
The PI Data Archive stores time-series data under tag names. PI Asset Framework (PI AF) adds the context: which pump, which line, which site. OSIsoft describes PI AF as a single repository for asset-centric models, hierarchies, objects and equipment. This guide explains the building blocks and how to set up an asset model that scales.
PI AF is also the basis for reusable PI Vision dashboards.
Elements and attributes
An element represents a piece of equipment or a process, and data is attached to it through attributes. According to the AVEVA glossary, an attribute can hold fixed information, such as the diameter of a tank, or use a data reference to get values from a PI point, a PI point array, a formula or a table. The AF SDK documentation adds constants and values from relational databases to that list.
Templates
Element templates define a set of base attributes for all elements based on them. You create the template once and as many elements from it as you need, which saves time and keeps the model consistent. Templates also exist for event frames, transfers, models and notifications (AVEVA documentation).
Analyses
An analysis reads values of PI AF attributes, performs calculations and writes the results to other attributes. The analysis types are expressions, rollups and event frame generation, and analysis templates keep analyses consistent across all elements of a type. The PI Analysis Service runs analyses on a schedule. AVEVA notes that saving analysis outputs to PI points also results in better PI AF performance.
Event frames and notifications
An event frame captures the start and end times and relevant asset data of a significant plant event, such as equipment downtime, a process excursion or a batch run. PI Notifications builds on them: notification rules are based on event frames and send email messages or call a web service.
Practices for a scalable asset model
OSIsoft’s best practices for PI AF still hold:
- Ask who will consume the data before you design the hierarchy.
- Start small, solve a specific problem first and build up. There is no single right way to build AF.
- Use templates, so every asset of a type has the same structure.
- Write the results of scheduled analyses to PI points.
Check the data underneath as well: tags with poor exception and compression settings weaken every model built on them, as our PI Tag Optimization page explains. For a real example, read how a PI AF template library helped Total affiliates get a head start on AF.
See your AF model and what depends on it
The Amitec Industrial Knowledge Graph scans your PI Data Archive, AF server and PI Vision and shows how PI tags, AF structures, PI Vision displays, PI analyses and Seeq workbooks relate, so you know what a change affects before you make it.
Frequently asked questions
What is PI Asset Framework (PI AF)?
PI AF is the part of the AVEVA PI Server that models assets, hierarchies and equipment and links them to time-series data in the PI Data Archive.
What is an AF element template?
A stored definition of an element type with its base attributes. Elements created from the template share the same structure.
What is a data reference in PI AF?
The source of an attribute value, for example a PI point, a formula, a table lookup or a value from a relational database.
What are PI AF analyses?
Calculations that read attribute values, compute results and write them to other attributes, for example expressions, rollups and event frame generation.
Do we need PI AF for PI Vision?
PI Vision works with PI points alone, but AVEVA recommends PI AF to get the most from PI Vision, for example asset swap and collections.
Related posts
How PI Vision custom symbols work, the rules of the extensibility framework, and when a ready-made add-on such as the PI Vision Suite saves the work.
Practices for AVEVA PI Vision dashboards: PI AF templates, asset swap, collections, multi-states, calculations, performance and display management.




