1.PhD Thesis and the role of HPC, 2014

Problem definition

Absence of waste engine oil in-situ treatment capabilities led to research the possibility of introducing one. This is especially significant when we talk about reusing the oil/lubricant of the career truck who use more than 3 tonnes of oil vs 5 qz in passenger cars

Conditions to introduce mobile/in-situ filtration

The treatment option had to be adapted to existing conditions on the car: meaning no change in car designs

Proposed solution

Use inorganic tubular membranes as cross-flow filtration which can be reused and recycled and prevent pressure loss in the lubrication system

How fine has to be the filtration media?

Depends on the impurities and in our study found to be between micro and ultrafiltration.

Research methodology

Lab site procedure of the work:

So the work started with collection of filtration media samples
Using scanning electron microscope (SEM) scan the tubular inorganic membranes from all sides
Measure the porosity based on SEM images
Collect waste oil samples and measure their viscosity and density

Numerical methods applied

Computation work procedure:

Recreate the 3D structure from SEM images:
- Image processing with ImageJ;
- 3D structure reconstruction using MatLab
Using PALABOS solver solve the CFD problem of average flowrate, and then visualize using ParaVIEW and 20 CPU HPC and 91% accuracy.

Vizualization and Findings

Results of the study in terms of the lubricant: reconstructed 3D homogenous inorganic porous media vs flowrate

Comparing the numerical results with other research setups, a case below is with water streamlines, to validate the findings

The HPC Center used for the study

The EnerGRID computational facility consists of a Transtec HPC cluster: 72 total virtual cores; 144 GBtotal RAM; 5.5 TB total disk capacity (3.0 TB failure free); double networking by both Infiniband and GBit. The tested peak performance (according to the TOP500 standard) is 376.09 GFlop/s with mvapich2/IB (which is roughly 60% of the theoretical peak performance 596.48 GFlop/s). We would like to acknowledge Thomas Zeiser of RRZE for the cluster facility start-up and maintenance.

8 Computational Nodes:

2 Quad-core Intel Xeon, 2.33 GHz processors, corresponding to 8 virtual cores;
16 GB RAM, leading to 2 GB/core ratio;
250 GB SATA hard drive, 7200 rpm.

Master Node:

2 Quad-core Intel Xeon, 2.33 GHz processors, corresponding to 8 virtual cores;
16 GB RAM, leading to 2 GB/core ratio;
2 x 250 GB ATA-2 hard drive, 7200 rpm;
4 x 1000 GB (1 for parity) SATA-2 hard drive, 7200 rpm.

Networking:

Infiniband Switch with 24 ports;
GBit Switch with 24 ports.

Page updated

Google Sites

Report abuse