Accession Number:

AD1079394

Title:

Data Analytics for Large Sensor Systems

Descriptive Note:

Technical Report,01 May 2015,30 Apr 2017

Corporate Author:

Virginia Polytechnic Institute and State University Blacksburg United States

Personal Author(s):

• Smith,Eric P.
• Alexander,William N.
• Davenport,William J.
• House,Leanna L.
• Leman,Scott C.

Report Date:

2017-09-27

Pagination or Media Count:

34.0

Abstract:

Modern wind tunnel testing involves multiple diverse sensor systems with anywhere from hundreds to thousands of individual sensors. These large-scale tests can be quite expensive and are usually time-sensitive. As such, any delays due to faulty instrumentation can have serious consequences. Equally serious is the possibility of discovering a sensor failure after the test has been completed, since time and effort will have been spent collecting what amounts to noise. Although sometimes correctable, the time involved in doing so distracts the experimenter from achieving the experimental goals. As such, any large scale sensor system needs methods to ensure that all the individual sensors are working as intended. Most commercial sensor systems contain rudimentary error detection for sensors within a given system, but these methods typically have to way of incorporating information about the ambient conditions under which they were run, or, more importantly, information from the output of other systems which are made of smaller, unique sensor subsets. By combining the information from diverse sensor systems into a global error detection process, we can measure the extent to which sensors across systems are correlated and use that correlation information to produce more powerful predictions and error detection capabilities. This report summarizes approaches that have been developed to evaluate sensor systems both for a single type of sensor as well as for systems that include multiple sensor types. A MATLAB macro is described to analyze arrays of data arising from a study utilizing a single sensor type. An approach to analysis of sensor systems using multiple types is described based on a Gaussian process model. Examples are provided to illustrate applications of the methodologies.

Descriptors:

• data analysis
• monitoring
• sensors
• arrays
• wind tunnel tests
• Gaussian processes
• data sets
• microphones
• statistical analysis
• graphical user interface

Subject Categories:

• Test Facilities, Equipment and Methods
• Statistics and Probability

Distribution Statement:

APPROVED FOR PUBLIC RELEASE