Uploaded by Wenxuan Tong

2018 OSIsoft TechCon Digital Plant Template 320 Version FINAL614

2018 OSIsoft TechCon Lab
A Digital Plant Template for
Operational Insights
An Enterprise Strategy
OSIsoft, LLC
1600 Alvarado Street
San Leandro, CA 94577 USA
Tel: (01) 510-297-5800
Web: http://www.osisoft.com
© 2018 by OSIsoft, LLC. All rights reserved.
No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or
by any means, mechanical, photocopying, recording, or otherwise, without the prior written permission
of OSIsoft, LLC.
OSIsoft, the OSIsoft logo and logotype, PI Analytics, PI ProcessBook, PI DataLink, ProcessPoint, PI Asset
Framework (PI AF), IT Monitor, MCN Health Monitor, PI System, PI ActiveView, PI ACE, PI AlarmView, PI
BatchView, PI Coresight, PI Data Services, PI Event Frames, PI Manual Logger, PI ProfileView, PI
WebParts, ProTRAQ, RLINK, RtAnalytics, RtBaseline, RtPortal, RtPM, RtReports and RtWebParts are all
trademarks of OSIsoft, LLC. All other trademarks or trade names used herein are the property of their
respective owners.
Use, duplication or disclosure by the U.S. Government is subject to restrictions set forth in the OSIsoft,
LLC license agreement and as provided in DFARS 227.7202, DFARS 252.227-7013, FAR 12.212, FAR
52.227, as applicable. OSIsoft, LLC.
Published: June 14, 2018
Digital Plant Template using PI Analytics and Operational Modes for the Enterprise – Hands-on Lab –
OSIsoft Users Conference 2018
Lead: Osvaldo Bascur, Principal Academia-Industry Innovation
Lead: Michael Tippett, Academia Field Systems Engineer
Assistant: Gina Laviste, Field Systems Engineer (PI Vision and PowerBI)
Instructor: Nicolas Peels, Academia Field Systems Engineer
Instructor: Erica Trump, Academia Instructional System Designer
Instructor: Ales Soudek, Global Systems Engineer
Table of Contents
Table of Contents .......................................................................................................................................... 3
Learning Objectives and Problem Statement ............................................................................................... 5
Part I Publishing and Creating the Digital Plant AF Database ..................................................................... 10
2. Creation of all process units or elements. ...................................................................................... 15
3. Explore the Process Unit Template Attributes................................................................................ 16
4. Explore the Process Unit Clock simulator and the Table for configuring the behavior of each
process unit. ........................................................................................................................................ 18
5. Explore the Process Unit Analysis Mode expression ...................................................................... 19
6. Explore the Process Unit Analysis Event Frame Generation........................................................... 20
Process Event Frame Template............................................................................................................... 21
Configure the Event Frames Search ............................................................................................................ 23
Summary of the process of setting the Digital Plant Template .............................................................. 27
Part II. Leveraging the data and event framed data with the insightsMy PI Syst generated by PI Analytics
.................................................................................................................................................................... 28
Analysis using PowerBI ........................................................................................................................... 32
PI Vision................................................................................................................................................... 28
Part III Industrial Adaptation for the Plant Template Tips .......................................................................... 43
PI Analytics for Unit Global variable calculations ................................................................................... 43
PI Analytics to create augment performance or softsensors estimates ................................................. 44
PI Datalink for data extraction using TimeDat algorithm ....................................................................... 45
PI Analytics Linear Regression Example .................................................................................................. 47
Using Microsoft Machine Learning Studio to create empirical models.................................................. 50
Key to success in modeling your process operational data. ................................................................... 52
Appendix I LAB Summary: ........................................................................................................................... 54
Appendix II PI System software components ............................................................................................ 54
Appendix III Extraction of the PI Event Framed Data. ............................................................................... 55
Appendix IV Connecting PI Vision to your Digital Plant Example ........................................................... 60
References .................................................................................................................................................. 61
Learning Objectives and Problem Statement
In this lab, we present you with a Digital Plant Strategy. This is a standard, proven approach to overall
process management using latest technology in the OSIsoft PI System. You will learn to apply this
strategy to your operations and get started on your journey towards digital transformation.
Figure 1. From raw time series data to operational insights.
The Digital Plant Strategy enriches real-time process data with asset and time context. Context is added
to process data based on key operational events for each unit in the plant. You will also learn how to
use the contextualized data to model the operation of a process in real time. The real-time model will
allow you to predict and optimize process performance, thus transforming data into actionable
Some customers have difficulty initializing designing the database schema that represents their process
and plant assets, often adding unnecessary items that complicate and delay the realization of tangible
benefits, such as reduced production costs or improved product yields.
In this example, we discuss optimum design and use via a simple PI Asset Framework (AF) database that
represents the production process and physical plant assets to meet key business objectives such as:
 Maximizing product yield and production throughput
 Optimizing energy usage during specific production events
 Visualizing real-time displays showing KPI metrics as well as process and production constraints
 Modeling the manufacturing process and calculating real-time variation tracking of expected
(predicted) vs. actual results.
 Extracting contextualized data from production events for viewing in advanced analytic
visualization tools such as Microsoft’s PowerBI or Tableau.
 Reducing unscheduled equipment failures and production downtimes
Further, this strategy bridges the gap between production planning and execution by identifying the
production variance as it occurs in real-time. The approach also simplifies the implementation of
production monitoring and analysis, leading to continuous improvements that result in increased
production yields and reduced operating costs. This integrated approach to Overall Production
Management engages Management, Engineering, Operations, and Maintenance teams. It has been called
the Follow the Money Strategy.
Several companies in the process industry have adopted this strategy. Overall Process Effectiveness is
improved when data from the plant is aligned with process scheduling. The approach also supports
collaboration by providing a unified vocabulary for planning and operating functions. It engages all
functions of the process plants as depicted in Figure 1 on the following page. Recent advances like the
evolution of networks, data speeds, and object programming are leading to increased adoption.
Figure 2. Enterprise integrated performance management (Production Targets versus Actuals).
In this lab, we apply Digital Plant Strategy to an Oil Refinery. We create a digital plant template that
models process units and operational modes. We use streaming data along with event context to
quantify production and consumables such as electricity, fuel, water, and others.
We use technologies including PI Asset Framework (AF) asset analytic template, PI Analysis Expressions,
and PI Event Frames to model a running plant. The Oil Refinery Process Block Diagram is a guide for
configuring the digital plant. It shows each process unit along with key data streams, such as process
feed rate, electricity consumption, and water consumption. A dynamic simulator creates the process
variables and operational events. The production targets from the plant planning and scheduling are fed
automatically into the Asset Framework using the Table database access subsystem. The integration of
process scheduling and operational execution is the feature of this exercise.
Figure 3. PI Vision showing a relative display generated automatically by the Digital Plant Template
PI Vision shows the operational modes and associated unit variables to see their impact on the
production compliance versus plant schedule. All operating events are visualized in the left side of the
navigational pane for fast event frame analysis.
Figure 4. Multidimensional ways to visualize plant production targets versus actuals.
This diagram shows the Multidimensional of the operating and time context data for improved decisionmaking. PowerBI provides the tools to slice the data into many variations so this how we can view the
operational events, data for the whole process plant.
Figure 5. PI Event Insights shown using Microsoft PowerBI Template.
PowerBI displays the PI Event Framed Data for all process units showing their Overall Production
Effectiveness for the selected time by Unit and Shift. There are many ways to present the raw data
using data slicers and visualization charts. PowerBI dashboards can be pushed to Azure where you can
use Cortana to assist in analyzing the operating modes and the data.
Figure 6. Real Time Operation Intelligence showing the continuous improvement loops
Figure 4 shows an overview of the business process automated by comparing the performance targets
with the process unit execution. It shows a modular, templatized (cookie cutter) approach to building all
the elements for a process plant. The PI Analytic engine classifies the operational modes in operational
modes, which generates the event frames to aggregate the data and to estimate the minor losses. The
PI Event Frame Template is used to generate the operational insights for further analysis. The
Operational modes are:
 “Running OK”: Operating with a specified production target range,
 “Trouble”: Operating off target by 10% (lost production and the associated minor utility
costs and missed opportunities). This Trouble operating mode is critical and often
ignored. These are all the minor losses, which are often unaccounted for unless the
plant information system does it automatically.
 “Idle”: Equipment is performing ok but the unit is not feeding material.
 “Down”: Equipment has had an unscheduled shutdown, loosing production and all
utilities associated with it, and,
 “Maintenance”: the time that a process unit was in scheduled to be in maintenance
Thus, we track the time that the plant is on target, as well as all the minor and major losses. The
strategy aggregates all utilities and consumables over the selected operational time events. Further
details are described in our manuscript called a journey towards a digital transformation in the process
industries available in PI Square.
The philosophy behind finding the operating modes is described in the Figure 5. Garbage in garbage out.
The raw data is transformed using performance calculations to validate, classify the raw data for be use
for model predictions (Bascur, 1988).
Figure 7. A data framework with distributed intelligence.
Figure 7 shows the process data hierarchy required to implement artificial intelligence strategies in the
process industries. After collecting the raw data the data is checked and validated, followed by a
process unit evaluation of its for operating modes. Then, empirical process models or performance
equations calculate augmented value for process data by operational mode. Once, the process
conditions for one unit are found a coordinated effort with the following chain supply units are adjusted
to the current constraints detected by the system. The entire process line for the plant’s best operating
conditions are identified based on the inventory levels and process-equipment capabilities. The
elimination of islands by business processes and continuous improvement strategy has been discussed
by Bascur and Kennedy, 1995. It emulates the theory of constraints proposed by Goldratt, E. 2014 in his
book The Goal, a process of ongoing improvement. For additional updated details you can visit PI
Square. https://pisquare.osisoft.com/docs/DOC-2971-a-journey-towards-a-digital-transformation-inthe-process-industries-title-preface-and-outline-v3pdf
Audience: All Process people and management. Business Process Improvement advocates.
Note: Appendix contains a detailed description of the value added digital plant strategy.
Part I Publishing and Creating the Digital Plant AF Database
The business objective of this example is to track the Production Variance for every process unit while
tracking the consumption of energy, water and other consumable. The Analytics use an expression to
10 | P a g e
detect the operating mode for every unit and to generate an event when going from one state to
Figure 2 shows several business process workflows in a typical process plant(s) in an Enterprise. In this
example, we will focus on getting the process targets coming from scheduling team to be check in real
time by the operating plant. The plant reports the event framed data for each process-operating mode.
These events are captured in the PI event frame database for business analysis. The validation and
classification of the data is perform automatically. Thus, the running mode state is the key one for
adding additional calculation for predicting performance indicators at each area of the plant.
Figure 8 – Simplified Oil Refinery Process Block Diagram
11 | P a g e
Oil Refinery block diagram example
Figure 9 – Applying the Digital Plant Template to an Oil Refinery
The figure shows a typical oil refinery block diagram. The mapping of all the areas into elements are built
using the PI Unit Template example.
1.- Publish the Digital Plant Template for the Oil refinery example
Start by using your Oil Refinery Block Process Diagram to compose the naming of all the necessary
process units.
These process units will use a UNIT Template to get the basis for PI Analytics expression, event frame
generation, notification and SQC. You can add AF elements as you desire. You can also add Inventories
or tanks.
Step by Step:
a. Find the OilRefineryExample.xml file. This file contains an integrated PI AF/Analysis/Event
Frame set up using a UNIT Template and the Event Frame Template.
b. Open the PI System Explorer. From the icon bar select Database
c. Create a New Database by finding the Database Icon on the PI System Explorer main menu
12 | P a g e
d. In the name field type the name of the AF Database that you will create. (This name is very
important because it the name that we will use to link these database with many associated
system to link them together.
e. Select the OilRefineryExample database and click OK.
f. From the File Menu Select Import database
g. Search in your Desktop directory for your OilRefineryExample.xml file.
13 | P a g e
h. Import the XML File to create the PI AF database and create the UNIT Template with PI Analytics
(expressions and event eeneration), event frame template, metadata tables with parameters
and all the tags associated to the attributes. Find the OilRefinery.xml file. This file contains an
integrated PI AF/Analysis/Event Frame set up using a UNIT Template and a Event Frame
i. From the navigation pane, select elements. In the elements list find the PIServer. Change the
name of the server to your PI Server. This will valuable when you get home to clone this
example into your PI System.
j. One you have the list of all the PI AF Elements go to the Menu bar and select Search.
k. Select the Attribute Search. Search for all Attributes in the database. Press Search.
Press Ok. This will bring the list of all the attributes that need to be configure.
14 | P a g e
Select the middle panel of the Attribute Search.
Select the CLOCK TAGS and Press Right side of the mouse. Select Reset Template.
Then, Create or Update PI Point.
Do the same for all ! mark Tags. Electricity Consumption, Water Consumption, Production
Rate and Mode. These are the global variables for the whole Process Plant.
q. All attributes with tags for the example are automatically generated.
The PI analytics (expression, event frame generation and notifications) will start working and generating
the values for all process variables necessary for the dynamic simulator to start work.
2. Creation of all process units or elements.
The example uses an oil refinery process block diagram to name all the elements that are configured
using one Process Unit Template. This template contains the analytics to create the simulation, the
expression to calculate the operation mode and to generate the events triggered for each unit based on
the scheduled triggered stored in a Table.
Step by Step
a. On the icon bar select, check in. All the attributes will connect to the tags created and the
simulator will start generating the process values and executing the expression for each
element. The event frames will be generated.
15 | P a g e
3. Explore the Process Unit Template Attributes
The following diagram shows the PI AF main library with the internal databases to build the digital plant.
The Unit template includes the attributes required. The PI Analytics calculations to calculate the real
time variance for each unit. The classification of the variance to define the operational modes of the
unit (“Running”, “Trouble”, “Idle”, “Down”, and “Maintenance”). The event frame generator set up the
time intervals for aggregation of the consumable variables in the Event Frame Template for estimating
the major operating costs and production for each operating mode for all the process units in the plant.
PI Analytics can have additional performance calculations using a data validation and classification
structure as discussed in section tree. PI Analytics do linear regression estimations for on line predictive
analytics based on validated and classified raw data. It can also do second order polynomial fit to
estimate the second order derivative to assess the overloading a process unit.
The PI Event Frame Template environment enables to associate the event frame, operating and the
consumables variables to aggregate the data according the identified operational modes. The event
The PI AF Tables are used to store the parameters to execute the generic algorithm for all process units.
In this example, we have two tables. One table is the Trigger Table that stores the trigger operating
mode values to classify the operational variance. The other table contains the parameters required to
model the process feed rate for each unit showing the different operating mode.
Operational events of importance can generate Notifications to send operational insights to other
people to collaborate in resolving problems.
The figure shows the Unit Template from which all elements are created equal. This UNIT Template is
applied to each of the selected elements that you will build in your plant. With this cookie cutter
approach you can be easily adapt this template for your plant for a fast start in your digital twin building
16 | P a g e
strategy. It will enable to track the production variance based on the operating states in your plant. To
continuously improve by checking where the operational constraints are and what is the value of your
losses while the happening. (Accelerate the time to avoid constraints or ongoing improvements to
eliminate them from your plant). Be able to identify the weakest link in the process chain supply.
From the PI System Explorer navigation pane, select Library.
On the right hand side if the figure you can see the PI AF Library. It includes the Element Templates,
Event Frame Templates, Tables, Enumerations, and Categories. The Unit Template is the key ingredient
to model a process plant. It contains 5 critical variables to get started: The process feed rate, total
energy, total water, operational mode (or mode), and the triggers attribute stored in the triggers table.
Additional variables can be added susch as the specific consumable variables derived from the basic
critical variables. Many additional performance equations variables can be include in additional analysis
as we will include later the basic unit template. These additions depends on the industry that you work.
Two most important additional analysis are the energy and water consumables totalizers, which are
required at each element to populate the global variables for the unit elements.
17 | P a g e
The Figure shows the Unit Template global attributes. These attributes are categorized using a
Consumables, Metrics, Process Variables, Triggers and Operational Mode. Click Group by: Category on
the right size of the screen.
The trigger attributes are stored in the Library Tables with the values that will be used to create the
mode for every unit. We will look at the Asset Triggers Table. This table contains the targets for each
unit based on the production scheduled.
4. Explore the PI Analyses Templates. Process Unit Clock simulator function and the Table for
configuring the behavior of each process unit.
The following figure shows the CLOCK expression. The Clock is the process tag simulator for each unit.
The model is configured from the process feed attributes table. This section generates the process feed
rate, electrical consumption and water consumption tags.
From the Unit Template select the analysis templates tab and look for the CLOCK expression.
The following table shows the parameters that are used to create the dynamic simulation data and
events for the variables for each unit. If you want to repurpose this to your plant you will need to
change the names of so they are the same as the elements that you want to create.
18 | P a g e
NOTE: In a physical plant, these attributes will be available directly from the sensors in the plant.
5. Explore the Process Unit Analysis Mode expression
Step by Step
a. Select the Analyses Tab and select the expression mode. This expression calculate the
operational variance modes when off from the production target set by planning team.
This section shows the expression that calculates the operating mode by comparing the process
variables Process Feed Rate, Electricity Consumption to create the 5 operational mode for all process
19 | P a g e
units. The metadata for each process unit is stored in the Library Triggers. The trigger data is based on
the production scheduling targets desired for each unit.
Go to the PI AF Tables and select the Asset Triggers Table.
The Table shows the 4 triggers values for each unit which are used in this example to compare the actual
process variables to generate the mode values and the state used to create the event frame to classified
the states of each process unit. This example assumes that each unit can run at a 100 as a target. As
such, the trouble mode is defined to be about 10% less than the production target. Having these
metadata parameters for each unit simplifies the implementation and maintenance of the analysis
algorithm. If you want to repurpose this to your plant targets you will need to change the names of so
they are the same as the elements that you want to create.
6. Explore the Process Unit Analysis Event Frame Generation
From the Unit Template tabs choose the Analyses and select the Running mode expression.
20 | P a g e
The Running mode sets the event frame by selecting the Event Frame Generator. This setting will
enable to aggregate all utilities and consumables variables once the Event Frame Template is designed
later. These frame intervals will account for the production and operational losses while in a certain
non-profitable mode. These time event marking shows the start and end times to generate the
aggregation of the production and consumables for each operational modes for all units in the plant.
Process Event Frame Template
This Library Database called Event Frame Template enables to create an object that will process all event
frames and aggregate the process variable data according to the desire degree of analysis required
based on the Operating Modes of each process unit. This is perhaps the most sophisticated tools to
transform process data into valuable information and business notifications.
From the PI AF Library select the Event Frame Template. You will explore the Consumable Variables
Template that does all the recalculations.
21 | P a g e
Event Frame Template
OAB This needs to include the Energy Consumption aggregations in the table
This Event Frame Template organizes all the units and the variables that will be augmented by
calculating the configured derived variables based on the integration algorithms in the PI System to find
the insights for each operational mode. These are the Average, Count, Delta, Maximum, Minimum,
Range Standard deviation and Total. These process derived variables are the operational insights
captured by each operation mode for all process units.
Explore the Electricity Consumption Minimum setup. Here the settings menu provide you with a list of
all Time Derived Variables algorithm to transform data into insights for each of the operating modes
found by the master algorithm for every unit in the plant.
You can select you desired aggregation
algorithm and Time Range calculation.
Explore the Table with all the metadata associated the triggers for each process unit.
22 | P a g e
The PI reads these values as attributes for each unit as you can see in the AF list of attributes.
These values are used by the PI Analytics expression to calculate the variance and to set the operating
mode for each unit. The time of the event is stored by the event frame generation.
As you can see this is the most interesting part of the lab. You are automating how to generate
insights based on the event framed data for further analysis using PI Vision and PowerBI or other
From the PI System Explorer navigation pane, select Elements. This view shows all the elements
configured from the Unit Template. This strategy enables to augment the raw data into finding
operational insights, augmenting the attributes by incorporating performance equations, predictive
analytics estimations and event frame analysis of the operational data and events.
Click on the FCCU Unit to see all the life process values generated by the simulator.
We have Process Feed Rate, Air Consumption, Electricity Consumption, Water Consumption and Mode.
We also see the Production Triggers, which are obtain from the Table Triggers. These values are used by
the Unit Template PI Analytics expression to calculate the operating mode and to generate the event
frames to find the operational insights. The same object is reused for all process unit in the Oil Refinery.
Our next step is to catch all exceptions from production targest and the associated data for analysis
using several tools available to us.
Configure the Event Frames Search
From the PI System Explorer navigation pane, select Event Frames database. Here we can find all the
event frames which having been generated by the PI System Analytics based on the data classification
defined and we can find the aggregate data at each of the operating modes for all the process units in
23 | P a g e
the plant. This information will be visualized internally, and then we will configured PI Vision and an
PowerBI continue the analysis process using the Visualization and Analysis tools. It is important to say
that the most important data transformation occurs in PI Analytics and the Event Frame subsystems.
You will find an empty table. We need to configure it by selecting the wheel tools on the top right hand
to configure it. From the menu list select, Select Attributes. Select the attributes from Template as
shown the next diagram.
Press the gear icon on the top right to bring the variables that contains the operating insights. These are
the results of taken the real data and classifying it into operational insights for every unit in the plant.
This Event Frame Search shows you the critical dimensions for analysis using PI Vision and PowerBI.
These dimensions are Asset or Unit, Production Feed Rate, Total Energy, Total Water, and mode. You
can as many as you see fit for your case.
24 | P a g e
This diagram shows the most valuable results. The raw transformed into operational insights for further
analysis and action.
From the PI System Explorer navigation pane, select Management Tools. These tools will enable to
control the way to calculate the past history if we change the business rules on how we transform data
and events into operational insights. It is a tremendous valuable tools for process operations
Back filling the Process Variables using the Management system. First start by doing the CLOCK
Analytics. This step is not required if you are using your physical plant. The tags will be coming live from
your sensor data. It is important to running the Clock first. Then, the Mode PI Analytics and then the
Event Frame Generator. It important to note that you can change your analysis and rerun it for other
parameters to test the results and data transformation into insights.
25 | P a g e
Then back fill your modes. These modes will generate the events which will create the time derived
variables and the event frame database for further analysis by other tools.
26 | P a g e
Summary of the process of setting the Digital Plant Template
This figure summarizes the strategy. One process unit template with the production rate, production
rate target, consumables, and stream assays is subject to an analysis, which generate the operating
mode. The associated data is aggregates using the PI Unit Event Frame Template for all units. These
operational insights are then used by PI Vision and by PowerBI in this LAB to identify operational
problems and solutions. The Running Operating Data is used to develop empirical models to augment
the value of the process data using LAB data or Mechanical data to generate softsensors to estimate
unmeasured variables to be able to push to the operating constraints.
27 | P a g e
The Process Unit Template is applied to the process block diagram of the plant. The data is aligned with
the assets using global variables for all units simplifying the analysis and optimization of the plant. The
PI AF enables to integrate the process stream data with other tabular data to augment the value of the
operating data using performance equations and aggregating the data at the right resolution defined
automatically by the event frame subsystem. The event-framed data provides the operational insights
to analyze and to optimize the process plant. These events can generate process notifications to send
emails to other colleagues or groups to be aware of the changes in operating modes to trigger a work
order or process improvement cycle to improve equipment reliability or to improve production
scheduling. The product tank inventories should also be included in practices. This exercise is left to the
student to implement. A basic tank inventory is left for exploration. We also will leave the PI
Notifications configuration to later.
New business process workflows are started at the sensor level enabling to reduce losses and improve
reliability analysis.
Part II. Leveraging the data and event framed data with the insights My
PI System generated by PI Analytics
We will share to ways of analyzing the data and events frames created by the Digital Plant Template
example. We will share an example using PowerBI and secondly we share the PI Vision strategy.
PI Vision
The section describes the use of PI Vision with the Digital Plant Template.
With PI Vision you can analyze data in multiple ways, seeing your data on any device, wherever you are,
whenever you need it. With its self-configuring display objects, PI Vision makes it intuitive to instantly
start working with your data, leveraging it in new ways, for data-driven decisions and new insights into
operations and business.
Please open PI Vision from the icon on your desktop.
We will generate a PI AF element relative display so we can look in detail at all the data in each of the
process unit in the Oil Refinery.
We will start by selecting the vertical bar indicator. For the Vacuum Tower we will select the Mode
Attribute and drag into the vertical bar indicator.
Select the trend icon and drag it to the canvas. Now select the mode, process feed rate, electricity
consumption, water consumption and air consumption to the trend.
28 | P a g e
Then, select the radial indicator from the bar and drag it to the canvas. Drag the Production Feed Rate
attribute to the radial indicator. You can change it appearance by editing it to you satisfaction. Now,
you can copy this object by selecting it CTRL C and then CTRL V to past into the canvas. Now you drag
the electricity consumption to the radial indicator. You can repeat for the other variables.
The following figure shows an example of what you can do.
You need to connect your Digital Plant PI AF in PI Vision.
Then Create a trend with the key variables and modes
You also define a Digital Refinery Overview.
You access the attributes you want show from the left pane
You can add a link to the Unit Event Frames for detailed analysis
You can add a link to your PowerBI Cloud dashboard
You can add a link to each of the units from a left hand side menu.
Now, you can pick the Asset and navigates for all area of the plant. All the variables will change to the
appropriate Unit Element and display the right information. This type of displays are created using the
PI AF Structure using the Digital Plant Template.
This display shows the whole plant with each unit-operating mode.
From here, a link can be set to go into PowerBI and to each process area of your digital plant. You can
add link to access the events frames for a particular unit for detailed process analysis. The Total Energy
and Water Consumption for the events are summarized the left pane attributes section.
We have decided to use the PI AF/Event Template to define the displays automatically to minimize
the set up time and maintenance. The users can modify and adapt to their needs for these basic
29 | P a g e
Now, becomes the interesting part. How to visualize the event framed data?
Click on the Event Frame icon on the left side of the display. You will get the list of all event frames
captured for you by the PI System.
Now, select on Event Frame with the right hand side of the dashboard.
Then, click the right side of the mouse.
Then, select Compare Similar Events by Type.
30 | P a g e
The display shows the Event Frames comparison for you. You can select a particular line in the Gantt
chart to get the details about it. You can see the total electricity consumption for this particular interval
of time. This is information is calculated form your PI Unit Event Frame Template. In addition, you can
see the minimum and maximum values for all the consumables variables associated with this Unit Event
Frame Template.
The Digital Plant Template provides these Event Frames directly from the PI AF Unit and Event Frame
Templates that you configured.
Now you can select the Event Details to get additional information for the event.
31 | P a g e
You can see all the details for a trouble operational mode for your vacuum tower. The trouble
operational mode usually goes undetected in process plants. You can add all the trouble times for your
unit for the electricity and calculate your energy intensity index. You will see that is very high compared
to your running OK mode. You can see the aggregation of the data using PowerBI Displays as show
previously. Now, that you have these tools to analyze you plant data you can many improvements to
improve your yields and reduce operating costs.
Optional instruction for a physical implementation after the simulator is transformed into deployment
with real sensors and events.
Analysis using PowerBI
Microsofoft Power BI is a suite of business analytics tools that deliver insights throughout your
organization. Connect to hundreds of data sources, simplify data prep, and drive ad hoc analysis.
Produce beautiful reports, then publish them for your organization to consume on the web and across
mobile devices. Everyone can create personalized dashboards with a unique, 360-degree view of their
business. And scale across the enterprise, with governance and security built-in.
The process event data generated by the PI Unit Event Frame Template is imported into PowerBI for
further analaysis and visualization.
The first step is to define you PI ODBC Connection to the AF Database. (Michael to Review, Dec 20,
1. Select ODBC 64 bit for your main search
32 | P a g e
Run your ODBC Configuration and select System DSN Tag. Technically, we are creating a DSN (Data source
name) which is pointing to the PI system.
3. Select add and select PI ODBC Driver
4. Configure your PI ODBC Driver as follows to connect to your PI AF Server.
a. ODBC Data Source Name = My PI System
b. Description = Your PI ODBC connection to your PI AF database
c. PI SQL Data Access Server Name = PISRV01
d. Data Source AF Server = PISRV01
Click OK and your done.
33 | P a g e
You will need the specific PI ODBC query to get the data to PowerBI. A step by Step if the PI SQL
Commander Wizard PI Event Frame query is provided in Appendix III for you.
This is equivalent to connect to your PI Vision to the PI AF to present the PI AF Analytics and Event
Frames into PI Vision. We will describe these steps in the next section.
Once you have your ODBC Connection set, then, open your PowerBI Desktop as show in the next
Open PowerBI and select GET DATA the option Other and then click on ODBC.
Scroll down to find ODBC. Select ODBC from the list and click CONNECT.
OAB Jan 2018> Need to get a better screen shot with the My Pi System in the menu.
Select you PI ODBC called My PI System and click OK.
34 | P a g e
Select Windows and Use my current credentials. Then, press CONNECT.
You will get all the PI AF Models that are in this PI AF Server. Please find your PI AF. In this case we have
to find the database called OilRefineryExample.
Select the PI AF Database View that you create earlier called OilRefineryPowerBI. Select DataT and find
Select the PI ODB PI AF OilRefinery Example and the ODBC query call OilRefineryPowerBI. Check it and
your will see the results as shown the previous diagram.
The Click LOAD to import the database into PowerBI for visualization and additional analysis.
Please see Appendix III to see how the PI SQL Commander enables to generate the PI AF View called
OilRefineryPowerBI. For this unique strategy, this is a simple way to extract the event-framed data. The
strategy enables to pass all the PI AF definitions into PowerBI for Cortana and R to work.
Click Load. The data will loaded into the desktop PowerBI.
35 | P a g e
You have loaded the PI Operational Insights into PowerBI. The data attributes from the PI Event Frames
Template are at the right hand of the application. Now, you have an empty canvas with a selection
visualization tools and customization tools of your PI Operational data. You will be able to view the
multivariate set of assets, variables and events using Business Intelligence. In addition, you can use
Machine-learning tools and use Cortana Artificial Intelligence once you Publish your creation to AZURE.
Your dashboard can have many pages. In addition, you have the tools to manipulate the data to create
additional views. We will show how can create Monthly, and Shift Event Selection for the time data set.
By having the PI Data in the cloud you can collaborate and share with others for additional insights.
This display shows you the PI Event Framed operational insights to generate your own analysis.
Let first create a few slicers for the data. Let’s create a time and asset slicers to analyze the process and
consumption variables by time, shift, month, operational modes and other variables.
Let start by selecting a visualization tool such as the Paretto Diagram. Select the Asset, Mode and the
Variable that you want to analyze for the whole plant.
36 | P a g e
Automatically you will get the following diagram.
Note: You need to use the format tool to change the selection of assets. Use the format tool from the
visualization panel. Select Selection Controls and switch the Single Select to off. You can also select the
Show Select all to view the whole plant at once. Then you can decide which units you want to have
closer look.
You can use the asset slicer to add more units into the Pareto chart. A Pareto chart is a bar graph. The
lengths of the bars represent frequency or cost (time or money), and are arranged with longest bars on
the left and the shortest to the right. In this way, the chart visually depicts which situations are more
significant. This simple example shows you the times that Diesel Hydrotreater has been running in OK
mode, troubles, idle, down or maintenance. The running Ok is most valuable contribution has it should
Once you publish your Dashboard in the Azure Cloud you can share, analyze and get the PowerBI to
extract automatically the Operational insights generated in the PI AF Analytics and event framed
operational insights. Check the Publish Icon on the PowerBI Desktop application.
We will show a few examples on how to create the visualization examples based on the Fields imported
from PI.
You will start by define the Time, followed by the list of assets and then using the Paretto Visualization
to compare the behavior of all units in the plant by process feed rate and their operating mode. Once
you create this few examples you can improve the visualization to your desire liking by using the Roller
icon on the Visualization Display options.
Attached are a few screen shots of sample PowerBI Template that has five section to visualize the
operational data. You can use our example power BI to learn more how to create better views. You can
reuse this template by important your plant data and change the asset, mode, consumption and process
variables connections to the visualization tools. This template includes:
An Overall Production and Consumption View,
A Monthly Production and Consumption Overview for all units based on Operational Mode
PowerBI Template,
An Overall Process Production Effectiveness shows the percentage of time for each unit for
each operating mode for the selected operating time by shift,
37 | P a g e
Individual Units analysis and Utilities by process feed rate. Energy Intensity and Water Intensity
Indices, and,
Many additional self-service views can be generated. In addition, you can use Cortana in the
Cloud to propose innovative ways to present the event-framed operational data.
We will live the opportunity to add more value. For example:
It would be nice if we could have some numbers in terms of dollars here, i.e. how much money was spent on
power/water during times when the unit was not operational. Or, the cost of raw material (or lost production)
from times that the unit was running in a bad state.
This would make the business impact of this type of analysis very clear. You can use the cost of
electricity to be around $0.10 per Kwatt. For the Crude Oil you can assume $60 per barrel.
Unfortunately, we are using 100 as a nominal value for the production feed. So, you might want
to wait until you get home to reconfigure the example using real tags from you plant.
Production and Consumption Overview for all units based on Operational Mode PowerBI Template
38 | P a g e
Monthly Production and Consumption Overview for all units based on Operational Mode PowerBI
Overall Process Production Effectiveness shows the percentage of time for each unit for each operating
mode for the selected operating time by shift.
39 | P a g e
Individual Units analysis and Utilities by process feed rate.
The next screen shot shows you the Azure PowerBI. So, you can Publish your PowerBI to the Cloud. Then, you will
see a screen as such: You will be able to enter you data questions into the PowerBI using Cortana. Cortana is a
given once you open PowerBI.
It is the easiest way for AI to check the PI Event Framed data extracted form PI. It does it automatically. It propose
you the questions that you might want to ask. It works better in your cellular phone.
Microsoft is working using Apple Apps due to rich environment to innovate with these brand new tools.
40 | P a g e
Once the PowerBI Desktop is uploaded to the Azure Cloud then Cortana becomes available. This last
screen shot show the use of Cortana include in PowerBI. Here you type your question about the data
and Cortana suggest a diagram to visualize the data by you.
The next screenshot shows you the PowerBI running on your cellular phone.
In this case the questions was: Provide an view with all assets production feed rate by operational mode
for the time rage in consideration: Asset: Process Feed Rate: Mode:. With these three identifiers that
whole plant is analyzed automatically using a Paretto Chart proposed by Cortana. In addition, Cortana
provides comments on your cellular phone with suggestions about unusual correlations of the data.
The answer is Paretto Diagram for all each process unit showing the process feed rate values for each of
the operating mode for the period selected. You can see that most of units have considerable minor
losses. These values shows that could improve their operating time in RUNNING OK mode considerable.
It also shows the losses of production once in DOWN TIME, IDLE and Maintenance modes.
41 | P a g e
This figure shows the simple way to configure the scheduling for capturing data into PowerBI continuous
evaluation of plant overall production effectiveness.
How to add the shift slicer calculation?
How to add the operating mode percent calculation?
42 | P a g e
Part III Industrial Adaptation for the Plant Template Tips
We will share a few tips on how to convert the digital plant simulator to using real sensors. You will
have to select the appropriate tags to create you global variables for each unit (Process Feed Rate,
Energy Consumption, Water Consumption and so on.
PI Analytics for Unit Global variable calculations
How to add all the Power or electricity tags associated to a process unit area?
How to create an analysis to generate the Electricity Consumption for a process area?
Using the PI Analytics you create a Consumption calculation section to add all power measurement in
the UNIT Area and associate to the Electricity Consumption. If you happen to have several Feed Rates
create a Process Feed Rate Analytics section to add all the process feed rates to the area.
Energy Consumption Global Variable calculation
This screenshots shows on the unit elements an all the electricity consumers attributes and the link to
measurement tags. This list of consumer electricity variables are then aggregated in the PI Analytics to
find the total electricity for the unit. This variable is a global variable which is used in the Process Unit
Template. The same procedure is used for all consumable variables.
43 | P a g e
Energy Total Estimate validation and Total consumption the Unit Area.
This figure shows the PI Analytics section called Consumption Aggregation. First, the raw value is check
for validity and the Electricity Consumption calculate variables is map into the global attribute for the
current element.
The Electricity Consumption is mapped to an output attribute as seeing the preceding diagram. The
Attribute now points to a Data Reference called Consumption Aggregation.
PI Analytics to create augment performance calculations and/or softsensors estimates
How to create special performance calculations or softsensors estimates?
You can create an analytics section to create especial calculations by first validating the key inputs, then
create and AND true to check that all variable are within reasonable limits. The most important check is
if the unit is under the Running OK condition. Then use the desire calculation and map the results to an
internal global variable for this area.
PI Attributes Example for Calculation of ConcFlow, TailsFow and Recovery. This is using a simple mass
balance based on the known feed flow and the metal assay in the feed, concentrate and tails. As such
the estimates for the concentrate and tailings flow for the area are calculated. Most important, the
metal recovery for the area is estimated for optimization guidance.
44 | P a g e
This PI Analytics screen capture shows how you can add performance equations to mass balance or to
evaluate softsensors. It uses the data validation and data classification algorithms to augment the real
time data into process operational insights. With this strategy, we will process augmented attributes
that we analyze and visualize using PI Vision or PowerBI in real time.
This is another example showing the use of the estimation of a variables when the Mode is in Running
and Trouble Conditions. This estimate is based on the solids and water mass balance around the mill. A
bias is added to account for the additional water to the circuits which is not measured.
PI Datalink for data extraction using PICompDat and PITimeDat algorithms
How to using the operational Running Mode operating condition to extract a sub set of data to develop
an empirical correlation?
PI DataLink directly integrates your PI Server data with Microsoft® Excel® so you can easily analyze all
your operational data using the powerful analytic features of your spreadsheets. From reports to models
to data analysis, the full power of both the PI System and Microsoft® Excel® are instantly available to
your spreadsheet users with PI DataLink. PI DataLink can be used to extract filtered data based on event
45 | P a g e
data for the right times. It will provide all the selected variables to mode a certain independent variable,
i.e., particle size for example and the associated process variables such a Process Feed Rate, Water
Additions, Power, % Solids to name a few.
The PIComDat extract the set of times when a certain PI Expression is true. So, we extract the time when
the process is Running OK. The dependent variables are extracted using the PITimeDat Algorithm. This
subset of data can be used to predictive analytics regression modelling.
This PI Datalink spreadsheet extracts the values data for the Running mode. It extracts all the variables
in the columns for the exact times that the Running Mode is true. As such, you end up with a good
candidate to generate an empirical process model to estimate the Holly Tag Value Y.
The Sub of Running data is used to model the data Y=f(X) using a Multiple Linear Regression using Excel
Data Analysis (Analysis Tool Pack Addin). The Model is used with PI Analytics to predict this variable
based on the operating variables. It is a robust way to create a soft sensor since the particle size
measurement is often having problems. The resulting equation is programed in PI Analytics as shown in
the next paragraph.
Particle Size = 87.75+ 0.01 * Process Feed Rate + 0.0009* Process Water + 0.01*Electricity Consumption
You can additional variables such as the total power of the mill and the ore type, which will improve
your correlation. The current one as an R of 71%.
46 | P a g e
PI Analytics Multiple Regression Estimation when Unit is in Running Mode
Now that we have obtained a multiple regression for your softsensor using the Unit Operating Mode =
to running. We can systematize the use of this strategy using PI Analytics.
We create a Child Element with the Input Variables and the output of the PI Analysis.
The following screenshot shows the PI Attributes required to esnter and to read the operating parater
and the equation coefficients for your Regression on line data.
The following screenshot shows the PI Analytics expression to run the MLR Prediction for a particle size
prediction in a grinding circuit. The shown example can be improved by adding a line stating the
calculation should be performed only with the unit Mode is in Running or Trouble as shown in the
previuos examples.
47 | P a g e
The outputResults are mapped to the output attribute for further analysis and visualization. We will
share a few additional examples which uses similar strategies to augment the use of raw data by using
the data classification and filtering strategy.
PI Analytics Linear Regression Example to estimate the slope of a set of measurements
How to perform linear multiple regression for a set of measurements using PI Analytics in real time?
PI Analytics can also be used to perform online linear regression of models as a softsensors to see
changes in the slope of a linear model.
A typical instrument that provides several measurements is a particle size analyzer. It is very dificult to
use the whole size distribution. It is good practice to model the size distribution to a Gaudin Schumann
model. As such, we can estimate the slope of the distribution. The a parameter is measure of the
hardness of the ore. The following figure shows the idea.
The PI AF Graphics shows the predicted particle size and the measured one.
The following figure show how to perform the estimation of the A Distribution Modulus and the Size
Modulus for three sizes measurements at the hydrocyclone overflow.
48 | P a g e
PI Analytics showing the linear regression model to estimate the slope of the size distribution or Gaudin
modulus. This parameter measures the shape of the size distribution. This shape affects the recovery of
the metal in the production process.
This PI AF screen shot shows two linear regression estimates of the slope of a particle size distribution
measures at the SAG Feed (Split Engineering) and the hydrocyclone overflow product (Outotec PSI 500I).
Changes in the slope indicates that the ore is harder or softer. It is a great soft sensor to characterize
the hardness of the ore to improve the metal recovery based on ore hardness or ore type.
49 | P a g e
Having the shape of the size distribution enables to improve the grinding performance to improve the
gold leaching or the flotation recovery in mineral processing plants.
Using Microsoft Machine Learning Studio to create empirical models
To improve the gamma of models available you can also import the created data set into Microsoft
Machine Learning Studio on the web.
Microsoft Azure ML Studio Regression Model.
The PI Datalink extracted data set can also be imported into Azure Machine Learning Studio. The
subset data is visualized to view constraint or bad data prior to use in a particular model. ML Studio
provides many types of models to choose from. The most widely use or multiple regression model or a
logistic function regression.
Once the process data is imported in ML Studio, you get a good depiction of the type of normality of the
process data. If you see a bell shape, you can determine that the data has a Gaussian normal
distribution. This means that it has been allowed to mode within a reasonable operating range to
accommodates fluctuations on the process disturbances with easy. Now, on the contrary, if you see
hyperbolas, you can infer that this variable is constraint against a certain value. These constraints are
the process restrictions imposed by the equipment or the process itself and should be avoided. The
data shapes tell a lot. We will not go into details but a good examination of the data sub set is required
to have a good predictive model.
We encourage you to try these examples to find an augmented set of process variables to move closer
to your process constraints to optimize your process plant.
The idea is to model the process chain supply so you can always find the optimal operating conditions
that maximizes the production while reducing operating costs.
50 | P a g e
This figure shows the Microsoft Azure Machine Learning Studio Model design.
First, you import the data set for your CSV file into the browser,
Then, you select the columns to select the desired variables to be modeled (Dependent variables, X).
This is shown in the screen capture. Then you select the type of algorithm that you would like to try
fitting your data. You select the Y, independent variable to be modeled.
In this environment, you can test several model and pick the one that best fits the data. We tested
several algorithms. The linear regression model was not the best but it provides a good correlation that
can be use with PI Analytics as a soft sensor.
51 | P a g e
The screenshot shows the featured weights or parameters for a multiple regression model to model the
Iron Yield based on the selected Process Variables.
In this case, an Iron Yield is obtained to evaluate the effect the selected variables and to estimate the
Yield for changes in Process Feed Rate, Feed type material and the water additions in the different
sections of the process.
Yield (Fe) = bias + 2.01*Concentrate Iron Content – 1.04 Iron Content Feed + 0.91 Iron Content in Tails –
0.69 Fe3O4 Content in the Feed + 0.5 Sizing Screen Sump Flow – 0.051 * R spirals Water Flow + 0.0 2* M
Spiral Water Flow – 0.01 * C Spirals Water Flow + 0.0097 * Mill Feed Solids Flow Rate
Key to success in modeling your process operational data.
This creation of real time data model a process where first a clear objective has to be set.
Example: maximize yields and reduce operating costs is a typical one.
Building this fishbone diagram shows you the process variables and all the events that could be used to
aggregate the data for achieving your desire outcome. As such, we need to configure these events to
build the right data sets to find the proper models from the data.
52 | P a g e
Fishbone (Taguchi) Cause and Effect Modeling for Maximization of Yields while reducing operating costs.
This is achieved by understanding the operational events affecting the outcomes. This is what we
have demonstrated in the example using the latest analytical tools available. The importance to
incorporate the People Events, Material Events, Equipment Events, which affect the results. We could
add weather, markets, and prices.
There is a lot we can do with Azure ML studio. Take a look at this video:
53 | P a g e
Appendix I LAB Summary:
Proposed Strategy
One Process Unit Template is used to digitize a process plant.
The business objective is to track the Production Variance for every process unit while tracking the
consumption of energy, water, and other consumables. The analytics is abstracted for all units using an
object that is configured to represent each physical asset.
One Process Event Template is used to aggregate the production and consumable variable for
enterprise tracking for every plant and process unit. The results are presented in PI Vision and PowerBI.
A. An integrated Unit Template provides the Analytics and Event Frame Generation performing the
online calculations and creation of insights for visualization, analysis, modeling of the operation.
A validation and classification strategy allows creating performance calculations using empirical models
or engineering analysis models. This strategy provides augment the value of the operational data for
process, maintenance, engineering and management to have one version of the plant for decisionmaking.
B. One Process Unit Event Frame template transforms the data into operational insights. These
are visualized, analyzed using PI Vision, PI Datalink, PowerBI and Machine Learning tools.
PI Vision enables the visualization and analysis of the process trends and pattern created from the
operational modes.
We use PowerBI to get data from the PI Event Frame Table. PowerBI Desktop facilitates the creation of
powerful visual and analysis to evaluate the seasonal and shift patterns created over time. An Overall
Production Effectiveness evaluation permits to differentiate operational, equipment maintenance and
scheduling problems. In addition, PowerBI can be published into Azure for Cloud for remote access and
analysis of the operational data guided by Microsoft Cortana. This creates a new synergy to share and
to collaborate at the enterprise level.
We use PI Datalink and the PI Time Data classification and extraction to align process variables to the
times that the process unit is running OK. The data set is use by the regression models or other
methods available. Next, we use Azure Machine Learning Studio to create Yield model based on feed
material characterization, process feed rate, water, energy consumption. The resulting predictive model
is used to create a mapping for the different type pf feed material and the effect the yields.
Appendix II PI System software components
The VM (virtual machine) used for this lab has the following PI System software components installed:
PI Data Archive
2017 R2
PI Asset Framework (PI AF) server
2017 R2
PI Asset Framework (PI AF) client (PI System Explorer)
2017 R2
54 | P a g e
PI Analysis & PI Notifications Services
2017 R2
PI Vision
2017 R2
PI OLEDB Enterprise and PI ODBC
2017 R2
PI Datalink
2017 SP1
PowerBI Desktop
For details on PI System software, please see: http://www.osisoft.com/pi-system/picapabilities/product-list/
Appendix III Extraction of the PI Event Framed Data.
The following paragraph will guide you to use PI SQL Commander to generate a view of the event
framed data to use PI ODBC to provide the data set to PowerBI or other tools. We will use the PI SQL
Event Frame Wizard to generate the procedure that will extract the data for us using the PI ODBC,
PI ODBC connection enables to use the Azure PowerBI scheduling tool to keep the PowerBI PI Data set
fresh while providing the BI tools for analysis from your devices anywhere in the world.
Step by Step
a. Open PI SQL Commander.
b. Select the AF Server
c. Connect using Windows NT integrated security by clicking OK
Select the PI AF that you want to connect to.
Select from PISRV01 your PI AF called OilRefineryExample (Generic Name)
Select the DataT .
Select Functions and open the arrow.
Select the PI AF Database to you want to extract event data from. In this case
Select Oil RefineryExample and find the DATAT folder from the list of files. Right click on top of
it. Select New Transpose function.
55 | P a g e
Select Dynamic Transpose Function (Event Frame) and Select Transpose Snapshot
Select the event frame template that we designed earlier. The results will be dependent on the
PI Event Frame Template configuration.
56 | P a g e
k. Select the Consumable Variables, then click Next.
l. The system will generate a name for you. Click Next.
m. Create function table
57 | P a g e
Click Next.
Click Next.
Click Execute and you will a message saying Execution completed successfully. Click Done.
58 | P a g e
You have created a Function Transpose for all Consumables variables analysis for the Event Frames in
your database.
The next step is to define SQL Query to tell the system to tell the times and events. This is done by
selecting and creating a VIEW. This view will create your PI ODBC connection to other Business
Intelligence tools such as PowerBI. Two steps are required here to create simple query and to provide a
view name for the PI ODBC to be found within the PI AF database.
1. Follow the suggested step. Take the SQL statement attached and paste in the View area <query
-----Copy paste this section in to CREATE VIEW <query> section-------------------------------------------------------SELECT ef.Name EventFrame, ef.StartTime, ef.EndTime, tc.*
FROM [OilRefineryExample].[EventFrame].[EventFrame] event
WHERE event.StartTime > '2018-2-1' AND event.endtime < '*'
) ef
INNER JOIN [OilRefineryExample].[DataT].[ft_TransposeEventFrameSnapshot_Consumable Variables] tc
ON tc.EventFrameID = ef.ID
59 | P a g e
2. In the <View Name> Section give a name to the VIEW NAME to your PI ODBC query so you can
find it when you will connect from the other system, i.e. PowerBI. The name in the case is
OilRefineryPowerBI. You can select any name.
Go to the Execute command and press it. It will say Command execute successfully.
If you execute the query you will get an equivalent set of data as you will get using the PI System
Explorer. Select New Query and paste the query as shown and press execute.
The results if the query is.
This PI Event Frames Table is now available for additional analysis and modeling in other external tools.
We have create a PI ODBC Query that now we can use to Get Data into PowerBI visualize the data event
frames by Unit, Mode, Process Feed Rate and the selected consumption variables for the Oil Refinery
example. Congratulations.
Appendix IV Connecting PI Vision to your Digital Plant Example.
Open PI Vision and go to admin section. Then, Select Manage Configuration and add the
OilRefineryExample AF data to it.
They go to your PI Vision and click the OilRefineryExample database that belongs to you. They you can
add dials, trends, values, links and so on. As always, the key here is that you will be able to navigate via
the whole plant very quickly and look at the unit event frames and the energy, water and production
values by mode.
60 | P a g e
Bascur, O.A. 1988. A control data framework with distributed intelligence. In Advances in
Instrumentation: Proceedings of the ISA/88 International Conference and Exhibit. Research Triangle
Park, NC: Instrument Society of America.
Bascur, O.A., and Kennedy, J.P. 1996. Measuring, managing and maximizing refinery performance.
Hydrocarb. Process. 75(1):111–116.
Bascur, O.A., and Kennedy, J.P. 1999. Real time business and process analysis to increase productivity in
the process industries. Presented at the 1999 ISA Conference, Houston.
Bascur, O.A., Plourde, M., Paquet, S., Morissette, S., and Gervais, D. 2017. A journey towards mine to
port operational intelligence. Presented at the 2017 SME Annual Conference and Expo, Denver,
February 19–22.
Steyn, J., Bascur, O.A., and Gorain, B. 2018. Metallurgy analytics: Transforming plant data into actionable
insights. SME 2018 Annual Meeting Proceedings, CO: SME. Minnesota.
61 | P a g e
Kellemer, J.D., Mac Namee, B., and D’Arcy, A., Fundamentals of Machine Learning for Predictive
Analytics, Algorithms, worked examples, and case studies, The MIT Press, Cambridge. MA.
Goldratt, E.M. 2014. The Goal: A Process of Ongoing Improvement. Great Barrington, MA: North River
62 | P a g e