Datasets for data science investigation of fused EM/RF
and vibroacoustic equipment monitoring

Tom Grimes - PNNL
Lynn Wood - PNNL
Karl Pitts - PNNL
Nathaniel Smith - Pacific Northwest National Laboratory
Jihee Yang - PNNL
Eva Brayfindley - PNNL
Elisabeth Moore - PNNL
Jan Irvahn - PNNL
Jeff Miller - PNNL
Jack Dermigny - Pacific Northwest National Laboratory
File Attachment
We present a dataset for enabling the use of deep learning for understanding the state of operating machinery. This dataset includes voltage and vibroacoustic measurements taken in a phase-locked fashion on a variety of common office and lab equipment. All included equipment connects to the wall via standard electrical cords with a strong emphasis on small hand tools. The distribution package includes the raw voltage and vibroacoustic data, the metadata, a Jupyter notebook that trains and evaluates a baseline deep learning model for performing useful sample tasks (e.g., building a classifier to determine if a single piece of equipment is on or off), a modifiable PyTorch dataloader, and a file to build an appropriate python environment. Using these tools, the next generation of data science practitioners can work toward newer and better approaches specifically for analyzing and understanding signals from operating machinery.