Much Research on composite solid propellants has been performed over the past few decades and much progress has been made, yet many of the fundamental processes are still unknown, and the development of new propellants remains highly empirical. Ways to enhance the performance of solid propellants for rocket and other applications continue to be explored experimentally, including the effects of various additives and the impact of fuel and oxidizer particle sizes on burning behavior. One established method to measure the burning rate of composite propellant mixtures in a controlled laboratory setting is to use a constant-volume pressure vessel, or strand burner. To provide high-pressure burn rate data at pressures up to 360 atm, the authors have installed, characterized and improved a strand burner facility at the University of Central Florida. Details on the facility and its improvements, the measurement procedures, and the data reduction and interpretation are presented. Two common HTPB/AP propellant mixtures were tested in the original strand burner. The resulting burn rates were compared to data from the literature with good agreement, thus validating the facility and related test techniques, the data acquisition, data reduction and interpretation. After more than 380 successful recordings, an upgraded version of the strand burner, was added to the facility. The details of Strand Burner II, its improvements over Strand Burner I, and its characterization study are presented.

The object of this investigation was to develop a general solution for the thermal stresses and strains in a hollow cylindrical case-bonded solid propellant. The heat conduction equation, as solved by Carslaw and Jaeger, was applied to a hollow composite cylinder. The temperature distribution from this equation was used in conjunction with the stress and strain for an elastic solid propellant. The boundary conditions were employed to solve for the constants and the general solution for the stresses and strains were obtained. In order to study the predictions of the general expressions, a numerical example was presented. It was found that the maximum stress and strain appeared at the inner radius of the grain. It was also observed that the stress and strain increased with an increase in the radius ratio "m”. Failure criteria for the grain under consideration were discussed. A method for obtaining the maximum allowable temperature variation (from cure temperature) was investigated. Knowing the stress and strain characteristics of the grain the equations developed would indicate failure conditions and also allow calculations of the maximum allowable temperature variations prior to grain failure. / M.S.

M.S.

Star grain configuration design has been widely used in solid rocket applications for several decades. Although a large number of surface cracks are detected in the rocket motor propellants, the mechanism of these cracks is sull not well known due to the complex geometry of the grain. A stress-freezing photoelastic investigation has been performed to study the deep cracks which emanate from the fingertips of the star-shaped cutout cylinders. Using three-dimensional photoelasticity and proper algorithms in fracture mechanics, the stress intensity factors (SIF's) and the stress singularity orders along the crack front have been calculated. A surface effect on the dominant singularity order is observed and some analytical results are employed as a comparison. Meanwhile, three-dimensional finite element solution to the circular cylinder is used to find the “equivalent” inner radius for the internal star cylinder and the variation of SIF's along the crack border shows a very similar trend to the experimental results once the "equivalent" radius is adopted. / Master of Science

Estimation of the total mission life of a spacecraft is an important issue for the communication satellite industries. For accurate determination of the remaining mission life of the satellite it is essential to estimate the amount of propellant present in the propellant tank of the spacecraft at various stages of its mission life. Because the annual revenue incurred from a typical commu-nication satellite operating at its full capacity is on the order of millions of dollars, premature removal of spacecraft from their orbits results in heavy losses. Various techniques such as the bookkeeping method, the gas law method, numerical modeling techniques, and use of capacitive sensors have been employed in the past for accurate determination of the amount of propellant present in a spacecraft. First half of the thesis is concerned with sensitivity analysis of the various propellant gauging techniques, that is, estimating the effects of the uncertainty in the instruments employed in the propellant gauging system on the onboard propellant estimation. This sensitivity analysis is done for three existing propellant gauging techniques – gas injection method, book-keeping method and the propellant tank heating method. A comparative study of the precision with which the onboard propellant is estimated by the three techniques is done and the primary source of uncertainty for all the three techniques is identified. It is illustrated that all the three methods — the gas injection method, the book-keeping method and the propellant tank heating method — are inherently indirect methods of propellant gauging, as a consequence of which, the precision with which the three techniques estimate the residual propellant decreases towards the end of mission life of the spacecraft. The second half of the thesis explores the possibility of using a new propellant tank configuration, consisting of a truncated cone centrally mounted within a spherical propellant tank, to measure the amount of liquid propellant present within the tank. The liquid propellant present within the propellant tank orients itself in a geometry, by virtue of its dominant surface tension force in zero-g condition, which minimizes its total surface energy. Study reveals that the amount of liquid propellant present in the tank can thus be estimated by measuring the height of the propellant meniscus within the central cone. It is also observed that, unlike gas law method, bookkeeping method or the propellant tank heating method, where the precision of the estimated propellant fill-fraction decreases towards the end-of-life of the spacecraft, for the proposed new configuration the precision increases.

It is well known that surface tension of a liquid has a decisive role in flow dynamics and the eventual equilibrium state, especially in confined flows under low gravity conditions and also in free surface flows. One such instance of a combination of these two cases where surface tension plays an important role is in the microgravity environment of a spacecraft propellant tank. In this specific case both propellant acquisition and residual propellant estimation are critical to the mission objectives particularly in the end-of-life phase. While there have been a few studies pertaining to the equilibrium state in given geometric configurations, the transient flow leading to final state from an initial arbitrary distribution of propellant is rarely described. The present study is aimed at analysing the dynamic behaviour of the liquids under reduced gravity through numerical simulation and also addresses the specific case of propellant flow transient in a cone-in-a-sphere type of tank configuration proposed by Lal and Raghunandan which is likely to result in both improved acquisition and life time estimation of spacecraft. While addressing this specific problem, the present work aims to study the transient nature of such surface tension driven flows in a general form as applicable to other similar problems also. Volume of Fluid (VOF) method for multiphase model in ANSYS FLUENT was adapted with suitable changes for generating numerical solutions to this problem. Simulations were run for three different cone angles of 17o, 21o & 28o with a flat liquid surface for full scale models to measure the rise height and time of rise. Two scaled models of ½ and 1/10th of the original dimensions with the same liquid configuration of the 28o cone angle case were simulated to see if the time scales involved would come down for experimental feasibility. A third simulation of the 1/10th scale model was run with the liquid spread in the tank to imitate the general conditions found in the propellant tank in microgravity. To understand the behaviour of liquids in the microgravity state to changing physical parameters, a set of simulations was run using liquid phases as water and hydrazine with different physical parameters of temperature and surface tension. The theory put forward by Lal and Raghunandan was found to stand firm. In the case of the cone angle of 28o it was observed that in the final equilibrium state the liquid collected towards the apex of the cone with the larger volume fraction of liquid accumulating inside the cone. An addition of a cylindrical section at the bottom of the cone seems to help although not uniformly for all case. The equilibrium settling times for all the three cone angle cases were in the order of 300 to 600 seconds for simulations on a spherical tank of diameter two metres which was close to the actual tank dimension used on spacecraft. Scaled down simulations of 1/10th and ½ the tank geometry with both flat liquid surfaces and spread out liquid volumes showed that the smaller models had equilibrium settling times which were considerably lower (in the order of tens of seconds) than the full scale models. Although smaller, these time scales are larger than the maximum time scales available in drop tower tests which provide a maximum free fall time of around 9 to 10 seconds. Validation of the proposed configuration by flying an aircraft in a parabolic flight path is a possibility that could be explored for the scaled down models since the zero-g duration for these flights is on an average between 15-20 seconds.

This thesis contains investigations in the area of polymers herein propellants binders are modified functionally to meet the requirements of future energetic propellants. Chapter 1 contains a broad introduction to the area of recent advances in solid propellants and the numerous applications of ‘Click Chemistry’. Chapters 2 details the materials, characterization tools and the experimental techniques employed for the studies. This is followed by Chapter 3, 4, and 5 which deals with functional modification of various propellants binders, their characterisation and evaluation in propellant formulations. Chapter 6 details with the thermal decomposition of diazides and its reaction with alkenes. The advent of modern rockets has opened a new era in the history of space exploration as well as defence applications. The driving force of the rocket emanates from the propellant – either solid or liquid. Composite solid propellants find an indispensable place, in today’s rockets and launch vehicles because of the inherent advantages such as high reliability, easy manufacturing, high thrust etc. The composite propellant consisting of inorganic oxidiser like ammonium perchlorate, (AP), ammonium nitrate (AN) etc), metallic fuel (aluminium powder, boron etc) and polymeric fuel binder (hydroxyl terminated polybutadiene-HTPB, polybutadiene-acrylic acid-acrylonitrile PBAN, glycidyl azide polymer (GAP), polyteramethylene oxide (PTMO) etc. is used in igniters, boosters, upper stage motors and special purpose motors in large launch vehicles. Large composite solid propellant grains or rocket motors in particular, demand adequate mechanical properties to enable them to withstand the stresses imposed during operation, handling, transportation and motor firing. They should also have a reasonably long ‘potlife’ to provide sufficient window for processing operations such as mixing and casting which makes the selection of binder with appropriate cure chemistry more challenging. In all composite solid propellants currently in use, polymers perform the role of a binder for the oxidiser, metallic fuel and other additives. It performs the dual role of imparting dimensional stability to the composite, provides structural integrity and good mechanical properties to the propellant besides acting as a fuel to impart the required energetics. Conventionally, the terminal hydroxyl groups in the binders like GAP, PTMO and HTPB are reacted with diisocyanates to form a polyurethane network, to impart the necessary mechanical properties to the propellant. A wide range of diisocyantes such as tolylene diisocyanate (TDI) and isophorone diisocyanate (IPDI) are used for curing of these binders. However, the incompatability of isocyanates with energetic oxidisers like ammonium dinitramide (ADN), hydrazinium nitroformate (HNF), short ‘potlife’ of the propellant slurry and undesirable side reactions with moisture are limiting factors which adversely affect the mechanical properties of curing binders through this route. The objective of the present study is to evolve an alternate approach of curing these binders is to make use of the 1,3 dipolar addition reactions between azide and alkyne groups which is a part of ‘Click chemistry’. This can be accomplished by the reaction of azide groups of GAP with triple bonds of alkynes and reactions of functionally modified HTPB/PTMO (azide/alkyne) to yield 1,2,3 -triazole based products. This offers an alternate route for processing of solid propellants wherein, the cured resins that have improved mechanical properties, better thermal stability and improved ballistic properties in view of the higher heat of decomposition resulting from the decomposition of the triazole groups. GAP is an azide containing energetic polymer. The azide groups can undergo reaction with alkynes to yield triazoles. In, Chapter 3 the synthesis and characterisation of various alkynyl compounds including bis propargyl succinate (BPS), bis propargyl adipate (BPA), bis propargyl sebacate (BPSc.) and bis propargyl oxy bisphenol A (BPB) for curing of GAP to yield triazoles networks are studied. The mechanism of the curing reaction of GAP with these alkynyl compounds was elucidated using a model compound viz. 2-azidoethoxyethane (AEE). The reaction mechanism has been analysed using Density Functional Theory (DFT) method. DFT based theoretical calculations implied marginal preference for 1, 5 addition over the 1, 4 addition for the uncatalysed cycloaddition reaction between azide and alkyne group. The detailed characterisation of these systems with respect to the cure kinetics, mechanical properties, dynamic mechanical behaviour and thermal decomposition characteristics were done and correlated to the structure of the network. The glass transition temperature (Tg), tensile strength and modulus of the system increased with crosslink density which in turn is, controlled by the azide to alkyne molar stoichiometry. Thermogravimetic analysis (TGA) showed better thermal stability for the GAP-triazole compared to GAP based urethanes. Though there have been a few reports on curing of GAP with alkynes, it is for the first time that a detailed characterisation of this system with respect to the cure kinetics, mechanical, dynamic mechanical, thermal decomposition mechanism of the polymer is being reported. To extent the concept of curing binders through 1,3 dipolar addition reaction, the binder HTPB as chemically transformed to propargyloxy carbonyl amine terminated polybutadiene (PrTPB) with azidoethoxy carbonyl amine terminated polybutadiene (AzTPB) and propargyloxy polybutadiene (PTPB). Similarly, PTMO was convnerted to propargyloxy polytetramethylene oxide (PTMP). Triazole-triazoline networks were derived by the reaction of the binders with alkyne/azide containing curing agents. The cure characteristics of these polymers (PrTPB with AzTPB, PTPB with GAP and PTMP with GAP) were studied by DSC. The detailed characterisations of the cured polymers for were done with respect to the, mechanical, dynamic mechanical behaviour and thermal decomposition characteristics were done. Propellant level studies were done using the triazoles derived from GAP, PrTPB-AzTPB, PTPB and PTMP as binder, in combination with ammonium perchlorate as oxidiser. The propellants were characterised with respect to rheological, mechanical, safety, as well as ballistic properties. From the studies, propellant formulations with improved energetics, safety characteristics, processability and mechanical properties as well defect free propellants could be developed using novel triazole crosslinked based binders. Chapter 6, is aimed at understanding the mechanism of thermal decomposition of diazido compounds in the first section. For this, synthesis and characterisation of a diazido ester 1,6 –bis (azidoacetoyloxy) hexane (HDBAA) was done. There have been no reports on the thermal decomposition mechanism of diazido compounds, where one azide group may influence the decomposition of the other. The thermal decomposition mechanism of the diazido ester were theoretically predicted by DFT method and corroborated by pyrolysis-GC-MS studies. In the second section of this chapter, the cure reaction of the diazido ester with the double bonds of HTPB has been investigated. The chapter 6B reports the mechanism of Cu (I) catalysed azide-alkene reaction validated using density functional theory (DFT) calculations in isomers of hexene (cis-3-hexene, trans-3-hexene and 2-methy pentene: model compound of HTPB) using HDBAA. This the first report on an isocyanate free curing of HTPB using an azide. Chapter 7 of the thesis summarizes the work carried out, the highlights and important findings of this work. The scope for future work such as development of high performance eco-friendly propellants based on triazoles in conjunction with chlorine-free oxidizer like ADN, synthesis of compatible plasticisers and suitable crosslinkers have been described. This work has given rise to one patent, three international publications and four papers in international conferences in the domain.

In this thesis, the thermal decomposition investigation of haloethanols namely 2-chloroethanol and 2-bromoethanol are reported both experimental and theoretical. Computational calculation of enthalpy of formation haloethanols using isodesmic and atomization reactions has also been reported. Finally, the chemistry of JP-10 ignition has also been investigated using shock tube. Chapter 1 gives a brief introduction to the experimental shock tube technique. Brief surveys of literature pertinent to haloethanols and JP-10 have also been discussed. The importance of thermal rate coefficient and detection techniques in shock tube chemistry is presented. Details of the theoretical methods used in the determination of thermal rate coefficients have been described at the end of the chapter. In Chapter 2, I have discussed experimental methods used in carrying out this work. The details of the experimental shock tube set-up employed in this work have been elaborated in this chapter. Kinetic simulations performed to understand the mechanism of chemical transformation of haloethanols at high temperature have also been presented. In chapter 3, thermal decomposition results obtained for 2-chloroethanol have been described. The kinetic data have been obtained in the temperature range of 930-1100 K behind the reflected shock wave in a shock tube. Analyses of pre and post shock mixture using FT-IR and gas chromatographic techniques are presented. Chemical kinetic simulation performed to simulate the product distribution is presented. The reduced kinetic model has also been presented which was obtained using the sensitivity analysis and was validated by comparison to the shock tube measurements. The details of the β-substitution effect have been shown. The kinetic parameters of the unimolecular elimination of HCl and H2O have been presented both experimentally and theoretically. Theoretical results were obtained by transition state theory using quantum chemistry calculations HF, MP2 (FULL) and B3LYP/6-311++G** level of theory. The details of intrinsic reaction coordinate calculation and potential energy surface calculations have also been described. These experimental and theoretical results suggest that the rate of HCl elimination is faster than that of H2O and HOCl elimination reaction. In chapter 4, I have reported thermal decomposition results obtained for 2-bromoethanol. The kinetic data have been obtained in the temperature range of 910-1102 K behind the reflected shock wave in a shock tube. Analyses of pre and post shock mixture using FT-IR and gas chromatographic techniques are discussed. Chemical kinetic simulation performed to simulate the product distribution is presented. The details of the β-substitution effect are explained. Both experimental and theoretical kinetic parameters of the unimolecular elimination of HBr and H2O have been presented. Theoretical results were obtained by transition state theory using quantum chemistry calculations at the HF, MP2 (FULL) and B3LYP/6-311++G** level of theory. The intrinsic reaction coordinate calculation and potential energy surface have been investigated in details. From this experimental and theoretical studies, it has been concluded that the rate of HBr elimination much faster than that of H2O. However, the experiments show that the rate of HOBr elimination is faster than that of the H2O. In chapter 5, I have reported the computational calculation of enthalpy of formation of haloethanols. The enthalpy of formation of haloethanols of the general formula XC2H4OH were calculated by the HF, MP2, B3LYP, G2, G3, G2MP2, G3B3, G3MP2B3, CBS-Q, CBS-QB3 and CCSD/cc-pVDZ level of theories applying isodesmic and atomization reactions. Results obtained using the Benson’s group and bond additivity methods have also been described at 298.15 K and at 1 atm in the gaseous state. In chapter 6, ignition delay measurement on neat jet propellent-10 (JP-10) and JP-10 + Triethyl amine (TEA) mixture have been reported. The JP-10 (Exo-tetrahydrodicyclopentadiene, C10H16) ignition delay times were measured behind a single pulse reflected shock wave in a shock tube. Experiments were performed over high temperature, high pressure, and three equivalence ratio and for different composition. It has been shown that the TEA can reduce the ignition delay of JP-10. A higher level quantum chemistry calculation has also been presented that were performed to obtain the bond dissociation energies of C-H bonds in JP-10. Chapter7 is the concluding chapter where the main work done in this thesis is summarized and future direction is presented.

<p>This work presents the development and/or application of several laser diagnostics for studying the flame structure of composite propellant flames. These studies include examining the flame structure of novel energetic materials with potential as propellant ingredients, the near-surface flame structure of basic composite propellants, and the global flame structure of propellants containing metal additives.<br></p><p><br></p><p>First, the characterization of the deflagration of various novel energetic cocrystals is presented. The synthesis and development of novel energetic materials is a costly and challenging process. Rather than synthesizing new materials, cocrystallization provides the potential opportunity to achieve improved properties of existing energetic materials. This work presents the characterization of the effect of cocrystallization on the deflagration of a 2:1 molar cocrystal of CL-20 and HMX as well as a 1:1 molar cocrystal of CL-20 and TNT. A hydrogen peroxide (HP) solvate of CL-20 as well as a polycrystalline composite of HMX and ammonium perchlorate (AP) were also studied. A physical mixture of each material was also tested for comparison. The burning rate of each material was measured as a function of pressure. Flame structure during self-deflagration was examined using planar laser-induced fluorescence (PLIF) of CN and OH. The burning rate of the HMX/CL-20 cocrystal and the CL-20/HP solvate closely matched that of CL-20, but the burning rate of the TNT/CL-20 cocrystal was between the burning rate of its coformers. All HMX/AP materials had a higher burning rate than either HMX or AP individually and the burning rate of a physical mixture was found to be a function of particle size. The differences in the burning rate of the physical mixtures and composite crystal of HMX/AP can be explained by changes in the flame structure observed using PLIF. Burning rates and flame structure of the cocrystals were found to closely match those of their respective physical mixtures when smaller particle sizes were used (approx. less than 100 um). The results obtained demonstrate that the deflagration behavior of the coformers is not indicative of the deflagration behavior of the resulting physical mixture or cocrystal. However, changes in the resulting flame structure greatly affect the burning rate.</p><p><br></p><p>Next, PLIF of nitric oxide (NO) was utilized to characterize the near surface flame structure of composite propellants of AP and hydroxyl-terminated polybutadiene (HTPB) containing varying particle sizes of AP burning at 1 atm in air. In all propellants, the NO PLIF signal was strongest close to the burning propellant surface and fell to a non-zero constant value within ~1 mm of the surface where it remained throughout the remainder of the flame. Distinct diffusion-flame-like structure was observed above large individual burning AP particles in the propellant containing a bimodal distribution of 400 and 40 um AP. In contrast, the flame of a propellant containing only fine AP (40 um) behaved like a homogeneous, premixed flame. The flame of the propellant containing a bimodal distribution of 200 and 40 um AP also showed similar behavior to a premixed flame with some heterogeneous structure indicating that, at this pressure, the propellant is approaching a limit where the particle sizing is small enough that the flame behaves like a homogeneous, premixed flame. Additionally, propellants containing aluminum were tested. No significant differences were observed in the NO PLIF behavior between the propellants with and without aluminum suggesting that, at these conditions, the aluminum does not have a significant effect on the AP/HTPB flame structure near the burning surface.</p><p><br></p><p>The effect of aluminum particle size on the temperature of aluminized-composite-propellant flames burning at 1 atm is also presented. In this work, measurements of 1) the temperature of CO (within the flame bath gas) and 2) the temperature of AlO (located primarily within regions surrounding the burning aluminum particles) within aluminized, AP-HTPB-propellant flames were performed as a function of height above the burning propellant surface. Three aluminized propellants with varying aluminum particle size (nominally 31 um, 4.5 um, or 80 nm) and one non-aluminized AP-HTPB propellant were studied while burning in air at 1 atm. A wavelength-modulation-spectroscopy (WMS) diagnostic was utilized to measure temperature and mole fraction of CO via mid-infrared wavelengths and a conventional AlO emission-spectroscopy technique was utilized to measure the temperature of AlO. The bath-gas temperature varied significantly between propellants, particularly within 2 cm of the burning surface. The propellant with the smallest particles (nano-scale aluminum) had the highest average temperatures and far less variation with measurement location. At all measurement locations, the average bath-gas temperature increased as the initial particle size of aluminum in the propellant decreased, likely due to increased aluminum combustion. The results support arguments that larger aluminum particles can act as a heat sink near the propellant surface and require more time and space to ignite and burn completely. On a time-averaged basis, the temperatures measured from AlO and CO agreed within uncertainty at near 2650 K in the nano-aluminum propellant flame, however, AlO temperatures often exceeded CO temperatures by ~250 to 800 K in the micron-aluminum propellant flames. This result suggests that in the flames studied here, and on a time-averaged basis, the micron-aluminum particles burn in the diffusion-controlled combustion regime, whereas the nano-aluminum particles burn within or very close to the kinetically controlled combustion regime.</p><p><br></p><p>The study of the effect of aluminum particle size on the temperature of aluminized, composite-propellant flames was then extended to characterize the same propellants burning at elevated pressures ranging from 1 to 10 atm. A novel mid-infrared scanned-wavelength direct absorption technique was developed to acquire measurements of temperature and CO in particle-laden propellant flames burning at up to 10 atm. The results from the application of this diagnostic are among the very first measurements of gas properties in aluminized composite propellant flames burning at pressures above atmospheric pressure. In all propellants, the flame temperature and combustion efficiency of the propellant flames increased with an increase in pressure. In addition, the propellants with smaller aluminum particle sizes achieved higher flame temperatures as the particles were able to ignite and react faster. However, the propellants containing nano-scale and the smallest micron-scale aluminum powders had similar global flame temperatures suggesting that at some point a decrease in particle size results in minimal gains in the overall flame temperature. The results demonstrate how well measurements of gas properties can be used to understand the behavior of the aluminum particle combustion in the flame.</p><p><br></p><p>Last, the design, development, and application of a laser-absorption-spectroscopy diagnostic capable of providing quantitative, time-resolved measurements of gas temperature and HCl concentration in flames of aluminized, composite propellant flames is presented. This diagnostic utilizes a quantum-well distributed-feedback tunable diode laser emitting near 3.27 um to measure the absorbance spectra of one or two adjacent HCl lines using a scanned-WMS technique which is insensitive to non-absorbing transmission losses caused by metal particulates in the flame. This diagnostic was applied to characterize the spatial and temporal evolution of temperature and/or HCl mole fraction in small-scale flames of AP-HTPB composite propellants containing either an aluminum-lithium alloy or micron-scale aluminum. Experiments were conducted at 1 and 10 atm. At both pressures, the flame temperature of the aluminum-lithium propellant, on a time-averaged basis, was 80 to 200 K higher than that of the aluminum-propellant (depending on location in the flame) indicating more complete combustion. In addition, the mole fraction of HCl in the aluminum-lithium propellant flame reached values 65-70% lower than the conventional aluminum-propellant flame at the highest measurement location in the flame. The measurements at both pressures showed similar trends in the reduction of HCl in the aluminum-lithium propellant flame but at 10 atm this occurred on a length scale an order of magnitude smaller than the flame at atmospheric pressure. The results presented further support that the use of an aluminum-lithium alloy is effective at reducing HCl produced by the propellant flame without compromising performance, thereby making it an attractive additive for solid rocket propellants.</p>

Thesis (MScIng)--University of Stellenbosch, 2003. / ENGLISH ABSTRACT: The emission spectrum measured in the middle infrared (IR) band from the plume of a rocket can be used to identify rockets and track inbound missiles. It is useful to test the stealth properties of the IR fingerprint of a rocket during its design phase without needing to spend excessive amounts of money on field trials. The modelled predictions of the IR spectra from selected rocket motor design parameters therefore bear significant benefits in reducing the development costs. In a recent doctorate study it was found that a fundamental approach including quantum-mechanical and computational fluid dynamics (CFD) models was not feasible. This is first of all due to the complexity of the systems and secondly due to the inadequate calculation speeds of even the most sophisticated modern computers. A solution was subsequently investigated by use of the ‘black-box’ model of a multi-layer perceptron feed-forward neural network with a single hidden layer consisting of 146 nodes. The input layer of the neural network consists of 18 rocket motor design parameters and the output layer consists of 146 IR absorbance variables in the range from 2 to 5 μm wavelengths. The results appeared promising for future investigations. The available data consist of only 18 different types of rocket motors due to the high costs of generating the data. The 18 rocket motor types fall into two different design classes, the double base (DB) and composite (C) propellant types. The sparseness of the data is a constraint in building adequate models of such a multivariate nature. The IR irradiance spectra data set consists of numerous repeat measurements made per rocket motor type. The repeat measurements form the pure error component of the data, which adds stability to training and provides lack-of-fit ANOVA capabilities. The emphasis in this dissertation is on comparing the feed-forward neural network model to the linear and neural network partial least squares (PLS) modelling techniques. The objective is to find a possibly more intuitive and more accurate model that effectively generalises the input-output relationships of the data. PLS models are known to be robust due to the exclusion of redundant information from projections made to primary latent variables, similarly to principal components (PCA) regression. The neural network PLS techniques include feed-forward sigmoidal neural network PLS (NNPLS) and radial-basis functions PLS (RBFPLS). The NNPLS and RBFPLS algorithms make use of neural networks to find non-linear functional relationships for the inner PLS models of the NIPALS algorithm. Error-based neural network PLS (EBNNPLS) and radial-basis function network PLS (EBRBFPLS) are also briefly investigated, as these techniques make use of non-linear projections to latent variables. A modification to the orthogonal least squares (OLS) training algorithm of radial-basis functions is developed and applied. The adaptive spread OLS algorithm (ASOLS) allows for the iterative adaptation of the Gaussian spread parameters found in the radial-basis transfer functions. Over-fitting from over-parameterisation is controlled by making use of leaveone- out cross-validation and the calculation of pseudo-degrees of freedom. After cross-validation the overall model is built by training on the entire data set. This is done by making use of the optimum parameterisation obtained from cross-validation. Cross-validation also gives an indication of how well a model can predict data unseen during training. The reverse problem of modelling the rocket propellant chemical compositions and the rocket physical design parameters from the IR irradiance spectra is also investigated. This problem bears familiarity to the field of spectral multivariate calibration. The applications in this field readily make use of PLS and neural network modelling. The reverse problem is investigated with the same modelling techniques applied to the forward modelling problem. The forward modelling results (IR spectrum predictions) show that the feedforward neural network complexity can be reduced to two hidden nodes in a single hidden layer. The NNPLS model with eleven latent dimensions outperforms all the other models with a maximum average R2-value of 0.75 across all output variables for unseen data from cross-validation. The explained variance for the output data of the overall model is 94.34%. The corresponding explained variance of the input data is 99.8%. The RBFPLS models built using the ASOLS training algorithm for the training of the radialbasis function inner models outperforms those using K-means and OLS training algorithms. The lack-of-fit ANOVA tests show that there is reason to doubt the adequacy of the NNPLS model. The modelling results however show promise for future development on larger, more representative data sets. The reverse modelling results show that the feed-forward neural network model, NNPLS and RBFPLS models produce similar results superior to the linear PLS model. The RBFPLS model with ASOLS inner model training and 5 latent dimensions stands out slightly as the best model. It is found that it is feasible to separately find the optimum model complexity (number of latent dimensions) for each output variable. The average R2-value across all output variables for unseen data is 0.43. The average R2-value for the overall model is 0.68. There are output variables with R2-values of over 0.8. The forward and reverse modelling results further show that dimensional reduction in the case of PLS does produce the best models. It is found that the input-output relationships are not highly non-linear. The non-linearities are largely responsible for the compensation of both the DB- and C-class rocket motor designs predictions within the overall model predictions. For this reason it is suggested that future models can be developed by making use of a simpler, more linear model for each rocket class after a class identification step. This approach however requires additional data that must be acquired. / AFRIKAANSE OPSOMMING: Die emissiespektra van die uitlaatpluime van vuurpyle in die middel-infrarooi (IR) band kan gebruik word om die vuurpyle te herken en om inkomende vuurpyle op te spoor. Dit is nuttig om die uitstralingseienskappe van ‘n vuurpyl se IR afdruk te toets, sonder om groot bedrae geld op veldtoetse te spandeer. Die gemodelleerde IR spektrale voorspellings vir ‘n bepaalde stel vuurpylmotor ontwerpsparameters kan dus grootliks bydra om motorontwikkelingskostes te bemoei. In ‘n onlangse doktorale studie is gevind dat ‘n fundamentele benadering van kwantum-meganiese en vloeidinamika-modelle nie lewensvatbaar is nie. Dit is hoofsaaklik as gevolg van die onvoldoende vermoë van selfs die mees gesofistikeerde moderne rekenaars. ‘n Moontlike oplossing tot die probleem is ondersoek deur gebruik te maak van ‘n multilaag perseptron voorwaartse neurale netwerk met 146 nodes in ‘n enkele versteekte laag. Die laag van invoer veranderlikes bestaan uit agtien vuurpylmotor ontwerpsparameters en die uitvoerlaag bestaan uit 146 IR-absorbansie veranderlikes in die reeks golflengtes vanaf 2 tot 5 μm. Dit het voorgekom dat die resultate belowend lyk vir toekomstige ondersoeke. Weens die hoë kostes om die data te genereer bestaan die beskikbare data uit slegs agtien verskillende tipes vuurpylmotors. Die agtien vuurpyl tipes val verder binne twee ontwerpsklasse, naamlik die dubbelbasis (DB) en saamgestelde (C) dryfmiddeltipes. Die yl data bemoeilik die bou van doeltreffende multiveranderlike modelle. Die datastel van IR uitstralingspektra bestaan uit herhaalde metings per vuurpyltipe. Die herhaalde metings vorm die suiwer fout komponent van die data. Dit verskaf stabilitieit tot die opleiding op die data en verder die vermoë om ‘n analise van variansie (ANOVA) op die data uit te voer. In hierdie tesis lê die klem op die vergelyking tussen die voorwaartse neurale netwerk en die lineêre en neurale netwerk parsiële kleinste kwadrate (PLS) modelleringstegnieke. Die doel is om ‘n moontlik meer insiggewende en akkurate model te vind wat effektief die in- en uitvoer verhoudings kan veralgemeen. Dit is bekend dat PLS modelle meer robuus kan wees weens die weglating van oortollige inligting deur projeksies op hoof latente veranderlikes. Dit is analoog aan hoofkomponente (PCA) regressie. Die neurale netwerk PLS-tegnieke sluit in voorwaartse sigmoïdale neurale netwerk PLS (NNPLS) en radiale-basis funksies PLS (RBFPLS). Die NNPLS en RBFPLS algoritmes maak gebruik van die neurale netwerke om nie-lineêre funksionele verbande te kry vir die binne PLS-modelle van die nie-lineêre iteratiewe parsiële kleinste kwadrate (NIPALS) algoritme. Die fout-gebaseerde neurale netwerk PLS (EBNNPLS) en radiale-basis funksies PLS (EBRBFPLS) is ook weens hulle nie-lineêre projeksies na latente veranderlikes kortiliks ondersoek. ‘n Aanpassing tot die ortogonale kleinste kwadrate (OLS) opleidingsalgoritme vir radiale-basis funksies is ontwikkel en toegepas. Die aangepaste algoritme (ASOLS) behels die iteratiewe aanpassing van die verspreidingsparameters binne die Gauss-funksies van die radiale-basis transformasie funksies. Die oormatige parameterisering van ‘n model word beheer deur kruisvalidering met enkele weglatings en die berekening van pseudo-vryheidsgrade. Na kruisvalidering word die algehele model gebou deur opleiding op die volledige datastel. Dit word gedoen deur van die optimale parameterisering gebruik te maak wat deur kruisvalidering bepaal is. Kruisvalidering gee ook ‘n goeie aanduiding van hoe goed ‘n model ongesiende data kan voorspel. Die modellering van die vuurpyle se chemiese en fisiese ontwerpsparameters (omgekeerde probleem) is ook ondersoek. Hierdie probleem is verwant aan die veld van spektrale multiveranderlike kalibrasie. Die toepassings in die veld maak gebruik van PLS en neurale netwerk modelle. Die omgekeerde probleem word dus ondersoek met dieselfde modelleringstegnieke wat gebruik is vir die voorwaartse probleem. Die voorwaartse modelleringsresultate (IR voorspellings) toon dat die kompleksiteit van die voorwaartse neurale netwerk tot twee versteekte nodes in ‘n enkele versteekte laag gereduseer kan word. Die NNPLS model met elf latente dimensies vaar die beste van alle modelle, met ‘n maksimum R2-waarde van 0.75 oor alle uitvoer veranderlikes vir die ongesiende data (kruisvalidering). Die verklaarde variansie vir die uitvoer data vanaf die algehele model is 94.34%. Die verklaarde variansie van die ooreenstemmende invoer data is 99.8%. Die RBFPLS modelle wat gebou is deur van die ASOLS algoritme gebruik te maak om die PLS binne modelle op te lei, vaar beter in vergelyking met die K-gemiddeldes en OLS opleidingsalgoritmes. Die toetse wat ‘n ‘tekort-aan-passing’ ANOVA behels, toon dat daar rede is om die geskiktheid van die NNPLS model te wantrou. Die modelleringsresultate lyk egter belowend vir die toekomstige ontwikkeling van modelle op groter, meer verteenwoordigde datastelle. Die omgekeerde modellering toon dat die voorwaartse neurale netwerk, NNPLS en RBFPLS modelle soortgelyke resultate produseer wat die lineêre PLS model s’n oortref. Die RBFPLS model met ASOLS opleiding van die PLS binne modelle word beskou as die beste model. Dit is lewensvatbaar om die optimale modelkompleksiteite van elke uitvoerveranderlike individueel te bepaal. Die gemiddelde R2-waarde oor alle uitvoerveranderlikes vir ongesiende data is 0.43. Die gemiddelde R2-waarde vir die algehele model is 0.68. Daar is van die uitvoer veranderlikes wat R2-waardes van 0.8 oortref. Die voor- en terugwaartse modelleringsresultate toon verder dat dimensionele reduksie in die geval van PLS die beste modelle lewer. Daar is ook gevind dat die nie-lineêriteite grootliks vergoed vir die voorspellings van beide DB- en Ctipe vuurpylmotors binne die algehele model. Om die rede word voorgestel dat toekomstige modelle ontwikkel kan word deur gebruik te maak van eenvoudiger, meer lineêre modelle vir elke vuurpylklas nadat ‘n klasidentifikasiestap uitgevoer is. Die benadering benodig egter addisionele praktiese data wat verkry moet word.