AI can create an industrial data visualization from an engineer’s request or suggest one using information already associated with an asset. The engineer does not necessarily need to write SQL or configure every chart, but still needs to check that the result shows the right equipment, measurement, and time range.
Consider a production engineer who wants to see how packaging line speed changed during the night shift. Assume the line speed is already modeled as an operational measurement. A useful chart needs the correct line and measurement, the shift’s start and end times, and enough data to interpret the result. A chart that looks plausible can still answer the wrong question.
What does “without writing SQL” mean?
In an AI-assisted visualization workflow, a user can describe what they want to see in ordinary language. The system interprets the request, finds available data, and produces a visual display. Depending on the product and configuration, it may generate and execute a database query behind the scenes. The user does not have to write that query for the workflow to involve SQL.
A generated panel is one visualization, such as a trend, gauge, or table. A complete dashboard can contain several panels arranged for a particular purpose. Generating one panel does not by itself mean the system has assembled a complete dashboard or selected every metric a team needs.
Give the system enough context to build a useful view
Start with the operational question. “Show the hourly average speed for Packaging Line 2 during last night’s shift” is more specific than “show production performance.” The first request gives the system a better chance of selecting the right measure and interval.
Before relying on a generated chart, make sure the underlying data has enough context:
- Equipment: Which line, machine, or sensor produced the measurement? Similar names across sites can lead to the wrong selection.
- Measurement and unit: Is the value a count, a percentage, or a rate? Does the system know its unit and what it represents?
- Time range: Does “last night” match the site’s shift boundaries and time zone?
- Aggregation: Does the question call for an hourly average, a total for the shift, or a maximum? These can tell different stories.
- Operating conditions: Was the line running, stopped, or changing products during the period?
This information can come from the request, the data model, or both. Industrial data contextualization describes how asset and measurement context makes operational data easier to interpret. If the relevant data is missing or the asset mapping is wrong, a more polished chart cannot fix the underlying problem.
Suggested panels and requested panels solve different needs
Some systems recommend visualizations before an engineer asks for one. These suggestions can be a useful starting point when someone opens an unfamiliar asset and wants to see which measurements are available. A user-requested panel starts with a more specific question, such as a trend for one shift or a comparison between two periods.
For example, a suggested panel might show a packaging line’s current speed. The production engineer may instead need an hourly trend for the previous shift, alongside line state. The suggestion helps identify a relevant measurement; the engineer’s request narrows the period and the purpose.
In TDengine IDMP, AI-Generated Panels are documented as suggestions based on an element’s attributes, template, and collected time-series data. Users can also describe a desired panel in natural language. Suggested panels can be reviewed and saved when useful. The availability of the workflow depends on the deployed version, data, permissions, and AI configuration; IDMP’s LLM connection documentation explains the connection requirement for LLM-based features.
How should an engineer check the generated visualization?
Use the chart as a starting point for inspection. Before sharing it or using it in a production review, check the following:
- Confirm the asset and signal. Check the source and measurement name. Confirm that the selected speed measurement and unit are appropriate for the line.
- Inspect the time window and data freshness. A shift crossing midnight or a delayed data feed can make a correct-looking chart misleading.
- Check units and aggregation. An hourly average can hide brief stops or speed spikes. Check whether stopped periods are included and whether the aggregation matches the production question.
- Look at the chart itself. Check the axis scale, missing intervals, and whether the chosen chart makes changes visible. An unmarked gap is not necessarily a zero.
- Compare with operating context. A change in line speed during a product changeover may need a different interpretation from the same change during steady production.
If the result does not match the question, refine the request or check the source data with the person responsible for the historian or asset model. For a production decision, compare the result with an established trend or other authorized evidence when available.
When is a generated panel useful?
A generated panel can help an engineer explore an unfamiliar asset, prepare a shift review, or make a first pass at a production question. It is less useful when the underlying measurement is ambiguous, the time series is incomplete, or the team cannot explain how the metric is calculated. In those cases, resolve the data or definition first.
The goal is a visualization that answers a specific production question and can be checked against its source. TDengine’s AI-Powered Insights page describes how its generated panels fit into a broader industrial data workflow.


