Sandia Labs FY21 LDRD Annual Report


Coupling power flow to target simulations in support of next generation pulsed power facilities. A one-year project under the Assured Survivability and Agility with Pulsed Power Mission Campaign explored problem reduction techniques for powerflow calculations in large pulsed power facilities to both improve time to solution and enable the characterization of electromagnetic drive in powerflow calculations. This work expanded the transmission line coupling capability in the ElectroMagnetic Plasma in Realistic Environments (EMPIRE) modeling and design tool— developed under the Plasma Science and Engineering Grand Challenge LDRD (FY18-FY20)—to include representations of Transverse Magnetic (TM) waves. TM waves, a generalization of the fundamental Transverse Electromagnetic (TEM) mode previously simulated, can represent pulse asymmetry in the system. In addition to a higher fidelity representation of waves

An EMPIRE-enabled simulation shows how transverse magnetic wave perturbation drives the lowest level magnetically insulated transmission line of the Saturn accelerator.

in transmission line models, the team applied the theory developed during the LDRD to produce new synthetic diagnostics that allow analysts to quantitatively assess the regularity of the electromagnetic fields in a powerflow simulation. Using the new capabilities, the team evaluated the pulse regularization of the driver used in the Saturn accelerator. With such prototype calculations, it will be possible to evaluate the science-based design plans for the next generation pulsed power facilities at Sandia. (PI: Duncan McGregor)

Data science for detection of genome editing. The current explosion of advances in gene editing technology and public accessibility to these techniques pose potential biosecurity threats. To accurately assess the risk to national security, bioengineering threats must be clearly distinguishable from natural genome variation, or edits. This LDRD project focused on a gap for detecting edits without prior information of the most likely regions in the genome where the edit effects will occur. The team employed three specific classes of algorithms: (1) decision-tree learners to identify subtle patterns and signatures that indicate an edit, (2) two complementary architectures of Deep Neural Networks learners to classify edit versus non-edit regions and learn grammatical patterns indicative of edit versus non-edit regions in next generation

deep sequencing DNA reads, and (3) novel Anomaly Detection algorithms to characterize edit versus non-edit regions. Using large data sets from various genome editing experiments, the team built a data- processing pipeline to handle raw data file formats and provide genomic noise counts, and ultimately created features now used in machine learning algorithms. Results show significant evidence of success in detecting targeted edits in the human genome. The developed capabilities will provide both decision support for the national security community in assessing potential biosecurity threat risks and possible consequences and options for mitigation and/or forensic investigation. (PI: Stephen Verzi)



Made with FlippingBook Ebook Creator