Best Practices for Performing Data Mining from Unstructured Data for Mechanical Integrity

Mechanical integrity program implementations require extensive design and operation data in order to be successful, however the standard methods for evaluating potential data sources, collecting data, reconciling to the laster equipment list and locating relevant and missing data have not substantially changed in several decades, as the standard process is to send engineers and inspectors to the field with scanners in backpacks to manually review and capture data in spreadsheets.

This presentation reviews best practices for accomplishing these tasks using modern methods which incorporate machine intelligence to auto-locate relevant data and execute data extraction, and use database rules to improve data quality prior to ingestion into predictive analytics software applications used to predict failure and better manage CAPEX and OPEX.  The benefits to the operator and engineering community include reduced costs to perform the same amount of work, in shorter time frames and allowing engineering resources to focus on higher level tasks.