Extracting Signature Features from Seismo-Acoustic Data for Automated Algorithm Development Title on narrative: Correlating Signature Features with Processes of Interest Using Collected Seismo-Acoustic Data Sets)

Persistently monitoring of process facilities by recording seismo-acoustic S-A) data over the long- term naturally results in very large quantities of data. After converting the data into more useable forms, skilled analysts must spend time systematically perusing all data for signals of interest, requiring a large investment of labor hours. Furthermore, manual analysis also makes it difficult to report the results in a timely manner. Quantums overall goal for Phase I is to demonstrate the ability to correlate signature features with processes of interest using collected data sets. Quantums ultimate goal in Phase II is to devise a set of algorithms and software that will automatically extract the uniquely identifiable feature, or group of features, from collected data to discover the specified processes of interest with high confidence. For Phase I, the objectives are: 1) understand the nature of the S-A signals associated with the processes of interest, and obtain the collected data sets to be used; 2) Analyze collected data to identify unique signature features indicative of the processes of interest; and 3) demonstrate the ability to correlate signature features with processes of interest, and draft a preliminary Software Design Document for the types of algorithms that will need to be developed in the Phase II effort. The key to a successful Phase I effort is the analytical skill and years of experience Quantums scientists have at actively extracting distinguishing features from a variety of S-A signatures to develop classification algorithms used in the S-A sensors Quantum produces. For example, Quantum desires to develop a capability to monitor oil and gas pipelines for S-A events such as detecting new leaks, validating operational changes, and alerting authorities to human tampering. Quantum is producing S-A sensor systems for the commercial security monitoring and surveillance markets. If successfully carried over into Phase II, the new automated algorithms developed for DOE will expand Quantums product offerings for government and commercial applications. Devising feature extraction algorithms will broaden the scope of the types of S-A signals Quantum systems can automatically detect, classify, and report. The methods and techniques used to extract signature features of processes of interest in this project will translate into improved classification algorithms in new areas of security monitoring and surveillance.