SNS Neutron Scattering

Grand Prize Winners

Willem Blokland (NSD), Joe Daws (UTK), Yun Liu (NSD), Armenak Petrosyan (CCSD), Viktor Reshniak (CCSD), Jeremy Trageser (CCSD), Mark Wendel (NSD), & Drew Winder (NSD)

The SNS target module is a consumable component which wears out due to radiation damage and cavitation erosion. Fiberoptic strain sensors have been installed on the last 6 target modules and are now routinely used to monitor the transient strain response of the mercury vessel to proton beam impacts. A large amount of data are collected on each target, with some of the sensors lasting the lifetime of the targets. One example of strain response is shown in Fig. 1. These data are expected to contain information that can be used to monitor damage to the mercury vessel and to improve the design of future targets so that higher beam powers can be used at SNS. As a first step, a spectral clustering algorithm was applied (Fig. 2) to the data collected on one of the targets to correlate beam power, measured frequencies, and the time after proton beam impact for the highest strain to occur. The goal was to discern possible changes in the target mechanical structure. The clusters did not reveal such patterns for this particular target, but several ideas for moving forward were identified that could be pursued in collaboration with various ORNL divisions. These ideas ranged from data processing for improving signal-to-noise to machine-learning applications with supercomputing and model optimization. Going forward, it is recommended to identify a funding source and develop a longer term plan of action for improving the data and using the data to improve target reliability.


Fig 1. Typical strain response to proton beam impact from one particular pulse at one particular location.	Fig2. Typical strain response to proton beam impact from one particular pulse at one particular location.

Runner Up Winners

Christian Balz (NSD), Arnab Banerjee (NSD), Andrei Savici (NSD), Hoang Tran (CCSD), & Barry Winn (NSD)

Adaptive Total variation Denoising of Neutron Scattering Data

A two-dimensional inelastic neutron scattering data set was acquired in such a way that the highly variable error per pixel was quantified. The resulting image was denoised using a regularization algorithm of the wavelet transform (WT) that weighted corrections using this error.

Most Integrated Project Between Facilities

William Godoy (CCSD) & Matthew Stone (NSD)

The Self-Shielding Background Problem

Powder inelastic neutron scattering measurements use aluminum sample cans to contain the sample. The sample can is an additional background in the measurement. Very often an empty can is measured to subtract this background from the original data. However, because the can is empty, the backside of the sample can receives more neutrons than the can which contained the sample. This effect is called self-shielding. To account for this a self-shielding factor ( a number less than one) is multiplied into the background measurement. Typically people will chose the self-shielding factor using a trial and error approach. This project determined an efficient way to experimentally determine the self-shielding factor using a numerical optimization technique.

Best Fusion of Community Codes

John Ankner (NSD), Jong Choi (CCSD), & Mathieu Doucet (NSD)

Layer Structure Determination with Clustering

The project addressed the issue of determining the layer structure of a thin film measured with neutron reflectometry. Reflectivity data was generated and put through a clustering algorithm to try to determine features that would indicate how many layers a thin film sample has.