Development of a system design methodology for robust thermal control subsystems to support responsive space

Derek William Hengeveld, Purdue University

Abstract

This dissertation presents design and implementation tools supporting the development of Responsive Space thermal control subsystems. An extended literature review revealed topics requiring immediate attention and others providing promising contributions. Consequently, several diverse works were carried out. Single hot- and cold-case design orbits that work well in the design of Responsive Space thermal control subsystems over a wide range of satellite surface properties and expected operating environments were identified. While hot- and cold-case design orbits for traditional missions are established from a well-defined spacecraft attitude and orientation, robust thermal control subsystem bounding orbit conditions must be based on a broad range of potential orbit definitions. A computational tool was developed which improves satellite thermal performance through intelligent component placement. Based on orbital-averaged electronic component power and environmental heat fluxes, the tool provides the optimal satellite face and location within that face for all components such that thermal gradients across the satellite are minimized. Results are determined using a two-step process. First, a global distribution optimization strategy was applied to determine the specific satellite face with which a component should be mated. After all component-face combinations are determined, a local placement algorithm was applied to determine specific component locations within each face. Both steps were based on genetic-algorithms that utilize a combination of elitist strategies, reproduction, local gradient searches, mutation, and best- and worst-fit heuristics. Tuning studies were carried out on the genetic-algorithms to determine appropriate convergence criteria, population size, and values for evolution parameters. The global distribution optimization strategy determines component-face pairs for multi-sided satellites, which achieve a more balanced distribution of heat flux. This depends upon panel orientation and the external thermal environmental fluxes placed upon them. Results showed that a strictly rule-based approach would provide the least computational expense and provide reasonable results for the cases considered here. However, this approach might not be suitable for cases not considered. Consequently, an approach that takes full advantage of the capabilities of the algorithm is recommended. However, this approach increases computational expense. Computational requirements, on average, are expected to be approximately 12 seconds or less for up to 54 components using a 2.5 gigahertz dual-core computer. The maximum computational requirements are expected to be less than approximately 22 seconds for up to 54 components using the same machine. Optimized, even, and worst-case distributions for a nominal distribution of 36 components in the hot-case orbit were found for total power of 100 W up to 1200 W. The second step provides rapidly optimized component placement on a given panel approaching a uniform heat flux distribution. Optimized results were obtained for 18 uniform and 11 non-uniform components within 20 s and 7 s, respectively, using a 2.5 gigahertz dual-core processor. Advantages of this method include no need for thermophysical properties and boundary conditions. Optimized results are obtained using only component averaged power and domain size. Consequently, this approach is ideally suited to situations where limited information is readily available. In addition, limiting the required inputs provides for relatively fast solutions. However, care should be taken to ensure that a uniform distribution of fluxes is required for optimized placement. This robust and fast approach can be utilized in a variety of applications including microelectronics and satellite development and is especially suited to those demanding low computational expense. Reduced-order models to predict satellite temperature responses for an 11-factor computer simulation model were developed. These surrogate models were shown to provide acceptable results over more computationally expensive computer simulations. Consequently, these models are ideally suited for rapid evaluation of a wide-range of satellite thermal control subsystem design approaches. Further, these models do not require users to have an extensive background in thermal modeling and/or access to expensive thermal simulation software. (Abstract shortened by UMI.)

Degree

Ph.D.

Advisors

Groll, Purdue University.

Subject Area

Aerospace engineering|Mechanical engineering

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS