How TDengine Historian Heatmaps Reveal Equipment Operating Patterns

In a typical factory scenario, 30 compressors report three Tags: temperature, pressure, and vibration. Over three months, that adds up to tens of millions of data points. The process director wants to know which operating range the compressors spend the most time in.

Most monitoring panels rely on Line Charts and Bar Charts. They can pinpoint a temperature spike on a specific day, but they cannot show how the data is distributed overall. That is not a flaw in Line Charts. Different chart types answer different kinds of questions.

Starting with TDengine Historian v1.0.19, the Visualization panel supports Heatmaps. The idea is simple: divide two Attributes into buckets, count how many samples fall into each pair of bucket ranges, and use color intensity to show density.

1. Different charts answer different questions

Common industrial monitoring charts each have their limits:

• Line Charts show trends: is the temperature rising or falling?
• Bar Charts compare values: how large is the energy consumption gap between Pump A and Pump B?
• Scatter Charts show relationships: how does rotational speed correlate with vibration?

In industrial monitoring panels, these charts often start from time. A Heatmap works differently. It looks at two Attribute dimensions at once and shows how data is distributed across the space where they intersect. It is not focused on what happened at a specific moment. It shows where historical data points tend to cluster.

Take a variable-speed device with rotational speed from 300 to 1200 rpm and load from 20% to 100%. After three months, you want to know where the device spends most of its runtime across different speed-load combinations. A Line Chart or Bar Chart will not answer that directly. A Heatmap will.

Bucket speed into intervals (say, every 100 rpm), bucket load into intervals (every 10%), count the samples in each intersecting cell, and color each cell by count. The darkest cell shows the dominant operating range. The entire process is just two-dimensional bucketed statistics. No complex algorithm is required.

2. How to use the Heatmap

To configure a Heatmap in TDengine Historian, choose one Attribute for the X-axis and another for the Y-axis, set the bucket size, choose a color scheme, and load the data. The axes can use any Element Attribute: rotational speed, temperature, pressure, vibration, current, or the time dimension itself.

Bucket size controls resolution. Coarse buckets (100 rpm per bucket,) give you the broad contour. Fine buckets (one per 20 revolutions) reveal more detail but need more data to produce a useful view. TDengine Historian supports two bucketing methods: fixed width, where each bucket covers the same value range, and fixed count, where the full range is divided evenly into a set number of buckets. There is no universal rule for bucket granularity. It depends on what you are trying to learn.

For color, TDengine Historian supports several gradient schemes, from single-color to multi-color gradients. Typically, darker colors represent more samples, while lighter colors represent fewer.

Usage recommendations:

Heatmaps complement Trend Charts. They do not replace them. Heatmaps are useful for finding distribution patterns and unusual clusters, but they cannot show exactly when something happened. When you spot an anomaly, switch back to the trend line to inspect the Time Series.

The selected data range affects what you see. Color mapping is calculated automatically from the minimum and maximum values in the current data range. A few extreme outliers can compress normal data into a narrow band of nearly identical colors. Narrow the query range or filter out the extremes.

Match the bucket count to the data volume. If buckets are too fine for the available data, most cells will be empty or contain only one or two points. The Heatmap becomes sparse and noisy. In general, aim for most populated cells to contain dozens to thousands of samples.

Use a logarithmic scale when values span multiple orders of magnitude. If trace leakage and normal flow rate differ by three orders of magnitude, a linear scale will make the smaller values almost invisible. With a log scale, each order of magnitude gets equal visual space.

Filter sparse cells so hotspots stand out. Scattered cells with counts of 1 or 2 are usually noise, not signal. Hiding them makes the real clusters easier to see. A small gap between cells can also help the eye distinguish one bucket from another.

3. Industrial applications

Current vs. temperature: is the relationship normal?

Motor current and winding temperature have a natural relationship: higher current usually means higher temperature. Set current on the X-axis with 5 A buckets and temperature on the Y-axis with 3 °C buckets, then load a week of data.

Under normal conditions, the dark region should run along the diagonal. If the entire dark region shifts upward (medium current, unexpectedly high temperature), cooling efficiency may be declining. If the dark regions become scattered and the diagonal breaks down, the relationship between current and temperature may be weakening.

This kind of two-dimensional relationship is hard to judge from a Line Chart. A Heatmap compresses it into a single density map, making it much easier to see whether the relationship still holds.

Spotting the hot motor in a fleet

Suppose 20 motors of the same model each report a winding temperature Tag. Set temperature on the X-axis with 3°C buckets and device name on the Y-axis. Each motor becomes a row. Color intensity within each row shows how many samples that motor recorded in each temperature range.

Normal motors cluster on the left, in the cooler range. A motor running hot clusters on the right. You can spot the problem by scanning across the rows. Even if no alarm Threshold has been triggered, the side-by-side comparison makes the difference obvious: one motor is running hotter than its peers.

When do alarms actually fire?

Now consider hundreds of devices, thousands of alarm records, and six months of data. Set the 24 hours of the day on the X-axis and the seven days of the week on the Y-axis. Color intensity shows the total alarm count for each time slot.

Two dark clusters stand out: Monday from 8 to 10 a.m. and Saturday from 3 to 5 a.m. The Monday cluster corresponds to the weekly startup surge, when equipment comes back online after a weekend shutdown. The Saturday cluster falls during the graveyard shift, when staffing is thinnest.

The operations team adjusted its inspection schedule around those windows. Two months later, a new Heatmap showed that the dark areas had become visibly lighter.

4. A chart hiding in plain sight

Heatmaps are not new. Bioinformatics uses them to visualize gene expression. Meteorology uses them to map temperature anomalies across space and time. Website operators use them to see where users click.

Their absence from the default chart menu in industrial monitoring software is not a technical limitation. Until recently, distribution analysis simply had not been a common requirement in many industrial monitoring workflows.

When data volumes are small, an experienced engineer can often infer the distribution by looking at the raw trends. But as the number of Tags grows and sampling frequency increases, intuition quickly reaches its limit. A Heatmap can find the densest operating range across millions of samples in seconds.

The Heatmap brings a proven distribution analysis method into industrial monitoring. It does not rely on complex math, just two-dimensional bucketed statistics. It gives users one more way to look at their data. Sometimes that extra view is enough to reveal a pattern that was already there.

The Heatmap feature has been officially available in the Visualization panel since TDengine Historian v1.0.19. For more information, visit idmpdocs.tdengine.com.

FAQs