Towards a Systematic Data Harmonization to Enable AI Application in the Process Industry

Abstract

Current methods of artificial intelligence may often proof ineffective in the process industry, usually because of insufficient data availability. In this contribution, we investigate how data standards can contribute to fulfill the data availability requirements of machine learning methods. We give an overview of AI use cases relevant in the process industry, name related requirements and discuss known standards in the context of implicit vs. explicit data. We conclude with a roadmap sketching how to bring the results of this contribution into practical application.

Process Digital Twin and Its Application in Petrochemical Industry

“…However, many data issues still exist, some connected with the volume and speed of data acquisition, some connected with reliability and uncertainty, some to do with dynamic model updating, and others related to data sharing and exchange standards, all of these hindering the chemical process industry to efficiently apply data-driven technologies for their assets in large scale. There need improve the data integration across the whole asset life cycle from process design over the functional design and asset specification up to the operation of the actual assets, building an Industrial Internet of Things (IIoT) platform to harness data integration and enable mashup applications like digital twins [41].…”

Definition, Characterization, and Modeling of Hybrid Modular‐Monolithic Process Plants

“…Furthermore, dependencies and interrelationships between elements in information models need to be modeled. Here, the data exchange in the process industry (DEXPI) activities should be mentioned [21], which are intended to facilitate the data exchange between simulation and P&ID programs in process industries. These models support the greenfield and brownfield engineering of hybrid plants by providing information about the monolithic and modular parts.…”

“…In the future, artificial intelligence (AI) tools may assist process steps in the engineering workflow [71]. The seamless data integration from process simulation with the export to DEXPI can enable their import into P&ID packages and AI-assisted engineering tools [21]. Machine-readable formats and structures are important for data transfer and analysis, which are currently under development [69].…”

“…In some way, seamless engineering is currently not possible due to different departments of process development in companies, engineering with different disciplines, and separated operation and production. A common database and machine-readable structure are essential for this approach [21]. There is a high effort for applying modules to other, already existing or conventionally planned, monolithic plant types, particularly to dedicated single-product plants.…”

“…Standards play a crucial role in the design and operation of such systems. Today, standards in cyber-physical systems and systems engineering are the key to higher efficiency in data analysis with consistent access [20,21]. Industrial standards were formed during the Industrial Revolution, e.g., the tube diameters or the Whitworth thread in 1841 based on company standards [22].…”