| The design of large-scale, complex systems such as plug-in hybrid electric vehicles (PHEVs) motivates the use of formal optimization methods from both multidisciplinary design optimization (MDO) and optimal control theory. Traditionally, MDO methods have been used to address the integrated design of engineering systems comprised of multiple, interacting components and/or disciplines for superior static system performance. Optimal control theory, on the other hand, is often used to select the best operation strategy of a given dynamic system for superior dynamic system performance. Although many times in practice the optimal design and control of such dynamic systems are addressed almost independently, this approach generally yields sub-optimal overall design solutions. This is because the system architecture, or physical design, is inherently coupled with its operation strategy, or control design. Combined optimal design and control techniques, also known as co-design, can address this issue by using an integrated approach to enable superior design solutions for dynamic systems. This thesis focuses on the co-design of large-scale systems, specifically PHEVs based on simultaneous multidisciplinary dynamic system design optimization (MDSDO) methods using direct transcription (DT). In order to enable a simultaneous approach for optimizing the design and control of the PHEV, a toolbox was developed to design all the critical component of a PHEV powertrain including: electric motor, generator, engine, transmission, and high voltage battery. This toolbox takes the size related design variables as inputs and by using the embedded analytical equations, generates the output performance characteristics of each component. The MDSDO problem formulation is then solved using GPOPS-II,a DT-based MATLAB software for solving multiple-phase optimal control problems. DT-based simultaneous problem formulations in MDSDO has already been successfully used in moderate scale problems, however there has been very few attempts to implement this method on large-scale problems. The current study addresses this issue and examines the practicality of DT-based simultaneous problem formulations in MDSDO for large-scale, complex dynamic systems. |
