Download files
Abstract
Over the years, unmanned underwater vehicles (UUVs) have proven to be an efficient, robust and versatile research tool for ocean exploration, military and industrial applications. In most cases, UUVs provide an added benefit of conducting various maritime operations for a greater time period and with less risk relative to the alternative of manned submersibles and professional divers. Therefore, it comes as no surprise that the development of more advanced, hydrodynamically efficient, and sea-worthy UUVs for ever changing maritime environment is an actively pursued research area.
The quest for understanding what lies below the water bodies has motivated the development of UUVs with a wide variety of shapes, sizes, working depth limits, sources of energy and means of propulsion. Hundreds of different UUVs have been designed over the past decades to meet the challenges of oceanographic exploration and exploitation programs. Previous attempts on UUV designs have focused primarily on functional designs with a view of fulfilling a certain set of requirements while very little research has been directed to identify optimum designs. Most of the developed UUVs are of a similar shape and draw heavily on past experience where non-optimal designs are largely accepted as the final design. Therefore, in this thesis, an optimization framework for the preliminary design of UUVs is presented. This research also addresses the internal clash-free arrangement of the on-board components for optimal position of the centre of gravity, reduced separation between centre of gravity and centre of buoyancy, etc., that eventually contribute to the stability of the designed vehicle.
The proposed framework incorporates geometry and configuration modules, a hydrodynamics module, several accepted maritime performance and characteristics estimation methods of UUVs, and a suite of state-of-the-art optimization algorithms. In the present work, an efficient integrated analysis/design system is developed through the seamless integration of a numerical analysis software (Matlab) and a computer-aided design software (CATIA). The developed framework provides an automated multidisciplinary design environment that allows various forms of single or multi-objective optimization formulations of the UUV design problems to be carried out utilizing commercial and in-house codes without user intervention. The modularity and catalogue driven structure adopted in the framework allows for the design of UUVs for various design requirements and/or with different propulsion and power system options.
The framework is used for design optimization of UUVs during concept and preliminary design phases. Several case studies consisting of single and multi-objective hull form resistance minimization problems for multiple classes (e.g. medium-, small- and micro-size UUV) are discussed. The performance of the UUV designed by using the developed framework is found better when compared with an existing one. While optimization methods can be applied to identify designs which are faster, more manoeuvrable and flexible to deal with various mission profiles, a practical realization requires the design to be robust, i.e. one with the best average performance under expected parametric variations. Therefore, to deliver practical designs, the formulations have been extended to yield robust optimal designs.
Prior research investigations in minimizing UUV hull form resistance mostly have used low-fidelity model (drag estimated using empirical formulas) and very limited studies in this regard could be found on the use of high-fidelity analysis methods (drag estimated using computational fluid dynamics, CFD analysis) within the course of optimization. The lack of a unified framework comprising different fidelity models for UUV design optimization has also revealed in the literature. Thus, the developed framework is extended to deal with high-fidelity estimates derived through seamless integration of a computer-aided design tool (CATIA), meshing and CFD analysis tools (ICEM and FLUENT). While it is understood that the high-fidelity analysis is more accurate, they also tend to be far more computationally expensive. Therefore, useful insights on possible means to identify appropriateness of fidelity models via correlation measures are proposed.
The inclusion of computationally intensive higher fidelity models has led to the development of surrogates or approximations that are commonly used instead of the expensive analysis in order to contain the computational cost within affordable limits. While there are a plethora of surrogate model types including Kriging and Co-Kriging, their performances are not always satisfactory for the highly non-linear problems where construction of a surrogate using solely the design variables is non-trivial. Therefore two interesting new methods to deal with the multi-fidelity data are proposed in this thesis, i.e., Additive Kriging and Modified Kriging. In the former approach, the function is approximated using a linear combination of the cheap function and a difference function constructed based on a Kriging model using cheap and expensive estimates. In the later approach, a low-fidelity estimate is provided as an input in addition to the design variables within a Kriging model. The performances of the proposed approaches are demonstrated using two examples: a well studied single variable mathematical problem and a real-world multi-variable submarine design optimization problem. For the first example, the performance of the Modified Kriging is found competitive with the existing Co-Kriging based approaches, whereas poor performance of the Additive Kriging is found in predicting the deceptive expensive function. On the other hand, both the models have performed well on the second example and resulted in nearly 35% savings in computational time.