0% found this document useful (0 votes)
257 views33 pages

SQC in The Modern Pi System: © 2021 Osisoft, LLC

This document discusses SQC (Statistical Quality Control) analysis in the modern OSIsoft PI System. It introduces the panelists and provides an agenda for the discussion. It then gives an overview of how SQC analysis can be built server-side in PI using tools like PI Vision, PI AF, and event frames. It discusses considerations for choosing source data for SQC analysis from PI tags and outlines common elements of control charting like determining the centerline and control limits. It also touches on approaches for batch processes.

Uploaded by

wadad
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
257 views33 pages

SQC in The Modern Pi System: © 2021 Osisoft, LLC

This document discusses SQC (Statistical Quality Control) analysis in the modern OSIsoft PI System. It introduces the panelists and provides an agenda for the discussion. It then gives an overview of how SQC analysis can be built server-side in PI using tools like PI Vision, PI AF, and event frames. It discusses considerations for choosing source data for SQC analysis from PI tags and outlines common elements of control charting like determining the centerline and control limits. It also touches on approaches for batch processes.

Uploaded by

wadad
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 33

SQC in the modern

PI System

© 2021
2020 OSIsoft, LLC 1
Panelists
Panelists

Rohan Vettiankal is the Product Marketing Martin Bryant has been working with PI Tadeas Marciniak is a Staff Pre-Sales
Manager for Visualization at OSIsoft. He is based since 1990 and closed his private consulting Engineer working from the OSIsoft Czech
in San Leandro, CA and has worked at the practice helping customers with SQC/SPC, office since 2009.
company since 2018. networks, and PI in 2003 to join OSIsoft.
Martin spent the first five years of his work His experience with PI dates back to the
Starting off as a Product Support Engineer, at OSIsoft doing very technical PI software mid-nineties when he worked for various
Rohan transitioned into the Product Marketing installations and migrations. OSIsoft partners on many PI projects in
Manager role after realizing his passion for different industries.
teaching and creating content to help highlight the In 2008 Martin began to focus exclusively
value OSIsoft’s products can bring. on customer value - including teaching,and He graduated at the Technical University
coaching. He was the primary creator of the (VŠB), Ostrava, Czech Republic.
Prior to his career at OSIsoft, he was working as "PI AF Jumpstart workshop" and some of
an Environment Engineer in his hometown of his learning curriculum work effects every
Chicago, where he performed environmental public class OSIsoft teaches and several
analysis at industrial facilities across the country. internal ones.
He started this role in 2015 after graduating from
the University of Illinois at Urbana-Champaign Martin lives outside of Philadelphia in the
with a bachelor's degree in Chemical Engineering. United States.

© 2021 OSIsoft, LLC 2


SQC in the modern PI System

Agenda

Introductions &
Getting Started

Demo 1: PI Vision Control Chart & Histogram, PI AF Template

Discussion: Your data, Your control limits, Your SQC Calculations


Demo 2: More options in PI Vision

Discussion: Western Electric Rules, Event Frames, logging excursions


Demo 3: Event Frames in PI Vision, Reason Codes

Discussion: Past & present of SQC in the PI Suite


• PI Data in 3rd party SQC tools & partner symbols in PI Vision
• The Services workshop with a focus on SQC/Centerlining

Take-aways
Q&A

© 2021 OSIsoft, LLC 3


SQC / SPC is about reducing variability in processes and achieving a target
level of performance with consistency.
Awareness of variability / consistency is often achieved through the use of several common
statistical charting tools that predate digital automation…

Control charts Histograms X/Y Scatter charts Other Tools


Compare process values to Give us a quick look at the Help us examine Provide other
an assigned or statistically distribution of values correlation in data perspectives on
calculated target, upper & through their range for a variability
lower control values given time

4 LLC
© 2021 OSIsoft, 4
The modern PI System

PI Data Archive PI AF / Analysis PI Vision

Capture real time data Built for decision support Accessible when & where you
need it
From the plant floor, field Dive deep into your data Data can be accessed from work
instruments, or IIoT devices with the contextual and or home, your phone, or the cloud
analytical power of and can be used for integration
Stored with the reliability PI Asset Framework & and advanced analysis
and performance of the Analysis
world’s most trusted
process streaming
database
5

© 2021 OSIsoft, LLC 5


Why build SQC analysis server
side?

Server discipline
and vigilance

Power and flexibility


Server-based calculations and event captures are
Increased visibility performed in real time – all the time – and are stored in
PI event frames and PI tags for quick retrieval

© 2021 OSIsoft, LLC 6


Why build SQC analysis server
side?

Server discipline
and vigilance
PI AF & Analysis allows you to structure calculations the
Power and flexibility way you want them - Set your control limits, calculate
them, or dynamically look them up from a 3rd party
database as they change
Increased visibility
Exploit years of history with backfilling and use
templates to apply your rule sets to easily capture a large
number of process values

© 2021 OSIsoft, LLC 7


Why build SQC analysis server
side?

Server discipline
and vigilance
Get access to SQC charts and data when and where you
need it with PI Vision: your mobile device, I-frames in
Power and flexibility corporate dashboards, at home on your laptop

Increased visibility Get notified wherever you are with email or text with PI
Notifications

Bring data & event frames into PI Datalink or to other


data analysis tools – like BI Tools or cloud-based AI

© 2021 OSIsoft, LLC 8


SQC source data

Can I use “raw” data from


PI?
There are several issues here:
• Quality of the process data
(should we filter for anomalies?)
• Frequency of the process data
(too much, too little data?)
• PI exception filtering & data
compression filtering

© 2021 OSIsoft, LLC 10


SQC source data

What else should I take into account?


Frequency
Integration of real-time data with data that is collected infrequently, such as lab samples for quality
analysis

Downtime
This will influence calculated means and standard deviations. If you have frequent or long downtimes
– consider filtering downtimes from your calculations

Averaging data
This removes variability – which is what we are studying. We don’t want to use averaged data or rolling
averages as the source data for SQC analysis.

Population studies
Some types of analyses are “population studies” (histograms). PI compressed data can be less
appropriate for these, but you can either use sampled data for this or better, use time in range for a
more appropriate calculation.

© 2021 OSIsoft, LLC 11


Centerlines + UCL & LCL

Control Charting
This is often the most useful type of real-time monitoring and is often called “centerlining” and
considered the heart of SQC/SPC

To do this we need several values:


• A sample or process value – often a fairly “raw” value

• A centerline

• Upper and Lower Control limits (UCL & LCL), which establish whether values are “out of
control” and set a reasonable operational range

© 2021 OSIsoft, LLC 12


Centerlines + UCL & LCL

How to determine the centerline + UCL & LCL


These can be determined several ways:

• Calculated (mean & standard deviation multiples) on a rolling basis or for


specific periods – like batch/product runs.

• Specifications or percentage of specifications, including the specifications


for a given product, grade, or batch (often looked up to a table)

• Choice of operations management and process analysis based on


consideration of both of the above and other factors and stored in an
attribute or looked up from a table

• Read from the process through the interface or connector or from a


process relational database via the PI RDBMS interface or PI AF table
lookup

© 2021 OSIsoft, LLC 13


Centerlines + UCL & LCL

Batch Processes

If you have a process where the


grade or product changes in
batches or grade runs, then SQC
analysis becomes a component of
a batch analysis and you should
probably contain all SQC analysis
within the batch

© 2021 OSIsoft, LLC 14


AF Templates and Analysis
Having SQC calculations based in AF provides many advantages:

• Utilizing PI AF templates, you can standardize your choice of methodology of SQC


analysis and apply it consistently to many process variables

• PI Analysis will allow you to calculate in real-time and store valuable data reflecting the
control limits, means, standard deviations, and excursions. You can also leverage PI Analysis
to backfill your calculations for any historical PI data.

• Your SQC data is not just part of a display, but part of your plant’s data history – ready for
comparison – between like systems, between similar periods of time (like batch or product runs,
and with different tools – include PI Datalink, multi-variate analysis tools, and even AI and
machine learning tools

© 2021 OSIsoft, LLC 15


Other SQC charts
CPK “capability” index
CPK measures the ability of the process to operate within control limits. It is generally appropriate
when those control limits are based on specification or other operational targets (and not
statistically)

Histogram
The histogram is a chart that shows the distribution of data values by counting the number within
subranges – usually one standard deviation each. This is often a snapshot chart for a given recent
time period and may be difficult to track over time. Keeping the data and presenting this in
alternative ways – or better yet, calculating and presenting the time that the process operated in
the selected subranges can be more interesting.

More…
If your SQC program interests extend to other measures of variability, potentially including: “P”
charts (tracking rejects), “R” charts (looking at the variability of the variability), “envelope” charts
(simple view of variability) at OSIsoft we’ve built the calculations and displays for all of these.

© 2021 OSIsoft, LLC 16


The Western Electric rules
are decision rules in statistical process control for detecting out-of-control or non-
random conditions on control charts.

They allow for more narrow control limits, because not every exception is seen as a
problem, but rather patterns of data that indicate problem variability. Some patterns
can even detect control situations, even though the limits are not yet violated (trending
patterns, all on one side… ).
20 Out of Spec High Out of Spec Low 5 Consecutive Decreasing
20 20
15 15

10 10
10
5
5 0
0 1 2 3 4 5 6 7 8 9 10
0 1 2 3 4 5 6 7 8 9 10 Target USL
1 2 3 4 5 6 7 8 9 10
LSL 5 decreasing
Target USL LSL OOS High Target USL LSL OOS Low

5 Consecutive Increasing 8 Consecutive Below Target


8 Consecutive Above Target
20 20 20
15 15 15
10 10 10
5 5
5
0 0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 0
1 2 3 4 5 6 7 8 9 10
Target USL LSL 5 increasing Target USL LSL 8 above target
Target USL LSL 8 below target

© 2021 OSIsoft, LLC 19


Configuring Western Electric Rules

PI AF’s implementation of the Western Electric Rules


Select the data

Choose which rules to apply

Choose the # of exceptions in rolling


samples of a size you choose

Like other parts of PI AF Analysis


evaluate test examples while you are
building…

Results are Event Frames with


captured data by template or
string PI Tag with pattern…

© 2021 OSIsoft, LLC 20


Configuring Western Electric Rules

Outside Captures events when the minimum number of values in a rolling sample size are
outside of the specified control limits.
of If you set this for 1 of 1 – every out of control value will be flagged, and will stay
flagged, for as long as the process is outside of control. This is the most commonly
Control used and most important test and recommended in most cases.

© 2021 OSIsoft, LLC 21


Configuring Western Electric Rules

“Two Sigma” may not be what you are thinking…


For this purpose: “sigma” is one sixth of the control range (Upper Control Limit – Lower Control Limit as
provided). Sigma is only related to standard deviation if your specified upper and lower control limits describe a
range of six standard deviations (which you can certainly calculate with PI).

A value is flagged as “outside two sigma” when the selected number of process values within a rolling
Outside of sample are greater than

two sigma Centerline +/ ( ( (Upper Control Limit – Lower Control Limit) / 6 ) * 2 )


or less than centerline – 2/6th of the control range.
In data samples that have normal distribution 95% of data falls within two standard deviations of the
mean

A value is flagged as “outside one sigma” when the selected number of process values within a rolling
Outside of sample are greater than

one sigma Centerline +/ ( ( (Upper Control Limit – Lower Control Limit) / 6 ) )


or less than centerline – 1/6th of the control range.
In data samples that have normal distribution 67% of data falls within two standard deviations of the
mean (one on either side).

© 2021 OSIsoft, LLC 22


Configuring Western Electric Rules

One Side of The process value is flagged when the specified number or more within a rolling
sample are all on one side of the centerline. This indicates a sagging or rising
Centerline process that may not have yet crossed a control limit and will alert frequently unless
processes are very well distributed around the centerline or large test value and
sample are requested. This test will very often ALSO be true if other tests are true
and if you are not careful this will replace those other (perhaps more important or
informative) tests unnecessarily.

Trend The process value is flagged when the specified number or more of consecutive
values are consistently value-to-value rising or falling. This indicates a sagging or
rising process that has not yet crossed a control limit

Stratification Stratification is counting the number of values within a rolling sample that are within
one sigma of the centerline (+/- 1/6th of control range). This test is generally looking
for a process that is not yielding the expected variability, which could indicate an
instrumentation (signal) problem more likely than a process problem.

Mixture Mixture is counting the number of values within a rolling sample that are outside of
one sigma of the centerline ( +/- 1/6th of control range) and includes values on both
sides of the centerline. This is generally the opposite of stratification and it
indicates a process that is more dynamic to the extent it is not producing values near
the centerline.
© 2021 OSIsoft, LLC 23
Configuring Western Electric Rules

Configuration
Recommendations

With so many options it’s difficult 1. Limit the number of tests you use (1-3 test should be sufficient for
to know what would work best most cases)
for your process. 2. Set your control limits consciously
That’s why we’ve come up with 3. Always use the “Outside Control Limit” test
a list of recommendations to
4. Refrain from using the outside one / two sigma tests
help guide you when configuring
these rules. 5. If appropriate, set up one or more of the “advisory” tests: “over
centerline”, “trend”, “mixture” or “stratification” and track those in a
separate rule

© 2021 OSIsoft, LLC 24


Configuring Western Electric Rules

The Power of Event Frames with SQC

PI Event Frames allow you to track each WE Rule Event as


Recommended tag setup a record. This data can then be used to better understand the
for Event Frames SQC event.
Write the results of your Once the SQC Event Frame data is captured, you can:
Western Electric rule analyses • View records as a list or in detail in PI Vision
to an attribute (string or enum) • Easily export the events (to PI Datalink or other analytical tools)
that is then written to a PI Tag
(string or digital). PI Vision also allows you to interact with these events by:
In a separate analysis, trigger • Acknowledging them
event frames from the attributes • Setting customized reason codes for them
above and capture the • Exchanging chat-like messages about individual events.
appropriate data in the template.
In addition to all of this you can also send an e-mail or text
notification based on the Event Frame and even use the
event frame records to create a "log" record of your SQC
excursions

© 2021 OSIsoft, LLC 25


PI Processbook SQC support
As has been previously announced, OSIsoft will be discontinuing some types of support for PI Processbook.
For thirty years, PI Processbook has offered an optional add-in which provided control chart, histogram,
& western electric rule support for PI Processbook display files.

Many PI Processbook displays can be migrated with a migration utility to PI Vision, however there are no
plans at this time to support displays built with the SQC add-in with full migration as:
• There is no "canned" control chart & histogram in PI Vision and no current plans to build them
• Western Electric rules and other calculations are supported in PI AF/PI Analysis not in the client display

PI AF, PI Analysis (with WE Rules & Event Frames), and PI Vision may not offer a one tool SQC Wizard
implementation, but the combination offers many advantages over the locally installed SQC wizard
• more flexibility and options to get the calculations and fit the SQC program the way you want
• sever discipline and vigilance and templates that can be flexibility applied to other process values
• SQC data & not just displays – backfill the calculations, export the data to PI Datalink or other analytical
tools
• more visibility through PI Vision & more interactivity with dynamic EF lists with reason codes & comments

© 2021 OSIsoft, LLC 28


PI System Jumpstart Workshop The workshop aims to rapidly transfer the skills and best practices
necessary to develop and implement the desired PI System configuration,
while focusing on a particular business use case or OSIsoft technology.

 The workshop scope is limited to a single business use case


 OSIsoft assistance and guidance is provided to create sample PI System
configurations for the identified use case. It is the customer’s responsibility to fully
complete and deploy the solution.

Virtual Agenda Participants: (maximum 8)


1. Introduction and Expectations OSIsoft facilitator(s) / PI AF experts
2. Create Templates
PI System Administrators
29 & IT Resources
3. Build Assets
4. Create Analytics
Use Case Subject Matter Experts
5. Create Visualization Plant / Project / Manager or Sr. Engineer
6. Review Work and Discuss Next Steps
© 2021 OSIsoft, LLC 29
Want to start implementing SQC at your organization?
PI Jumpstart workshop for SQC
Quick start the implementation of SQC / Centerlining in your environment with the help of our
experts

1-3 Days* By the end of this course, you will be able to:
 Employ prepared PI AF templates with control limit retrievals,
and key calculations for a standardized approach and work
Intermediate efficiency

 Employ only standard PI AF & PI Vision components with no


Flexible remote custom programming or custom PI Vision symbols
sessions
 Employ the PI AF Analytics implementation of the Western
Electric Rules for SQC SPC and PI Event Frames for customized
Administrators 24 x 7 monitoring of excursions

*Currently workshops are being conducted remotely in 2 to 3 hour sessions. A day equivalent is generally 2 or 3 of these sessions.

Contact your Account Manager or Customer Success Manager for more


information!

© 2021 OSIsoft, LLC 31


OSIsoft Partners
If you'd still like a more customized SQC control chart, histogram, or other display
symbol – ask about PI Partners that can develop what you want for you with PI Vision
extensibility custom symbols

If you'd like more help in migrating your existing PI Processbook SQC displays or
deploying new SQC Displays than our workshop implies PI Partners can also help
with those services

© 2021 OSIsoft, LLC 32


If you want to use another charting tool with PI Data…

Export Data on Query / Demand using our ODBC Driver and PI SQL tools

AFSDK
Application

3.
PI OLEDB
Application x
Enterprise

Application
Get the 2018 Update –
Higher performance
HTTP/
HTTP/
NetTcp
NetTcp

Application
PI ODBC
or JDBC PI SQL DAS
PI SQL DAS
(RTQP) Simpler Queries
RTQP!
Driver (OLEDB)
PI SQL Application
Client
Client
Client or PI SQL Data Access AF Server Client
Server

OSIsoft can also coach in the export of PI Data via our PI Integrator family or family of SQL Query tools to export data to the
Third Party SQC/SPC application of your choice
© 2021 OSIsoft, LLC 33
If you want to use another charting tool with PI Data…
Continuously update your SQC/SPC Analysis with a PI Integrator
to a relational database or cloud-based “big data” analysis database…
PI Integrator for Business Analytics • PI AF is the schema
PI Integrator for • SQL Server builder
Business Analytics • Hadoop (HIVE & HDFS)
• Oracle RDBMS • All configuration – no
• File coding
• ODBC client
• Performance &
reliability even with
PI Integrator for SAP HANA
• SAP HANA high volumes
• Smart Data Integration (SDI)
PI System PI Integrators • Data filters to help
Time, Event Tabular “scrub” data
and Asset Context
Context PI Integrator for Microsoft Azure • Scheduled updates or
• SQL Data Warehouse
• Azure SQL stream-to-stream
• Azure Data Lake

OSIsoft can also coach in the export of PI Data via our PI Integrator family or family of SQL Query
tools to export data to the Third Party SQC/SPC application of your choice
© 2021 OSIsoft, LLC 34
Why build SQC analysis server
side?

Server discipline
and vigilance

Power and flexibility


Server-based calculations and event captures are
Increased visibility performed in real time – all the time – and are stored in
PI event frames and PI tags for quick retrieval

© 2021 OSIsoft, LLC 35


Why build SQC analysis server
side?

Server discipline
and vigilance
PI AF & Analysis allows you to structure calculations the
Power and flexibility way you want them - Set your control limits, calculate
them, or dynamically look them up from a 3rd party
database as they change
Increased visibility
Exploit years of history with backfilling and use
templates to apply your rule sets to easily capture a large
number of process values

© 2021 OSIsoft, LLC 36


Why build SQC analysis server
side?

Server discipline
and vigilance
Get access to SQC charts and data when and where you
need it with PI Vision: your mobile device, I-frames in
Power and flexibility corporate dashboards, at home on your laptop

Increased visibility Get notified wherever you are with email or text with PI
Notifications

Bring data & event frames into PI Datalink or to other


data analysis tools – like BI Tools or cloud-based AI

© 2021 OSIsoft, LLC 37


and remember the PI Jumpstart Workshop for SQC
• 1 to 3 day-equivalents remotely (often over two to three weeks)
• a handful of your PI Knowledgeable and/or SQC knowledgeable people working with our experts
• prepared templates ready to jumpstart the effort
• learn how to customize & deploy and then complete the project yourself

Contact your Account Manager or Customer Success Manager

© 2020 OSIsoft, LLC 38


Thank you

Q&A
© 2021 OSIsoft, LLC 39

You might also like