This thesis presents a way to maximise the photographic coverage of Phobos, one of the two Martian moons, as part of the space mission Mars Moon eXplorer. This coverage is performed by the French hyperspectral imager MacrOmega, and two criteria are selected: the area covered and the resolution of the pictures. The approach considered is a greedy algorithm, and elements of basic theory are provided. This greedy approach is compared to a chronological algorithm, whose results were already approved for the mission. / Detta examensarbete presenterar ett sätt att maximera den fotografisk täckningen av Phobos, som är en av Mars två månar, som en del av rymduppdraget Mars Moon eXplorer. Den fotografiska täckningen ska utföras av den franska hyperspektralavbildaren MacrOmega, och två kriterier har valts ut: (i) området som omfattas samt (ii) bildens upplösning. Metoden som testas är en girig algoritm och baselementen i algoritmen presenteras. Den giriga algoritmens resultat jämförs med resultat från en kronologisk algoritm, vars resultat redan godkänts för uppdraget.

This work has as an objective to structure the content of a course on Attitude determination methods, part of an Aerospace Engineering Master program. A selection of books, papers, theses, web sites and films was reviewed to identify the most relevant topics within the areas of Static and Dynamic Attitude Determination and the ways to present them in a educational context. Theory is presented in a simplified way and examples were gathered to illustrate the theoretical part.  Finally, a discussion is carried out on the main learning goals and challenges, required time for instruction and exercises and suggestion for a grading system.​ / Detta arbete har som mål att strukturera innehållet i en kurs om Attitydbestämningsmetoder inom flyg- och rymdteknikmastersprogram. Ett urval av böcker, artiklar, avhandlingar, webbsidor och filmer granskades för att identifiera de mest relevanta ämnena inom statisk och dynamisk attitydbestämning och de olika sätten att presentera dem i ett utbildningssammanhang. Teorin presenteras på ett förenklat sätt och några exemplar visas för att illustrera den teoretiska delen. Avslutningsvis, diskuteras de huvudsakliga lärandemålen, nödvändig handledning och övningstid, samt betygsättning.

This report presents the bending modes study conducted on a heavy launcher. The controller of the launcher takes as inputs the attitude and attitude rate measurements given by the Inertial Measurement Unit (IMU). Since the bending modes generate measurement errors at the IMU location, the study of deformations due to these bending modes is critical to assess the stability of the launcher during the atmospheric flight phase. The goal of this master thesis project is to detect and then select the most excitable bending modes among the large number of modes provided by a detailed structural analysis of the launcher. Only these relevant modes will be later used to generate reduced dynamical models of the launcher in order to efficiently design an appropriate controller. Indeed, considering all the bending modes will dramatically increase the calculation time and will not significantly improve the representativeness of the model at the control law frequency range of interest. To reach this objective, an extended excitability (the maximum of the module of the transfer function between the effective deflection and the considered mode generalized coordinate transported at the IMU location) is defined and computed for each mode. A criterion has been implemented to choose only the relevant modes. The sensitivity study conducted during this master thesis project has shown that with around 20 modes over 200, one can reproduce the dynamic behavior of the complete system. / Denna rapport presenterar studien för böjningslägen som utförts på en bärraket. Eftersom böjningslägen genererar mätfel vid IMU-platsen är studien av deformationer på grund av dessa böjningslägen avgörande för att bedöma stabiliteten under den atmosfäriska flygfasen. Målet med detta examensarbete är att upptäcka och sedan välja de mest spännande böjningslägena bland det stora antalet lägen som tillhandahålls av en detaljerad strukturanalys av bärraketen. Endast dessa relevanta lägen kommer senare att användas för att generera reducerade dynamiska modeller av bärraketen för att effektivt utforma en lämplig styrenhet. Faktum är att övervägande av alla böjningslägen dramatiskt ökar beräkningstiden och kommer inte att förbättra modellens representativitet väsentligt. För att uppnå detta mål definieras och beräknas en utvidgad excitabilitet (det maximala för överföringsfunktionens modul mellan effektiv avböjning och den övervägda modaliserade koordinaten som transporteras på IMU-platsen) för varje läge. Ett kriterium har implementerats för att bara välja de relevanta lägena. Känslighetsstudien som genomfördes under detta examensarbete har visat att med cirka 20 lägen över 200 kan man återge det dynamiska beteendet hos hela systemet.

Optimal orbital trajectories are obtained through the solution of highly nonlinear large scale problems. In the case of low-thrust propulsion applications, the spacecraft benefits from high specific impulses and, hence, greater payload mass. However, these missions require a high count of orbital revolutions and, therefore, display augmented sensitivity to many disturbances. Solutions to such problems can be tackled via a discrete approach, using optimal feedback control laws. Historically, differential dynamic programming (DDP) has shown outstanding results in tackling these problems. A state of the art software that implements a variation of DDP has been developed by Whiffen and it is used by NASA’s DAWN mission [Mystic: Implementation of the Static Dynamic Optimal Control Algorithm for High-Fidelity, Low-Thrust Trajectory Design" , AAS/AIAA Astrodynamics Specialist Conference, (Keystone, Colorado), American Institute of Aeronautics and Astronautics, Aug. 21, 2006]. One of the latest techniques implemented to deal with these discrete constrained optimizations is the Hybrid Differential Dynamic Programming (HDDP) algorithm, introduced by Lantoine and Russell in [A Hybrid Differential Dynamic Programming Algorithm for Constrained Optimal Control Problems. Part 1: Theory", Journal of Optimization Theory and Applications, vol. 154, pp. 382-417, issue 2, Aug. 1, 2012]. This method complements the reliability and efficiency of classic nonlinear programming techniques with the robustness to poor initial guesses and the reduced computational effort of DDP. The key feature of the algorithm is the exploitation of a second order state transition matrix procedure to propagate the needed partials, decoupling the dynamics from the optimization. In doing so, it renders the integration of dynamical equations suitable for parallelization. Together with the possibility to treat constrained problems, this represents the greatest improvement of classic DDP. Nevertheless, the major limitation of this approach is the high computational cost to evaluate the required state transition matrices. Analytical derivatives, when available, have shown a significant reduction in the computational cost and time for HDDP application. This work applies differential algebra to HDDP to cope with this limitation. In particular, differential algebra is introduced to obtain state transition matrices as polynomial maps. These maps come directly from the integration of the dynamics of the system, removing the dedicated algorithmic step and reducing its computational cost. Moreover, by operating on polynomial maps, all the solutions of local optimization problems are treated through differential algebraic techniques. This approach allows us to deal with higher order expansions of the cost, without modifying the algorithm. The leading assumption of this work is that, treating higher than second order expansions, grants larger radii of convergence for the algorithm, improved robustness to initial guesses, hence faster rates of convergence. Examples are presented in this thesis to assess the performance of the newly constructed algorithm and to test the assumptions.​ / Optimala omloppsbanor erhålls genom lösningen av mycket storskaliga olinjära problem. I fallet med låg framdrivningskraft så drar farkosten nytta av hög specifik impuls och därmed större slutlig farkostmassa. Dock så kräver dessa rymduppdrag flera omloppsvarv och uppvisar därför ökad känslighet för olika störningskrafter. Lösningar på dessa problem kan hanteras via ett diskret tillvägagångssätt med hjälp av optimal reglering. Historiskt har differentialdynamisk programmering (DDP) visat enastående resultat för att hantera dessa problem. En toppmodern programvara som implementerar en variation av DDP har utvecklats av Whiffen i ["Mystic: Implementation of the Static Dynamic Optimal Control Algorithm for High-Fidelity, Low-Thrust Trajectory Design" , AAS/AIAA Astrodynamics Specialist Conference, (Keystone, Colorado), American Institute of Aeronautics and Astronautics, Aug. 21, 2006] och används av NASA:s rymduppdrag Dawn. En av de senaste teknikerna som implementerats för att hantera dessa diskreta och begränsade optimeringar är en hybrid differentialdynamisk programmeringsalgoritm (HDDP) som introducerades av Lantoine och Russell i ["A Hybrid Differential Dynamic Programming Algorithm for Constrained Optimal Control Problems. Part 1: Theory", Journal of Optimization Theory and Applications, vol. 154, pp. 382-417, issue 2, Aug. 1, 2012]. Denna metod kompletterar pålitligheten och effektiviteten hos klassiska olinjära programmeringstekniker med robusthet mot dåliga initiala gissningar och den reducerade beräkningskostnaden för DDP. Nyckelegenskapen hos algoritmen är utnyttjandet av en procedur för andra ordningens övergångsmatris för propagering av de erforderliga partiella derivatorna. Denna procedur frikopplar också dynamiken från optimeringen. Genom att göra så blir integration av de dynamiska ekvationerna lämpliga för parallellisering. Tillsammans med förmågan att ta itu med begränsade problem representerar detta den största förbättringen av klassisk DDP. Ändå är den stora begränsningen av detta tillvägagångssätt den höga kostnaden för beräkningar som krävs för att utvärdera tillståndsövergångsmatriserna. När de är tillgängliga, har analytiska derivatorer visat en signifikant minskning av beräkningskostnaden och tiden för HDDP-tillämpningar. Detta arbete tillämpar differentialalgebra på HDDP för att klara av denna begränsning. I synnerhet införs differentialalgebra för att erhålla tillståndsövergångsmatriser som polynomavbildningar. Dessa avbildningar kommer direkt från integrationen av systemets dynamik och därför är det möjligt att ta bort det dedikerade algoritmiska steget och minska beräkningskostnaden. Vidare behandlas alla lösningar av lokala optimeringsproblem genom olika algebraiska tekniker genom att använda polynomkartor. Detta tillvägagångssätt tillåter oss att hantera högre ordningens expansionstermer av kostnadsfunktionen utan att ändra algoritmen. Det främsta antagandet i detta arbete är att behandling av högre än andra ordningens expansionstermer ger större konvergensradier för algoritmen, förbättrad robusthet mot sämre initiala gissningar och följaktligen snabbare konvergensnivåer. Exempel presenteras i denna examensarbete för att bedöma prestandan hos den nybyggda algoritmen och för att testa antagandena.

Designing a flight control system for an aircraft is a multifaceted task which has to take many requirements and constrints into account. Being able to quickly evaluate the feasibility of achieving design targets in terms of – for example – maneuvering capability and stability mar-gins is valuable when faced with the task of designing a flight control system for a new aircraft or aircraft configuration. This thesis presents methods to quickly assess the maneuverability capability of an aircraft with certain dynamics and control servo constraints by inverting the aircraft dynamics. Limitations related to the achievable robustness properties are also de-scribed with an emphasis on unstable aircraft configurations. The influence from variations of key flight mechanical parameters such as Cmα is investigated and the results are exempli-fied for a linearized model of the Admire aircraft.

The aerospace industry relies on nondestructive evaluation (NDE) to ensure aircraft safety and will benefit from methods that allow for early damage detection. Photoluminescence piezospectroscopy (PS) has demonstrated stress and damage sensing of substrates when coupled with alpha-alumina nanoparticles in a polymer matrix applied as a sensor coating. Alpha phase alumina exhibits photoluminescent spectral emission lines (R-lines) that shift due to changes in the stress state of the alumina. The coatings' capability for sensing early subsurface damage suggests the potential for implementing stress sensing paint for integrity monitoring of aircraft structures. To achieve a viable stress sensing coating that can be applied as a paint, materials for optimal sensing and processing need to be tailored for aircraft applications. In addition, advances in optics technology for area measurement and faster data collection are needed. In this work, manufacturing of the sensing paint was achieved by introducing alumina nanoparticles into an aircraft grade topcoat using 3 different processing approaches and the paint with the best dispersion was identified using quantitative luminescence intensity results. To maintain the ease of application through spraying, dispersant was added to the paint. Tensile tests on composite and aluminum substrates resulted in spectral shifts with applied loading that reveal non-uniform and non-recoverable stresses within the paint. Scanning electron microscopy showed microcracks verifying that the sensing paint experienced damage during loading. R1 peaks shift as the paint was heated and cooled, indicating the possibility that the paint is sensitive to temperature changes. Future iterations of the sensing paint will focus on improvements in polymer mechanical properties and homogeneity on application, particle-to-polymer bonding and enhanced adhesion. Area measurement was achieved through the development and calibration of a hyperspectral imaging system using a laser with wider aperture. The long-term goal is to establish a standardized paint-based PS coating and optics technology for structural integrity monitoring of aircraft structures.

Boron carbide (B4C) is one of the most important opaque boride ceramics that has high hardness and Young's modulus that along with low density lead to a significant resistance to ballistic impact and, thus, B4C is broadly used as a protective material. B4C has also high neutron capturing cross section; therefore, it is used as control rods and neutron absorption shielding in nuclear reactors. In this work thermal, electrical and mechanical properties of dense B4C ceramics (99%) sintered using Spark Plasma Sintering (SPS) were investigated. The Young's modulus of B4C measured by three different techniques – IE, RUS, and nanoindentation showed a very good overlap in values, which ranges from 419.2 ± 47.3 GPa for nanoindentation to 458.7 GPa for RUS measurements at room temperature. The mean contact pressure-contact depth plots obtained from load-displacement nanoindentation data indicated pop-in events during loading and an "elbow" event during unloading, both of which are indicative of possible structural changes in B4C structure during nanoindentation. The appearance of "elbow" deviations in load-displacement nanoindentation curves of B4C was detected for the first time. The 4-point bending strength of the B4C ceramics was equal to 585 ± 70 MPa with Weibull parameter of 9.9 and scale parameter equal to 611 MPa. The biaxial strength of B4C was measured to be much lower and equal to 238.6 ± 122 MPa with Weibull parameters of 2.2 and scale parameter equal to 271 MPa. To the best of our knowledge the biaxial strength of B4C was also measured for the first time. In this work it was determined that failure of B4C occurred by fully transgranular fracture, with no intergranular failure present on fracture surface. B4C's fracture toughness Klc = 3 ± 0.19 MPa x m1/2 was measured using SEVNB technique, which is similar to previously reported values.

Ordinary differential equations can describe many dynamic systems. When physics is well understood, the time-dependent responses are easily obtained numerically. The particular numerical method used for integration depends on the application. Unfortunately, when physics is not fully understood, the discrepancies between predictions and observed responses can be large and unacceptable. In this thesis, we show how to directly implement integration of ordinary differential equations through recurrent neural networks using Python. We leveraged modern machine learning frameworks, such as TensorFlow and Keras. Besides offering basic models capabilities (such as multilayer perceptrons and recurrent neural networks) and optimization methods, these frameworks offer powerful automatic differentiation. With that, our approach's main advantage is that one can implement hybrid models combining physics-informed and data-driven kernels, where data-driven kernels are used to reduce the gap between predictions and observations. In order to illustrate our approach, we used two case studies. The first one consisted of performing fatigue crack growth integration through Euler's forward method using a hybrid model combining a data-driven stress intensity range model with a physics-based crack length increment model. The second case study consisted of performing model parameter identification of a dynamic two-degree-of-freedom system through Runge-Kutta integration. Additionally, we performed a numerical experiment for fleet prognosis with hybrid models. The problem consists of predicting fatigue crack length for a fleet of aircraft. The hybrid models are trained using full input observations (far-field loads) and very limited output observations (crack length data for only a portion of the fleet). The results demonstrate that our proposed physics-informed recurrent neural network can model fatigue crack growth even when the observed distribution of crack length does not match the fleet distribution.

Presented within this work is a new method for inertial orbit estimation of an object, either known or unknown, adaptable to a network of low-cost observation satellites. The observation satellites would only require a monocular camera for line of sight measurements. Using the line of sight measurements of each observer, a pair of orthogonal geometric planes that intersect both the observation satellite and the target are created. The intersection of the two planes in the inertial frame defines the new measurement model that is implemented with multiple observation nodes. Total system observability is analyzed and the instantaneous (per node) observability is used to remove "bad" measurements from the system. The measurement model is used in an extended Kalman filter framework and the measurement noise nonlinear transformation is addressed. Three cases are presented; first, the minimum number of required observation nodes to produce accurate results if determined. Then, a smaller number of observation nodes is analyzed to highlight the use of the instantaneous observability and its deleterious effect on the filter performance. Finally, the method is expanded out to multiple observation satellites in a constellation. For all cases, the results show that this method is capable of producing accurate orbit estimation that converges in a short time.

Modern flight control laws are designed utilizing modeled plant aerodynamics, and as such the closed-loop system is sensitive to the actual aerodynamics and flight environment. Control laws such as dynamic inversion rely on an onboard aerodynamic model of the flight controller that is not always accurate because of simplifications in the modeling process or unmodeled dynamics. Typically, the most accurate estimation of the aerodynamics is determined with an expensive wind tunnel test (WTT) supplemented with aerodynamic finite element modeling. The WTT takes a considerable amount of the research and development budget, yet it may not provide an aerodynamic model suitable for flight control. This issue can be overcome by implementing a linear robust control law augmented with an online adaptive control law. The linear robust control law can be designed by any established methods, but in this work we present a new parameter space method that guarantees a desired gain and phase margin. The new method is developed to obtain a desired performance and stability in the presence of the aerodynamic uncertainty. Unlike the conventional s-domain parameter space method that utilizes the pole-placement technique analytically, the new method designs the controller in the frequency domain numerically using the stability margin specification. The linear robust control is enhanced by an adaptive control system that is designed by the online Feed-Forward Neural Network (FFNN). The FFNN adaptive control compensates for the aerodynamic uncertainty and imperfect modeling of aircraft dynamics, and it gradually replaces the linear controller as the network gains converge to a value that minimizes the linear control law. Although the FFNN adaptively adjusts the controller gains, an additional stability augmentation system is designed by Sigma-Pi Neural Network (SPNN) for compensating for the nonlinearity of the aircraft dynamics. The SPNN predicts the control input at a specific flight condition by memorizing the previous flight empirically. The SPNN adapts both the engine speed and elevator commands in the aircraft speed/altitude control. Training the SPNN is performed using a recursive least square estimator, and the control design is demonstrated on a six-degree-of-freedom (6DOF) digital simulation.