Analytics & Visualization · PI Vision Custom Symbol

How to Add a Histogram to AVEVA PI Vision

Everything you need to get the PIForge Histogram symbol running on your AVEVA PI Vision server — installation, license activation, and a tour of every feature with live video.

What this solves

Installation — 5 steps, about 10 minutes

1. 1

   Download the package

   Purchase the symbol at piforge.app and download the ZIP from your profile. It contains four files: the symbol script (.js), two templates (.html) and an icon (.svg).

2. 2

   Copy the files to the PI Vision server

   Extract all four files into the PI Vision extension folder on the server: %PIHOME64%\PIVision\Scripts\app\editor\symbols\ext\ (create the ext folder if it does not exist yet). Use %PIHOME64%, not %PIHOME% — on a PI server those are different trees and PI Vision is only in the 64-bit one.

3. 3

   Restart the PI Vision application

   Recycle the PI Vision application pool in IIS (or run iisreset), then hard-refresh the browser (Ctrl+F5) so the new symbol script is loaded.

4. 4

   Add the symbol to a display

   Open a display in edit mode. The new symbol appears in the symbol palette on the left. Select it, then drag a tag or AF attribute onto the canvas.

5. 5

   Activate the license

   Open Format Symbol, paste the license key from your purchase email into the License field, and save. The watermark disappears immediately — no restart needed.

What you can do with it

Overview

A Trend Cannot Tell You This

Distribution analysis on the data you already historise

A trend shows where a value went. It does not show how the values are spread, where they cluster, or whether the process can hold its spec. Drop one attribute and Histogram builds the distribution straight from the archive.

• Count, mean, standard deviation and bin width computed in the display
• Numeric or categorical data — strings, booleans and enumerations included
• Works on any tag or AF attribute you already collect
• No server component, no scripting

Binning

Bin Width Decides What You See

By hand, or by the rule your statisticians expect

Too few bins hide structure; too many turn the chart into noise. Choose the width yourself, or let an established rule pick it from the data.

• Sturges, Scott and Freedman–Diaconis rules
• Or fix the bin count and range manually
• Count, percent or density on the Y axis
• Exact counts printed on every bar

Statistical Overlays

The Shape, Against the Ideal

Normal curve, central tendency and spread

Overlay the normal distribution the data would follow if it were well behaved, then compare. Where the bars pull away from the curve is where the process is telling you something.

• Normal (bell) curve fitted to the data
• Mean, median and standard-deviation bands
• Configurable percentile markers
• Multistate colour rules and labelled limit lines

Process Capability

Can This Process Hold the Spec?

Cp, Cpk, Pp, Ppk and Cpm against your limits

Enter the lower and upper spec limits and Histogram returns the numbers a quality review actually asks for — short-term and long-term capability side by side, with the loss rate they imply.

• Short-term Cp / Cpk from within-subgroup variation
• Long-term Pp / Ppk from overall variation, plus Cpm against target
• Sigma level and DPMO
• Out-of-spec: what actually happened vs what the normal model predicts
• Verdict badge: capable at Cpk ≥ 1.33

Mixed Populations

When Cpk Is a Lie

It detects the case that makes capability meaningless

A capable-looking Cpk on two blended populations is worthless — and it is the most common way capability studies go wrong. Histogram fits a two-component Gaussian mixture to the data and tells you when one tag is really two operating states: two grades, two shifts, two units.

• Two-component mixture fitted by expectation-maximisation
• Bimodality coefficient plus separation and weight tests before it will claim anything
• Each population drawn with its own curve, mean and share
• States plainly that a pooled Cpk on two populations is invalid
• This is the check that stops a bad number reaching a report

Normality Check

Capability Assumes Normal Data

It says so when the assumption fails

Cpk on skewed or heavy-tailed data understates the real reject rate. Histogram measures skew and kurtosis and warns you on the same panel that shows the number.

• Skew and kurtosis evaluated on every capability run
• On-screen warning when Cpk should not be trusted
• Actual vs predicted out-of-spec exposes the gap directly
• You still get the number — with its caveat attached

Pareto Mode

Fix the Vital Few

Reject codes, alarm causes, downtime reasons

Point it at a text tag and it switches to categories on its own. Bars are sorted by frequency, a cumulative curve is overlaid, and the vital few that account for most of the losses are highlighted.

• Automatic categorical detection for string, boolean and enumeration tags
• Sorted bars with cumulative % curve and 80% line
• Vital few highlighted, trivial many dimmed
• Top-N cap for long code lists
• The classic tool for alarm, defect and downtime rationalisation

Styling & Export

Yours to Match, and to Hand Over

Site styling plus CSV and JSON export

Bars, background, grid and type follow your display standard. And every number on the chart leaves with you.

• Bar, background and grid colours; font size
• Colour rules and labelled limit lines
• CSV export of every bin or category with counts and percentages
• JSON export adds the full statistics and the entire capability result
• Audit-ready evidence for a quality review, in one click

Frequently asked questions

Does it check for normality before reporting Cpk?

Yes, and it reports the result. Cp and Cpk assume a normal distribution, so a capability number without that check can be confidently wrong.

Where do the specification limits come from?

They are configured on the symbol, so the same tag can be assessed against whatever specification applies to the product currently being made.

Does Histogram work with my PI Vision version?

Yes — PI Vision 2019 R2 through 2023+ are supported. The symbol uses the standard PI Vision extensibility model, so regular version upgrades do not break it, and one year of compatibility updates is included.

Do I need to install anything on client PCs?

No. The symbol is installed once on the PI Vision server; every browser client gets it automatically from there.

How is it licensed?

One license covers one PI Vision server, with unlimited displays and unlimited users on that server. The key is delivered instantly by email after purchase.

Can I try it before buying?

Every feature is demonstrated in live video on the product page, and a free symbol (Combo Chart Free Edition) is available on GitHub to validate the install workflow end to end. If you need a formal quotation or have to go through procurement, use the Request a Quote button on the product page.

Other Analytics & Visualization symbols

All PI Vision symbol guides · Browse the marketplace