(728g) The Use of Asset-Oriented Data Models for Data Integration Enables Advanced Analytics in the Process Industry

Industrial organizations in resource extraction, continuous or discrete manufacturing, and power generation each generate large volumes of data during the operation of their plants and facilities. The effective use of this data to drive business outcomes is a central theme behind recent trends in IIoT or digitization of industry as well as cross-cutting efforts such as Industry 4.0. Machine- or sensor-generated process time series data, engineering design data, maintenance records, and business records are often separate data silos within large organizations. The existence of these disparate data silos has historically hindered the effective implementation of advanced analytics in areas such as asset benchmarking, process monitoring, fault detection, and machine learning for predictive maintenance or other applications.

This work shares practical lessons from the industrial application of advanced analytics in prototype and production phases. It also highlights real analytics use-cases and the transformation in data management and information architecture necessary to enable these activities. Several commonly observed challenges in enabling data scientists and quantitative analysts to be effective with process data are discussed. The utility of asset-oriented data models is shown in an example application with a realistic set of databases; including the data from instrumentation, control systems, and transactional business records. The data and data model requirements of business intelligence technologies and advanced analytical applications are discussed, with a focus on how a shared data layer must be flexible to support a variety of analytics activities.

A property graph structure stored in a graph database can effectively support the storage and construction of relationships relevant to an asset-oriented data model. The ingestion of structured information complying with established standards for data interchange can accelerate the construction of such a graph model. This is illustrated with the use of piping and instrumentation diagrams (P&IDs) to establish the process topology as well as link data streams from instrumentation with process units and higher-level views of a plant.