Histogram+

Overview

When to use it

Use Histogram+ when you need to:

• See the spread and the most common range of a process value.
• Compare how closely a distribution follows a normal curve, with the mean, median, standard deviation lines and quantiles.
• Judge whether a process fits its specification limits (Cp, Cpk, Pp, Ppk).
• Find the few states of a digital or text value that account for most events.
• Compare the distributions of several data sources on the same bars.

Choose Control Chart+ when you need to know whether the process is stable over time, and Combo Chart+ to follow the values themselves over time.

Before you begin

• Prepare numeric PI Points or AF attributes with recorded values in the display time range. A digital state, boolean or text source gives a count per state instead of bars.
• Choose a time range that holds enough values for the bars to mean something. A few dozen values give sparse bars.
• For capability against AF limits, the attribute needs the limit traits HiHi, LoLo and Target (or other limit traits you choose). Otherwise enter the limits by hand.

Add and set up the symbol

1. Select Histogram+ in the PI Vision symbol gallery.
2. Drag a numeric PI Point or attribute onto the display.
3. Set the display time range, for example the last 8 hours.
4. Right-click the symbol and select Configure Histogram.
5. Open the Axis X section and set Number of Bars to 20.
6. Close the pane. The symbol shows one bar per range with its count, and the mean and standard deviation under the title.

Common tasks

Show the distribution against a normal curve

Use this to see how far a process value departs from a normal distribution.

1. Right-click the symbol and select Configure Histogram.
2. Open Trace Options and expand Statistics.
3. Set Bell Curve (1) to ON.
4. Set Mean (2) to ON.
5. Set Standard Deviation (3) to 2.

The chart draws a normal curve with the same mean and standard deviation as your data, a line at the mean, and lines at 1 and 2 standard deviations on each side. The closer the bars follow the curve, the closer the values are to normal. Here the bars are flat on top, so the tank level is spread more evenly than a normal curve.

To mark other points, switch Median on, or select + under Quantiles and enter a fraction between 0 and 1. A quantile of 0.95 draws a line where 95 % of the values lie below it.

Choose the number of bars

Use this when the bars are too coarse to show the shape, or so fine that they look ragged.

1. Right-click the symbol and select Configure Histogram.
2. Open Axis X.
3. In Binning rule (1), choose Manual to set the bars yourself, or a rule that calculates them from the data.
4. With Manual, enter the count in Number of Bars (2), from 1 to 100.

With a rule, Number of Bars is hidden. On about 5,800 values over 8 hours, the rules drew roughly 15 (Sturges), 29 (Scott) and 37 (Freedman-Diaconis) bars.

To fix the range the bars cover, set Scale Range to Custom in the same section.

Check capability against specification limits

Use this to find out whether a numeric process value stays inside its specification. It works per data source, for numeric data in Histogram mode.

1. Right-click the symbol and select Configure Histogram.
2. Open Trace Options, select the data source, and expand Statistics.
3. Set Specification limits (1) to Fixed values.
4. Enter the upper limit (2), the target (3) and the lower limit (4), in the unit the chart shows.
5. Keep Capability readout (5) and Specification lines (6) ON.

Dashed lines mark the lower limit, the target and the upper limit. A line above the plot gives Cpk and Ppk, each colored by its verdict: 1.33 or more is capable (green), 1.00 to 1.33 marginal (amber), below 1.00 not capable (red).

Hover the readout to see every index. Cpk compares the limits with the short-term variation between neighboring values, Ppk with the overall spread. In the example Cpk is 6.99 and Ppk 0.60: the level is steady from one moment to the next but wanders over the whole range, so it does not stay inside the limits.

Two warnings can follow the readout: not normal and possibly two populations. The indices assume a normal distribution, so treat a flagged result with care. For an AF attribute, choose AF limit traits in step 3 to read the limits from its HiHi, LoLo and Target traits. While capability is on, the mean and standard deviation on the chart count every value once, so they can differ from the figures without capability.

Rank the states of a digital value as a Pareto chart

Use this to find the few states, such as stop reasons or machine modes, that account for most events. The data source must be a digital state, boolean or text value, and the only data source on the symbol.

1. Right-click the symbol and select Configure Histogram.
2. Under General, set Chart Mode (1) to Pareto.
3. Enter a number in Top N (0 = all) (2) to keep only the largest categories and merge the rest into Other. Enter 0 to show every category.
4. Keep 80 % line (3) ON to draw the 80 % reference line.

The bars run from the most frequent state to the least frequent, and a line shows the cumulative percentage on the right axis. The states up to the one where the line reaches 80 % are drawn solid, the rest dimmed. Other is always last.

To rank by how long each state lasted instead of how often it occurred, open Data Retrieval and set Basis to Time Weighted. This suits downtime reasons. Basis is available with the Recorded mode.

Compare data sources on the same scale

Use this to compare how several data sources are distributed, even when they have different numbers of values.

1. Right-click the symbol and select Configure Histogram.
2. Under General, set Stacked (1) to OFF, so each data source gets its own bar in every range.
3. Open the Axis Y section, expand Value Scales, and set Y value (1) to Percent.
4. Open Trace Options, select each data source, and set its Color and Label.

Each data source shows its share of its own values (the unit in the title belongs to the first data source), so the bars of one data source add up to 100 % whatever its number of values. The legend names the data sources.

With Stacked ON, the data sources pile up in one bar, which shows the combined count. Choose Density in step 3 to compare the bars with a normal curve.

Use the symbol

• Hover a bar to see its range and count. In a Pareto chart the tooltip adds the share and the cumulative percentage.
• Fullscreen opens the chart in a large dialog. Download saves a picture (PNG, SVG, PDF) or the binned data (CSV, JSON, XLSX).
• If the toolbar is not shown, set Toolbar > Show to ON.
• Time range: the bars follow the PI Vision display time range. Histogram+ also supports the Time Bar for its own range.
• No data: the chart shows empty axes. Check the time range and the data source.
• The symbol only displays data. It does not write to the PI System or change other symbols.

Configure the symbol

Right-click the symbol and select Configure Histogram. Sections appear in this order.

Trend Options

Axis X

Axis Y

Multi-state section

Available for numeric data. Sets how multi-state limits added to the symbol are drawn.

Data Retrieval

Trace Options

Select a data source in the list first. The settings below apply to that data source.

Capability needs at least 10 values. Cp and Pp need both limits. An index that cannot be calculated shows its reason. A limit trait that changes during the time range withholds the capability, because no single limit applies.

Time Bar, Export Data and Numbers

Troubleshooting and limitations

Limitation: Neither axis offers a logarithmic scale.

Limitation: In Recorded mode the median, the quantiles and Freedman-Diaconis are estimated from 100 fine bins. In Interpolated mode they are exact.

Limitation: Capability counts every value once, also when Basis is Time Weighted.

Limitation: Histogram+ has no zoom. The toolbar offers Fullscreen and Download only.