What Causes Data Quality Issues in PI System?

PI data-quality problems can originate at the instrument, collection layer, archive, asset model, calculation, or display.

Understanding the failure categories helps teams investigate in the correct order.

Instrument and source-system problems

Examples include:

• Sensor failure

• Calibration problems

• Frozen values

• PLC or DCS communication loss

• Incorrect source configuration

Always verify the source before changing PI configuration.

Interface and connector problems

Collection problems can create stale data, gaps, or bad states.

Review service status, communication errors, buffering, source connectivity, and point mapping.

PI Point configuration problems

Incorrect point settings can affect data meaning and storage behavior.

Common examples include:

• Wrong source address

• Wrong engineering units

• Incorrect span

• Poor exception or compression settings

• Incorrect point source

Asset Framework mapping problems

AF can reference the wrong PI Point even when both the tag and AF attribute are technically healthy.

Template substitution errors and migrations are common sources of mapping drift.

Calculation problems

Analyses can fail, lag, or continue to use obsolete inputs.

A calculated output can remain numeric and appear valid after the source logic has become incorrect.

Display and reporting problems

PI Vision and external applications can reference obsolete tags, old servers, or deprecated AF attributes.

A display error is not always a source-data error.

Human and governance problems

Many technical failures start with normal engineering work.

Examples include:

• Tag renames without dependency review

• Undocumented template changes

• Manual data entry without validation

• Duplicate signals created during projects

• Old objects never retired

Scale and ownership problems

Large PI environments accumulate issues when ownership is unclear and checks are manual.

Use standards, automated inventory, recurring health checks, and clear ownership to reduce this risk.

The practical troubleshooting order

When a value is wrong, trace from source to use. Check the source, collection, PI Point, AF mapping, calculation, and display in sequence.

This avoids changing the wrong layer and helps the team identify the actual failure point faster.