James B. Fitch 1, Kanishka Sharma 1, B. Taylor Newill 2, Bhaskar Bhatnagar 1 1 Navistar, 2 NOESIS Solutions

Transcription

1 Automation of Java file creation and batch mode submission for StarCCM+ Simulation models to be used in a Design of Experiments and multi objective optimization. James B. Fitch 1, Kanishka Sharma 1, B. Taylor Newill 2, Bhaskar Bhatnagar 1 1 Navistar, 2 NOESIS Solutions Exploring the design space of a simulation requires the preparation of multiple CCM+ simulation files. Setting various run dependent conditions such as inlet velocity and rotating body angular velocity can be done using CCM s batch mode and a internally defined java macro file. When the run dependent conditions need to altered to conform to a statistically defined Design of Experiments or as part of a multi objective optimization the amount of simulation preprocessing that needs to occur becomes unmanageable for the CFD analyst. This paper investigates the process of using commercially available third party software to integrate with CCM+ in order to automate the creation of the Java macro files that will define the CCM+ run. The OPTIMUS software package from NOESIS Solutions was used to quickly set run conditions and automatically submit the runs to an external cluster. The setup was easy and repeatable and enables many simulations to be run with low user interaction. In addition the validation of this process enables the setup for more complex automated simulations. The work shown will also demonstrate the time savings and performance that was realized by using this tool.

2 Automation of Java file creation and batch mode submission for StarCCM+ Simulation models to be used in a Design of Experiments and multiobjective optimization.

3 Introduction Optimization is necessary at Navistar CCM+ makes it easy to create Java files Scripting is still required to automatically modify run conditions Time consuming Not scalable OPTIMUS is a tool that can quickly automate the entire process OPTIMUS allows the user to easily integrate with distributed computing OPTIMUS drives efficient optimization

4 Innovating industrial transportation since 1831 International ProStar is the most aerodynamic and fuel efficient Class 8 truck ever First Class2c 3 vehicle to receive CARB certification as a ZEV

5 Noesis Solutions 15+ years of strong Optimization fundamentals A unique provider of simulation engineering more than 15+ years 100+ persons/year experience in Optimization and Simulation Process Automation. The largest OEM provider of embedded optimization A solid innovation partner for more than 2000 manufacturing companies auto, aerospace and other industries. A worldwide presence, a strong financial track record 20+ offices in Europe, US and Asia and double digit profitable growth for 15+ years.

6 CFD Model Generic Semi Truck Rotating Wheels 25 m/s Vary yaw angles between +/ 10 degrees

7 Geometry Conditions Modify cab width at rear of cab Modify windscreen angle Maintain element quality Import deformed mesh

8 Automation It was taking 1.5 hours to modify settings for each additional run condition Check mesh for each run Then automate java file creation Change Geometry Change Flow Conditions Update Java File Solve Model

9 CCM+ Control Requirements Read in template java file Identify variables to change Replace blocks of java file if necessary Execute CCM+ analysis at a remote site

10 Shape Change Mesh Analysis CFD Analysis

11 OPTIMUS CCM+ Interface Automatically identify model parameters Launch Analysis

12 Batch Submission Optimus uses a Remote Management System Sequences each analyses separately (deform mesh, run solver, post process, etc) One click integration to submit jobs to Purdue PBS system Then provide credentials and run time commands

13 Design of Experiments Quickly setup and run a Minimax Latin Hypercube of 25 experiments The design points are chosen such that the maximal distance of any point to its closest design point (i.e. the covering radius) is minimal

14 DOE Results

15 Multi Objective Optimization Minimize overall drag for a given yaw US Semi trucks must get 7 MPG by 2014 Maximize volume of Cab Size of the cab influences driver comfort Non Dominated Sorting Evolutionary Algorithm (NSEA+) Advanced algorithm simulates evolution in nature

16 Baseline DOE Best Pareto Front

17 Results (Performance Animation)

18 Results (Time Savings) Operation for a given yaw 1 st Experiment Next 199 Experiments With OPTIMUS Prepare Model 18 Hours 0 Hours 18 Hours Create Java File 2 Hours 0 Hours 2 Hours Update / Workflow 0 Hours 0.5 Hours / exp 5 Hours Prepare and Submit 2 Hours 1 Hour /exp 0 Hours Total 22 Hours Hours 25 Hours Savings Hours

19 Results (Performance) Experiment Volume (m 3 ) Drag (normalized) Baseline DOE Optimal RSM Optimal Optimal Improvement 6.7% 9.7% 0.75 m 3 improvement is equal to a double bed 10% Reduction in drag could reduce fuel consumption by 2000 gallons / year

20 Conclusion OPTIMUS enables quick, automated, setup for design space analyses and optimization Entire shape deformation, run conditions, and analysis automated Significant improvements were found in drag and driver comfort