Optimizing a New Upper Stage Rocket Engine for 5 Missions in 6 Weeks
Executive Summary
- Industry: Aerospace
- Products: ModelCenter Integrate , ModelCenter Explore
- Applications: MBE: Model Based Engineering , MDAO: Multi-Disciplinary Analysis and Optimization
The Air Force Institute of Technology (AFIT) is investigating the innovative Dual-Expander Aerospike Nozzle (DEAN) engine architecture in support of initiatives like the Integrated High Payoff Rocket Propulsion Technology (IHPRPT) and Next Generation Engine (NGE) to improve satellite launch operations. J. Simmons’ research project at the AFIT, the Air Force’s graduate school, was to build a simulation model to optimize the DEAN for use as a replacement for the RL10, the second stage engine on both the Atlas V and Delta IV space launch systems. Simmons began by automating the DEAN model, generating enough data for an American Institute of Aeronautics and Astronautics (AIAA) journal article in the process. Simmons and two AFIT masters students continued evolving the DEAN automation and over the course of their graduate studies, followed by exploration of the DEAN trade space including an optimization of the DEAN for a baseline NGE mission. With this optimization study complete, Simmons sent an email to his committee to update them on the status of the project six weeks before the end of his Ph.D. research. One of the replies asked him to expand the project by optimizing DEAN for two more NGE missions, the Air Force’s X-37 space plane and two configurations for NASA’s Space Launch System (SLS). Fortunately, Simmons built the simulation model using the ModelCenter framework to automate the workflow for optimizing the DEAN architecture. In only 41 days he optimized the DEAN for the five new missions, demonstrating that the new engine architecture has the potential to outperform traditional upper stage engines through a wide range of mission scenarios.
Background
DEAN’s innovative architecture
In a traditional rocket engine expander cycle, the fuel is pumped through a cooling jacket for the chamber and nozzle. The energy transferred to the fuel from cooling the chamber and nozzle is then used to drive the turbine turning both the fuel and oxidizer pumps before the fuel is introduced into the chamber. In the DEAN, the fuel and oxidizer each drive their own turbines to power their own pumps. The DEAN also uses an aerospike nozzle that runs through the middle of the rocket’s propellant flow and up into the chamber, leaving the ambient atmosphere to form the outer boundary for the flow. The interaction with the ambient atmosphere gives aerospike nozzles automatic altitude compensation, making them more efficient over a range of altitudes than similar bell nozzles which must be designed for a single, specific altitude.
Challenge
The professor conducting research into the DEAN engine worked with several graduate students to study its suitability for one IHPRPT mission. The graduate students used NASA’s Numerical Propulsion System Simulation (NPSS) software to represent the pumps, turbines, pipes, valves, chamber, nozzle and cooling channels in analyzing the performance of the DEAN engine. The NPSS model captured the engine geometry and other inputs as hard-coded values so each variable had to be recoded each time it was mentioned in the code to evaluate a new design point. This process took weeks, making it impossible to evaluate more than a few design points. The students demonstrated that the DEAN thrust, specific impulse (change in momentum delivered per unit of propellant consumed), and thrust to weight ratio exceeded the performance of the current generation RL10 engine. But they had no idea how close their baseline design was to the optimal design for any particular mission.
Solution
DEAN simulation workflow in ModelCenter
Simmons had previously demonstrated his ability to set up workflows using the ModelCenter framework to automate a range of different optimization problems so he was asked to get involved in the project. He started out by recoding the NPSS DEAN model to convert hard-coded values such as the chamber length, inner and outer chamber radii and oxygen to fuel ratio (O/F) into parameters that could be accessed from the command line. He created a ModelCenter script to generate the geometry of a design point and calculate the geometric parameters needed to drive NPSS and display the design. He used ModelCenter to create an automated workflow that drives the parameterized DEAN model to run trade studies. The cycle time for analyzing a single design point with this workflow is only 4 minutes. Simmons used the local load balancing feature of ModelCenter to run 5 cases in parallel to further increase the speed at which trade studies consisting of large numbers of design points could be run.
DEAN optimization results for original design case
Simmons used the workflow to run trade studies examining the effect of chamber length, O/F and total mass flow on thrust and specific impulse to establish the rough boundaries of the design space. This study helped reduce the size of the DEAN’s base design by 40% by eliminating unneeded engine length. He then expanded the workflow to capture additional performance parameters including materials constraints, engine weight, engine geometry and turbomachinery performance. He ran a new set of trade studies on the expanded workflow to evaluate the sensitivity of the performance parameters to the design variables. Using the results of the sensitivity analysis he formally defined a multi-objective optimization process utilizing Designs of Experiments (DoEs) and the Darwin optimizer in ModelCenter. One of the key steps in the optimization process was to seed the optimizer with the best designs from a DoE study. The workflow iterated to an optimal design for the original case in two days. Around this time, a professor asked him to optimize DEAN for five additional missions. With the model parameterized and the workflow created, this was simply a matter of adjusting the objectives and design parameter boundaries to match each additional case and re-running the optimization studies.
Benefit
Optimized DEAN engine exceeds requirements and outperforms traditional engines for three missions based on the IHPRPT and NGE programs
Optimized designs of the DEAN are smaller, lighter, and better performing than their traditional counterparts. The table above compares the DEAN to the RL10 engine for missions ranging from 25,000 lbf to 35,000 lbf vacuum thrust. In all three cases the DEAN is one quarter the length and one fifth the radius of a comparable RL10 while exceeding the specific impulse of an RL10 by as much as 5 seconds. These performance gains are achieved by doubling the chamber pressure of an RL10 and leveraging the compact size of the DEAN’s aerospike nozzle. Similar performance gains are seen for the X-37 and SLS missions, giving the DEAN improved performance over traditional upper stage engines over a vacuum thrust range of 6,500 lbf to 100,000 lbf.