,0.002,0.02 DESVAR,4,longwid,0.015,0.005,0.03 DESVAR,5,longthic,0.01,0.005,0.03 $ Relate the Design Varibles to change in stiffener and shell thicknesses DVPREL1,1,PSHELL,1,4,0.002,,,,+ +,3,1.0 DVPREL1,2,PSHELL,4,4,0.002,,,,+ +,3,1.0 DVPREL1,3,PSHELL,3,4,0.002,,,,+ +,3,1.0 DVPREL1,4,PBARL,1,12,0.005,,,,+

65

+,1,1.0 DVPREL1,5,PBARL,2,12,0.005,,,,+ +,4,1.0 DVPREL1,6,PBARL,1,13,0.005,,,,+ +,2,1.0 DVPREL1,7,PBARL,2,13,0.005,,,,+ +,5,1.0 $Define Objective DRESP1,1,DRUCK,FRDISP,,,1,77.8586,5951 DRESP1,2,DRU,FRDISP,,,7,77.8586,5951 DRESP2,100,BETA,100, ,DRESP1,1,2 $1234567$1234567$1234567$1234567$1234567$1234567$1234567$1234567$123456 7 DEQATN 100 OBJ(R,I) = 20.0 * LOG10(SQRT((R * * 2) + (I * * 2)) /(2.0E-5) ) DCONSTR,117,100,,70.0 $ Define Constraint on weight DRESP1,33,WEIGHT,WEIGHT, $DCONSTR,5,22,100.0,2500.0 $ Define Constraint on frequency DRESP1,3,frequ,FREQ,,,1, DCONSTR,55,3,23.0 $ Override default optimization parameters DOPTPRM,DESMAX,20,p1,1,p2,15,CONV1,1E-6 ,IPRINT,3 $ Grid points of the Torpedo GRID 1 -0.30 0.00 0.00 GRID 2 -0.3000 0.0869 0.0233 GRID 3 -0.3000 0.0779 0.0450 GRID 4 -0.3000 0.0636 0.0636 GRID 5 -0.3000 0.0450 0.0779 . . . . . . . . . . . . . . . GRID 1213 1.6000 0.0000 0.0000 GRID 1214 1.7500 0.0000 0.0000 GRID 1215 1.9000 0.0000 0.0000 $ $Grid Points of the surrounding water $ GRID* 1501 2.9100000000 0.0000000000 * 0.0000000000 -1 GRID* 1502 1.9996380000 0.9959744000 * 0.0000000000 -1 GRID* 1503 1.9996380000 0.9959744000 * 0.0000000000 -1 . . . . . . . . . . . . . . . GRID* 24964 -0.2567208000 0.3616001000 * -0.1620268000 -1

66

GRID* 24965 0.7309163000 * -0.2201162000 $ $ Grid points of the Air $ GRID 25001 GRID 25002 GRID 25003 . . . . . . GRID 31663 GRID 31664

1.7692490000 -1

0.05000 0.11314 0.11314-1 0.05000-0.11314 0.11314-1 0.05000 0.09868 0.12595-1 . . . . . . . . . 0.66061-0.10952-0.10952-1 0.68030-0.10952-0.10952-1

$ This Completes Grid Data $Now Elements for the Torpedo CTRIA3 1 1 2 3 1 CTRIA3 2 1 3 4 1 . . . . . . . . . . . . . . . . . . CTRIA3 23 1 24 25 1 CTRIA3 24 1 25 2 1 CQUAD4 25 1 2 3 27 26 CQUAD4 26 1 3 4 28 27 . . . . . . . . . . . . . . . . . . . . . CQUAD4 1199 3 1176 1177 1201 1200 CQUAD4 1200 3 1177 1154 1178 1201 CTRIA3 1201 3 1178 1179 1202 CTRIA3 1202 3 1179 1180 1202 . . . . . . . . . . . . . . . . . . CTRIA3 1223 3 1200 1201 1202 CTRIA3 1224 3 1201 1178 1202 CBAR 1225 1 194 195 1203 0.00000-0.00790-0.00212 0.00000-0.007080.00409 CBAR 1226 1 195 196 1203 0.00000-0.00708-0.00409 0.00000-0.005780.00578 . . . . . . . . . . . . . . . . . . . . . . . . CBAR 1839 2 1081 1105 1209 -0.00000-0.00817-0.00000-0.00000-0.008170.00000 CBAR 1840 2 1105 1129 1209 -0.00000-0.00817-0.00000-0.00000-0.008170.00000 CONM2 1841 196 1.9321 CONM2 1842 199 1.9321 . . . . . . . . . .

. . .

67

. CONM2 CONM2 RBE2 1110 1118

. 1943 1944 2000 1111 1119

. 1078 1081 1202 1112 1120 1128 1 176 184 192

1 1113 1121 1129 1 177 185 193

. 1.9321 1.9321 1106 1114 1122

1107 1115 1123

1108 1116 1124

1109 1117 1125

1126 RBE2 174 182 183 190 191 186 187 188 189 1127 3000 175 170 178 171 179 172 180 173 181

$Now Elements for water CTETRA 5001 400 CTETRA 5002 400 . . . . . . . . . CTETRA 127228 400 CTETRA 127229 400 $ Elements for Air CHEXA 130001700 25457 27057 27025 CHEXA 130002700 25489 27441 27057 . . . . . . . . . CHEXA 25392 CHEXA 25223 135576700 25225 25226 135577700

1728 1728 . . . 20508 20508

4392 3106 . . . 22572 22572

3106 1729 . . . 21832 24965

4391 4391 . . . 21161 21832

25001

25003

25053

25052

25425

25003

25004

25065

25053

25457

. . . 31280

. . . 31664

. . . 26320

. . . 26352

. . . 25380

. . .

31664

26256

25872

26320

25392

25211 25225 $Infinite Boundary Condition CAABSF 150001 4000 1501 CELAS1 140001 5000 1501 1 CAABSF 150002 4000 1502 CELAS1 140002 5000 1502 1 . . . . . . . . . . . . . . . CAABSF 154372 4000 9637 CELAS1 144372 5000 9637 1 CAABSF 154373 4000 9638 CELAS1 144373 5000 9638 1 $End of infinite boundary condition data

68

PAABSF,4000,,,,,2.2572e12 PELAS,5000,9.7169e-5 PSHELL 1 101 0.00635 101 PSHELL 4 102 0.00635 102 PSHELL 3 103 0.00635 103 PBARL,1,300,,BAR ,1.5e-2,1.0e-2 PBARL,2,300,,BAR ,1.5e-2,1.0e-2 MAT1 1017.0E+010 0.33 2780.00 MAT1 1027.0E+010 0.33 2780.00 MAT1 1037.0E+010 0.33 2780.00 MAT1 3007.0E+010 0.33 2780.00 SPC1 10 123 1 SPC1 10 1 2 THRU 25 SPC1 10 2 175 187 SPC1 10 3 181 193 MAT10,800,2.2+9,1026.0 PSOLID,400,800,,,,,PFLUID $Acoustic Source Dafinition for source one ACSRCE,70,500,,,2000,1.21,142355.3 TABLED1,2000,,,,,,,,+T1 +T1,0.0,0.0,77.76,0.0,77.86,1.0,77.96,0.0,+T2 +T2,1000.0,0.0,ENDT DAREA,500,25151,1,0.904778,29150,1,0.140919 FREQ,15,77.858604 $Material properties for the fluid MAT10,900,,1.21,343.0 PARAM,GRDPNT,0 PSOLID,700,900,,,,,PFLUID PARAM,POST,-2 PARAM,PREFDB,2.-5 PARAM RMS,YES ACMODL,diff,,,,0.04 PARAM,AUTOSPC,NO Param,Prgpst,No ENDDATA

69

References
1. Allen, M. J., Vlahopoulos, N., Integration of Finite Element and Boundary Element Methods for Calculating the Radiated Sound from a Randomly Excited Structure, Computers and Structures, Vol. 77, No. 2, pp. 155-169, 2000. 2. Everstine, G. C., Henderson, F. M., Coupled Finite Element/Boundary Element Approach for Fluid-Structure Interaction, Journal of Acoustical Society of America, Vol. 87, 1990. 3. Zienkiewicz, O. C., Newton, R. E., Coupled Vibrations of a Structure Submerged in a Compressible Fluid, Proceedings of International Symposium on Finite Element Techniques, Stuttgart, pp. 359-379, 1969. 4. Everstine, G. C., Schroeder, E. A., Marcus, M. S., The Dynamic Analysis of Submerged Structures, Nastran Users Experience, NASA TM X-3278, Washington, 1975. 5. Pinsky, P. M., Abboud, N. N., Transient Finite Element Analysis of the Exterior Structural Acoustics Problem, In: Numerical Techniques in Acoustic Radiation, American Society of Mechanical Engineers, New York, pp. 35-47, 1989. 6. Marcus, M. S., A Finite Element Method Applied to the Vibration of Submerged Plates, Journal of Ship Research, Vol. 22, pp. 94-99, 1978. 7. Everstine, G. C., Finite Element Formulations of Structural Acoustic Problems, Computers and Structures, Vol. 65, pp. 307-321, 1997. 8. Ruzzene, M., Baz, A. Finite Element Modeling of Vibration and Sound Radiation from Fluid-Loaded Shells, Journal of Thin-Walled Structures, Vol. 36, pp. 21-46, 2002.

70

9. Everstine, G. C., A Symmetric Potential Formulation for Fluid-Structure Interaction, Journal of Sound and Vibration, Vol. 79, pp. 157-160, 1981. 10. Everstine, G. C., Finite Element Solution of Transient Fluid-Structure Interaction Problems, 19th NASTRAN Users Colloquium, NASA CP-3111, Washington, DC, pp. 162-173, 1991. 11. Everstine, G. C., Structural Analogies for a Scalar Field Problem, International Journal of Numerical Methods in Engineering, Vol. 17, pp. 419-429, 1981. 12. Hunt, J. T., Knittel, M. R., Barach, D., Finite Element Approach to Acoustic Radiation from Elastic Structures, Journal of Acoustical Society of America ,Vol. 55, pp. 269-280, 1974. 13. Hunt, J. T., Knittel, M. R., Nichols, C. S., Barach, D., Finite Element Approach to Acoustic Scattering from Elastic Structures, Journal of Acoustical Society of America, Vol. 57, pp. 287-299, 1975. 14. Everstine, G. C., Henderson, F. M., Lipman, R. R., Finite Element Prediction of Acoustic Scattering and Radiation from Submerged Elastic Structures, Proceeding of 12th NASTRAN Users Colloquium, pp. 194-209, 1984. 15. Zienkiwicz, O. C., Taylor, R. L., The Finite Element Method, McGraw-Hill, Vol. 2, 1989. 16. Choi, K. K., Dong, J., Design Sensitivity Analysis and Optimization of High Frequency Radiation Problems Using Energy Finite Method and Energy Boundary Element Method, 10th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference, AIAA 2004-4615, 30th August-1st September, Albany, New York, 2004.

71

17. Zhang, W., Wang, A., Vlahopoulos, N., An Alternative Energy Finite Element Formulation Based on Incoherent Orthogonal Waves and its Validation for Marine Structures, Finite Elements in Analysis and Design, Vol. 38, pp. 10951113, 2002. 18. Geers, T. L., Felippa, C. A., Doubly Asymptotic Approximations for Vibration Analysis of Submerged Structures, Journal of Acoustical Society of America, Vol. 57, Issue. 4, pp. 1152-1159, 1975. 19. Everstine, G. C., Taylor, D. W., A NASTRAN Implementation of the Doubly Asymptotic Approximation for Underwater Shock Response, NASTRAN Users Experiences, pp. 207-228, 1976. 20. Rosen, M. W., Acoustic Studies on Power Transmission in Underwater Missile Propulsion, 1st ed., Arlington: Compass Publications, pp. 357379, 1967. 21. Kinsler, L. E., Frey, A. R., Coppens, A. B., Sanders, J. V., Fundamentals of Acoustics, Wiley, NY, 1982. 22. Dudley, D. W., Introduction to Acoustical Engineering of Gear Transmission, in Underwater Missile Propulsion, 1st ed., Arlington: Compass Publications, pp. 349357, 1967. 23. NASTRAN Users Manual, Mac-Neal Schwendler Corporation, Los Angeles, CA, 2001. 24. Pierce, A. D., Acoustics: An Introduction to its Principles and Applications, 1st ed., Acoustical Society of America, 1989. 25. DOT Users Manual, Vanderplaats Research & Development, Inc., Colorado Springs, CO, 2001.

72

26. Junger, M. C., Feit, D., Sound, Structures, and Their Interaction, 2nd ed. MIT Press, 1986. 27. The NASTRAN Theoretical Manual, Mac-Neal Schwendler Corporation, Los Angeles, CA, 2001. 28. The NASTRAN Advanced Dynamics Guide, Mac-Neal Schwendler Corporation, Los Angeles, CA, 2001.

73