- Home
- Research
- Reduced-order modeling
Research
Reduced-order modeling
Reduced-order modeling (ROM)?
- The dimensionality of the system can be very large → very expensive!
- Reduced-order modeling (ROM): there is a correlation between parameters. Reduce the dimensionality of the system from M to r
- Dimensionality reduction techniques: Principal Component Analysis (PCA), Singular Vector Decomposition (SVD), CUR, kernel-PCA, ...
PC-transport ROM
- Reduce the high-dimensional composition space of the combustion system to a lower-dimensional principal component (PC) space
- We apply PC-transport ROM to a complicated combustion system, represented by a highly-transient compression ignition of large hydrocarbon fuels (>100 species) at high thermodynamic pressure (40-100 atm)
TDB-CUR
- Time-dependent-bases with CUR decomposition (TDB-CUR) has the following unique features (i) the bases of the ROM is adapted over time to minimize the low-rank approximation error, (ii) offline training dataset & training process are not required, (iii) stiffness of the multi-physics systems can be mitigated
- We introduced the TDB-CUR compatible with combustion systems and validated a hierachy of the combustion systems, ranging from 1-D laminar flame to 3-D turbulent flames.
TDB-L-CUR
- Currently working