Sensitivity analysis of a multi-parameter performance function and application to a nuclear rocket engine system.
| dc.contributor.author | Logan, Robert J. | |
| dc.contributor.corporate | Aerospace Systems Laboratory | |
| dc.contributor.department | Aerospace and Mechanical Sciences | |
| dc.date.accessioned | 2012-11-01T22:55:38Z | |
| dc.date.available | 2012-11-01T22:55:38Z | |
| dc.date.issued | 1970 | |
| dc.description.abstract | The use of computers to optimize free parameters of a system has become relatively widespread in many areas of engineering. Parameter optimization codes have been written for that purpose, and make it possible for a design engineer, once he has developed the mathematics of a system, to optimize its parameters according to some criteria. But of equal interest to the design engineer is the sensitivity of the optimized criteria to departure of the parameters away from their optimal value. The purpose of this thesis is to show ways in which a parameter optimization code may be augmented to yield such sensitivity information. A Fletcher and Powell version of Davidon's variable metric optimization search technique was employed to optimize multi-parameter functions. Their method is useful in that it computes the inverse Hessian matrix (or matrix of second derivatives), which completely describes the curvature of the function at the optimum. Equations were developed so that the sensitivity could be expressed in a meaningful output format. This was made possible through the use of matrix inversion and eigenvalue analysis subroutines which were obtained from the scientific subroutine library of the IBM 360 91 and used in conjunction with a digital computer code employing the Fletcher and Powell technique. Equality constrained optimization problems were also considered by employing the penalty factor method proposed by Courant and used by Kelley, Equations analogous to the use of Lagrangian Multipliers were used to determine the cost of the equality constraint. Example problems are offered showing the optimal solutions, sensitivity data at the optimum, and interpretation of that data. The well known Rosenbrock function was used to exhibit the accuracy of the methods employed. A typical engineering problem was solved involving the sensitivity of an optimal nuclear rocket engine used to inject a payload onto an interplanetary trajectory. The results indicated that the thermal power of the reactor and the ratio of length to diameter of the core could be varied considerably from their optimal values with little cost. The power density however was relatively fixed for optimal operation. | |
| dc.description.uri | http://archive.org/details/sensitivitynalys1094515158 | |
| dc.identifier.oclc | o640322333 | |
| dc.identifier.uri | https://hdl.handle.net/10945/15158 | |
| dc.language.iso | en_US | |
| dc.publisher | Princeton University | en_US |
| dc.subject.lcsh | Aeronautics | en_US |
| dc.title | Sensitivity analysis of a multi-parameter performance function and application to a nuclear rocket engine system. | en_US |
| dc.type | Thesis | en_US |
| dspace.entity.type | Publication | |
| etd.thesisdegree.discipline | Engineering | en_US |
| etd.thesisdegree.grantor | Princeton University | en_US |
| etd.thesisdegree.level | Masters | en_US |
| etd.thesisdegree.name | Master of Science in Engineering | en_US |