Diagnosing Performance Equation Evaluation Issues in PI System: OR vs AND Logic and Data Freshness

Performance Equations (PEs) in OSIsoft PI System are powerful tools for real-time data analysis. However, subtle configuration choices can have significant effects on the results, leading to outcomes that may initially seem perplexing. In this post, we address a common scenario: a PE intended to flag a process as "OK" if either of two tags is "OK", but which sometimes returns unexpected values. We'll explore the nuances of Boolean logic and the critical role of triggering and data freshness.

---

The Scenario

Suppose you have two status tags, tag1 and tag2, and you want to create a PE that returns "OK" if either tag is reporting "OK", otherwise "FAILED". It's tempting to write something like:

IF TagVal('tag1', '*') = "OK" OR TagVal('tag2', '*') = "OK" THEN "OK" ELSE "FAILED"

Yet, in practice, you find the PE returns "OK" even when only one tag is "OK" and the other is "FAILED" or, conversely, it doesn’t return as expected in some situations. What’s going on?

---

1. Confirming the Boolean Logic

First, let’s clarify the logic:

• OR means either tag can be "OK" for the result to be "OK".
• AND means both tags must be "OK" for the result to be "OK"; otherwise, you get "FAILED".

Truth Table Example:

If the goal is to show "OK" when either tag is "OK", then the original equation is logically correct. But why might the PE return unexpected results?

---

2. Data Freshness and Sampling Synchronization

A less obvious—but common—culprit is data freshness due to how and when calculations are triggered in PI AF Analytics or PE Scheduler:

• If your tags are updated at different intervals or scan classes, one may still have an old value when the calculation runs.
• For example, if tag1 updates every minute and tag2 every 10 seconds, a PE triggered by new data from tag2 might use the most recent (possibly stale) tag1 value.

This means that, depending on which tag triggers the calculation and their update frequencies, your PE might:

• Use a current value for one tag but a stale one for the other.
• Return incorrect results based on non-synchronized tag values.

---

3. Solutions and Best Practices

a) Triggering Mechanism

• Natural Scheduling: Calculation triggered when new data arrives on either tag (default). Can lead to the use of stale values.
• Clock-based Scheduling: Forces calculation at fixed intervals using the most recent values for both tags, reducing data skew.

Recommendation: For multi-tag logic, favor clock-based scheduling.

b) Data Age Checks

You can explicitly check the time difference between now and the latest sample for each tag. If a tag’s value is too old, output an error indicator (e.g., "I/O Timeout" or "Calc Failed").

Example:

IF SecSinceEvent('tag1','*') > 60 OR SecSinceEvent('tag2','*') > 60 THEN "I/O Timeout" ELSE IF TagVal('tag1','*')="OK" OR TagVal('tag2','*')="OK" THEN "OK" ELSE "FAILED"

This will alert you to stale data situations.

c) Re-check Logic Requirements

Make sure the Boolean logic matches your operational requirements:

• Do you want OK if either tag is OK (use OR)?
• Or only if both are OK (use AND)?

---

4. Key Takeaways

• Boolean logic in PEs must align with process requirements (OR vs AND).
• Calculation schedules and data age can impact PE results—especially with tags updating at different rates.
• Clock-based triggering and data age checks help ensure accurate, up-to-date evaluations.

---

Conclusion

Performance Equations are central to process status monitoring in PI. When results seem off, it's often not a logic error but a matter of data freshness and triggering. By understanding and adjusting the scheduling and adding data age validation, you can make your PEs more reliable and insightful.

Have you run into unexpected results with your PI Analytics or PE setups? How did you resolve the issues? Share in the comments below!