Machine Learning and Predictive Analytics

Guidebook for Engineers

Leverage Domain Expertise to Drive Optimization – Without Needing a Data Scientist
Today, staying competitive
means progressing on a digital
transformation journey, including 100%
machine learning and predictive
Process Setpoints % Throughput
analytics. Not only can industrial
Recipe Settings % Quality
organizations capitalize on the IoT Line Machine
Raw Materials Properties % Uptime
opportunity, optimize operations
and generate greater profitability, Visual representation of OEE optimization
but engaging in the latest
Figure 1: Engineers can use analytics to optimize OEE for a product on
technologies also helps to attract a line/machine/plant and gain new improvement insights in just hours.

and retain the best talent.

Fortunately, the journey to success with machine learning To drive analytics and improve processes, process engineers can align domain expertise to five capabilities:
and analytics doesn’t mean that process engineers need to be 1. Analysis - automatic root cause identification accelerates continuous improvement
data scientists. Proven processes and software technologies
make analytics do-able for every industrial organization. 2. Monitoring – early warnings reduce downtime and waste

Engineers Are the Process Experts 3. Prediction – proactive actions improve quality, stability, and reliability

Process engineers have exceptional domain expertise to put 4. Simulation – what-if simulations accelerate accurate decisions at a lower cost
together process models – or Process Digital Twins – and 5. Optimization – optimal process setpoints improve throughput at acceptable quality by up to 10%
be able to interpret the models. This is the foundation for
improving competitive advantage and success with analytics.
Advanced analytics techniques are available to industrial
process engineers to align their domain expertise to these five
capabilities. To support the journey to machine learning and
analytics, GE Digital provides analytics technology training in the
form of self-serve detailed demo videos and application advice.
Additionally, while today’s software features enhanced ease of use and no-code implementation extensible
with Python, process engineers can still lean on product experts in combination with their own domain
expertise to mine data and leverage analytics to improve operations.

Success with Predictive Analytics

As an example, a leading food manufacturer was able to drive down customer complaints by more than 33%
through analytics. The manufacturer had struggled with weight control on a cube-shaped product. Make the
cubes too heavy, and the manufacturer was giving away product or producing watery product if the excess
weight was due to too much water. When the cubes were too light, the company was in regulatory jeopardy as
well as having trouble compacting the product into a stable cube shape.

The team used Proficy CSense to get a complete, correlated-by-lot and period picture of: ingredient specs,
process variables as run, and lab data – using the software to look for controllable factors that correlated to Figure 2: A leading food manufacturer was able to drive down customer
excess giveaway and then comparing periods with better weight control to the factors that were true then. complaints by more than 33% through analytics, using Proficy CSense.
Now, when the team sees how a raw material variance was successfully corrected for or a process disturbance
was overcome, that understanding is embedded into a new material spec, recipe or SOP. The smart analysis
with Proficy CSense yielded other benefits as well. Another example involves applying a smart predict project at a pulp
and paper manufacturer to predict Critical to Quality (CTQ) KPIs to
improve productivity and eliminate wastewater regulatory issues. As
a final example, a partner in mining delivered an Advanced Process
Control solution that increases throughput by 10% using smart
optimization technology.
 What is a control loop?
Predict Product Quality Demo
A simple form of a process controller is the thermostat which maintains the temperature of a room according
to a given setpoint. It operates as a closed loop control device, trying to minimize the difference between the
room temperature and the desired one.

The industrial version is the PID (Proportional-Integral-Derivative) control loop - an essential part of every
process applications. PID loops have been around for a very long time. The first pneumatic instruments
featuring a proportional controller were developed by Taylor Instrument Companies at the beginning of the
20th century,

Watch Video Nowadays, loop controllers are available as standalone devices called single loop controllers, but the
most common version is a piece of code that resides in a PLC (Programmable Logic Controller) or a DCS
(Distributed Control System). It makes it easier to combine them to create advanced control diagrams
like cascade or feed-forward control, or split range required for the complex control of food and beverage,
chemical, oil and gas operations, and more.

Predict Product Quality with Analytics

See how a process engineer can create a machine learning model to
predict product quality that will help reduce laboratory testing.

Analytics and Control Loop Optimization

As a use case example, process optimization is key in manufacturing,
and control loops are the critical components. “Out of tune” loops can
affect the quality of the product, the material and energy consumption,
and ultimately increase the risk of contamination in regulated
industries. AI and machine learning can be used to improve and
optimize control loops to generate big savings and reduce risks. Figure 3: AI and machine learning can be used to analyze and improve control
loops to optimize performance KPIs, generate big savings and reduce risks.
Control Loops: Small Tweaks, Big Impact
A lot of literature exists that describe the behavior of PID loops and how to tune them. However, it still Two apparently similar machines might require different settings as they are
represents a challenge for many as all the processes are different. equipped with sensors that will react to change in a slightly different way.
This might be because they use different technologies – a glass vessel vs a
The complexity of the process is obviously one of the criteria. Heat jacketed devices such as kettles, dryers,
steel vessel, which by nature have different inertia - or simply because their
reactors or pasteurization units can be hard to control. Using steam, the heat transfer is not uniform which
characteristics vary over time: aging valves, deviating sensors, etc.
might result in an overshoot during uptimes, making the control loops difficult to tune. Note that this is less
prevalent using water. Traditional cascaded loops will only solve part of the problem. An advanced analytics Loop tuning therefore doesn’t happen once. It must take place on a regular
system such as GE Digital’s Proficy CSense can help by looking at historical data to create a model of the actual basis if done manually and customized for each asset. Another option is real
profile and recommend new settings accordingly. The model will take into account the change of parameters time monitoring using AI and machine learning. Analytics make use of a suite
such as viscosity and steam pressure which affect the heat transfer coefficient and the flow pattern. of components to determine and understand the causes of process deviation
in industrial environments. Engineers and data scientists can analyze, monitor,
predict, simulate, and optimize and control set points in real time.

For process optimization, analytics solutions need to provide multiple

capabilities: process modelling and troubleshooting as well as online
deployment and real time monitoring.

Figure 4: An advanced analytics system can solve problems by looking at historical

data – including from multiple sources – to create a model of the actual profile and Figure 5: Analytics solutions need to provide multiple capabilities: process modelling and
recommend new settings accordingly. troubleshooting as well as online deployment and real time monitoring.

Data is prepared, visualized, and rules-based, data driven process models can
be constructed. Using these models, root causes of process deviations are
identified, so processes can be optimized.
Predictive Analytics and Bad sensor data can mean lost product, downtime, compliance issues, and safety risks as well as a dirty data foundation for
digital transformation and continuous improvement programs. Industrial organizations need to have good data that can be

Sensor Health leveraged for operations, ad hoc analysis, and enterprise analytics.

Over time, sensors tend to deviate, impacting processes and operations. But, it’s time consuming – and impossible for most
Applications for predictive analytics are organizations – to manually determine if and why sensors are working or failing ahead of increasing risk.
endless, but a possible first step: engineers
Predictive Analytics and Sensors
can use analytics to monitor sensor health.
Engineers can employ a predictive analytics app such as Proficy Sensor Health to continuously monitor and analyze sensor
data. Users can target anomalies and minimize their potential impact. The analytics app provides an easy way to automate the
detection of bad sensors, where data is deviating from normal conditions.

When an anomaly is detected, the app can generate alarms to speed repairs, replacements, and recalibrations.

By using predictive analytics to monitor sensor health, engineers can:

• Reduce downtime: Sensors are often used to provide indications that equipment is
running correctly. Incorrect readings can lead to equipment failure or damage. Early
detection of a sensor that is no longer giving accurate or consistent results can provide
advance warning that enables maintenance to replace or recalibrate the sensor before
the worst happens.

• Improve product quality and compliance: Sensors are often used for measuring
the results of a product or to ensure the ambient surroundings of a process are within
specification. If the sensors used to measure the product or environment are not
accurate or functioning correctly, it can lead to a product being out of specification.
Providing early warnings can reduce costs related product recalls or scrapped product.

• Ensure sensor data quality: Ensuring data quality in downstream analytics is part of
IoT-fueled improvements. If the intent is more advanced use of analytics for a process,
the need for ensuring data quality is critical.
Figure 6: Follow these strategy steps to good sensor health, leveraging predictive analytics
and machine learning.
How to Select the The following advice provides guidance for engineers.

Right Predictive 1 Make sure that “predictive analytics” isn’t buzz hiding risky, unproven software and

Analytics Solution
newbie “experts.”

For example, GE Digital has offered and implemented analytics solutions for more
than 15 years, serving industrial organizations around the world across a diverse set of
Every industrial organization knows that it’s industries.
imperative to move ahead at light speed with
predictive analytics. Consequently, new analytics Work with a partner that you can trust and know will support you for the long term.

startups and consulting companies are showing up

every week. 2 Select industrial advanced analytics with a troubleshooting component that enables
engineers to rapidly troubleshoot continuous, discrete, or batch manufacturing process
How does an engineer select the right predictive performance by mining insight from available sensor and
analytics solution? production data.

Seamless connectivity, rich visualization, and predictive analytics enable engineers to

analyze operating scenarios, quantifying the impact that operational changes will have
on key performance metrics and identifying causes for performance variation.

Additionally, to support the full IoT value journey, look for capabilities from simple
calculations to predictive machine-learning models to real-time optimization and
advanced-control algorithms.

3 Be sure that the analytics package enables engineers to rapidly develop analytic
solutions – supporting improvements in production throughput, yield, quality, and
efficiency with significant margins.

A comprehensive analytic solution-development environment provides visual analytic

building blocks to build and test calculations, predictive analytics, and real-time
optimization and control solutions with connectivity to real-time and historical data
sources and drag-and-drop access to rich functional libraries.
Figure 7: To support the full IoT value journey, look for capabilities
from simple calculations to predictive machine-learning models to Plug-and-play connectivity to historical and real-time data sources and automation
real-time optimization and advanced-control algorithms. systems make for faster configuration. Built-in support for data quality makes real-time
data cleaning and validation easy.
4 Confirm that the analytics package can speed deployment with templates for
greater efficiency.

Engineers should be able to save analytics solutions as reusable templates

for easy deployment to similar assets or process units. Additionally, while
the analytics troubleshooting component should enable engineers to find
answers faster with analytics-guided data mining and process-performance
troubleshooting, the development/configuration capabilities should allow them
to more easily capture expert knowledge and best practices into high-value
analytic templates for rapid enterprise-wide deployment.

Focus on analytics solutions developed with engineers – not just data scientists
From Small Projects to
5
– in mind.

Multi-Plant Optimization
With an analytics package that is accessible to engineers, teams can create
a Process Digital Twin for smarter operations. Visual drag-and-drop analytics
accelerate time to value and reduce dependence on data scientists and
programmers. Online demos enable rapid mastery of the software with easy- All automation and process engineers can and need to develop capabilities in
to-follow demonstrations and guided simulations. analytics and machine learning to remain competitive – both at an individual
professional level as well as to help their industrial organization – in our world of
As previously mentioned, a rapid development environment is critical. The best
digital transformation.
solutions provide rapid wizard-driven data mining for engineers for fast time-to-
insight, an easy visual drag-and-drop environment for subject matter experts Over time, engineers can go from small projects to pilots to multi-plant optimization
and engineers, and analytic solution templates without programming for simple with deep application of analytics. Engineers’ deep domain expertise provides a
calculations, data cleaning, maths, statistics, machine-learning models, real- foundation for modelling processes and developing the analytics that are game
time optimization, and advanced process control. changers in very specific applications. The combination of applied analytics
technology with those Process Twin models uncovers hidden opportunities for
improvement over-and-over again.
If you’re ready to optimize with analytics, GE Digital’s Proficy CSense turns raw data
into real-time value with a Process Digital Twin. The software uses AI and machine
learning to enable process engineers to combine data across industrial data sources
and rapidly identify problems, discover root causes, and automate actions to
continuously improve quality, utilization, productivity, and delivery of
production operations.
Advanced Analytics Case Studies
Global References for Proficy CSense
Major Food Company uses analytics to
reduce waste, improve quality, save on
raw materials, and optimize OEE

The Problem What was done? Value to customer?

The company was looking to apply analytics to MES data to Proficy CSense was used to analyze combined Proficy Plant • Reduced product waste by 75% with $240K per year
optimize OEE, reduce waste and improve product quality. Applications data (Raw Materials, Production Context, savings
Final Product Quality) and Proficy Historian data (Process/
• Improved Quality reducing Customer complaints by 38%
Sensor) to identify best operating settings and raw material
properties to ensure good quality product, avoid waste, • Reduced Raw Materials cost with $65K per year savings
maximize OEE & minimize cost.
• Improved OEE by 9%

Conditions to AIM for to avoid waste and ensure quality

FIBO Wall Panels Manufacturer gains insight
in hours about how to improve quality
https://fibosystem.com/

The Problem What was done? Value to customer?

The company was looking to apply analytics to MES data Proficy CSense was used to analyze combined Proficy Improvement insights in hours by combining proven
to optimize process OEE performance, throughput and Plant Applications MES and Proficy Historian data in hours analytics and MES software.
product quality. identifying and confirming critical process variables and
thresholds, and raw material properties and highlighting or
confirming multiple potential improvement insights.

Insights like which combinations of raw material vendors

work best for them, or what are the optimal raw material
quality thresholds, beyond which their risk of defects increase
significantly, or what are the critical process variables and
their thresholds to produce best final product quality.
Skjern Paper Uses AI to Improve
Product Quality and Reduce Waste
Skjern Paper, located in Skjern, Denmark,
has always been ahead of its time.
The only paper mill in Denmark, Skjern Paper started
production in 1967 with the idea of manufacturing paper
exclusively from old newspaper, becoming an innovative
leader in sustainability.
Today, Skjern Paper is owned by Buur Invest A/S and manufactures 75,000 tons of paper and board
products each year from 100% recycled fiber.

The majority of its products serve end customers as:

• Cardboard cores in toilet rolls and kitchen towels

• Composite cans

• Cardboard in binders

• Solid board boxes

• Gift wrap paper rolls

• Sheet interlayers for pallet goods

The company’s environmental and societal commitments extend to Circular Economy and membership
in the UN Global Compact and its Nordic network.

Supporting its commitments to the environment and customers, Skjern Paper relies on innovation in its
production. Using the latest technologies and processes, the company provides high quality, flexibility and
just-in-time delivery to its customers – with delivery within 24 hours to most of Europe.

Recently, the plant manager at Skjern Paper looked to Artificial Intelligence (AI) and Machine Learning (ML)
to take production to the next level.
A Strategy to Leverage
AI in Production
“I’ve been looking at the area of AI for some time,” explains Skjern Paper’s
Technical Manager Erik Møller. “The industry has gained insight into improved
efficiency using AI and ML. We decided to make digitization and production
improvements through AI and ML part of our strategy.”

GE Digital partner, Novotek, introduced Møller to Proficy CSense, an industrial

advanced analytics software package that predicts future asset and process
performance. Møller was able to easily explore CSense’s technical capabilities
through a series of YouTube videos and demos.

“Proficy CSense looked very user friendly,” Møller says. “I saw that it has the
capabilities that we needed; the price was right, and GE Digital was willing to provide
six hours of free consulting to help us get started. I was interested to see what
possible production issues we could identify when using the CSense product.”
Accelerating an Initial AI Project
Paper plants have hundreds of PID control loops that can
cause process variation and contribute to quality issues, if not
maintained in a healthy condition.
At the same time, while Skjern Paper has extensive quality assurance systems, the team does
not have many real-time ways to measure paper quality, making real-time quality control
difficult. Operators would check quality samples for a whole reel of paper at the end of a
production run, which involved a delayed lab analysis and the inability to adjust production
earlier in the process.

With a goal of avoiding or reducing 5% of quality rejects, especially when switching between
different products, Møller took advantage of free consulting with a GE Digital AI and ML expert to
jump start the analytics project.

In six hours of consulting, Skjern Paper captured insights from the Quality Prediction and Analysis: By examining data and applying the capabilities of CSense,
initial project. The team used Proficy CSense to: Paper Burst Strength Møller was able to discover possible causes of errors in the
production line. When producing paper, manufacturers add
1. Analyze: Used available data to discover causes of quality The team built a model in CSense around a quality parameter
dewatering chemicals. Møller discovered in the production
variation and rejects related to the Mullen burst strength of the paper. The Mullen
process that they are producing scrap due to adding too much
Burst Test is an industry standard to measure the paper’s
2. Monitor: Monitored the health of PID control loops to reduce chemical to the pulp mix.
physical strength and fiber bond. Skjern Paper has 20 data
process variation
points inputting to the CSense model, which predicts the burst “The model provides instant feedback on the differences in
3. Predict: Created a predictive model from available data to strength parameter. the level of the chemical,” Møller says. “It was the error in the
predict product quality in real time, enabling real-time quality production. When we had problems with dewatering, we would
Burst strength is a sensitive measurement, according to Møller,
control to reduce quality rejects add more chemical, but in this case, we can add less. We had
and a hard one to start with – however, even with that challenge,
thought that more chemical was better in the production, but we
the team saw early predictive success.
have new insight with CSense. Now, we can reduce the amount of
“We built a rather fine model and compared data from the model chemical used and reduce scrap – which decreases our costs. It is
with actual quality data later,” Møller explains. “It is not completely a great capability that gave us benefit straight away.”
aligned but showing good results. I’m quite pleased with it.”
Quick Insight and Results
As the company moves from these initial “Decreasing scrap and chemical usage and increasing
insights and into production, operators will production capacity through CSense are all ways that we are
helping the environment,” Møller explains. “Also, as we use
benefit from the real-time AI optimization.
CSense more, we can gather data from the supply chain and
This new insight from AI also supports the company’s optimize that way too. The capabilities are there, so it is just a
commitment to Circular Economy. matter of structuring the data and the model correctly.”

In summary, after just six hours of consulting, Skjern Paper

was able to realize:
“When we finish implementing and adjusting 1. Analyze: New insight was gained about how dewatering
the model, we will see a reduction in the chemicals can affect product quality
scrap that we are producing,” Møller notes. 2. Monitor: It was then shown how Proficy CSense can be
“We will get an early warning when quality is configured to monitor PID control loop health to detect
changing, and the operator will get an early suboptimal PID control loops early to avoid process
indication. The master plan is to have the variation

indicators in the SCADA system. We will be 3. Predict: It was shown how a predictive model can be
able to solve the reasons for issues by using created from available data and can be deployed to predict
product quality in real time, enabling real-time quality
the CSense visualization.”
control to reduce quality rejects and waste
Next Steps
While the initial project was about gaining some quick wins with a small model
and realizing the potential of AI, Møller has a plan for next steps.

These include:

Taking the insights already learned and driving the realized process changes into
production

• Constructing data that will lead to more insights

• Tackling downtime – starting with locking the data related to downtime, doing
some manual manipulation of the data, then investigating how CSense can use
that data

• Training additional team members on using CSense and having them take over
from Møller’s foundation of work

• Build a model to gain insight into increasing capacity

Skjern Paper will also be implementing Proficy Operations Hub for main PID loop
monitoring and visualization.

AI Recommendations and Insights

With its journey into AI, Skjern Paper is continuing its tradition of being at the
forefront and ahead of its time.

What does the team recommend to other companies?

“I would recommend Proficy CSense to other companies,” Møller says. “Also, to

proceed with AI, my advice is to start looking at how to get the dataset. We were
quite lucky that I have been working with the data in the mill. The dataset that I have
been producing is very compatible with CSense.”

Lastly, Møller foresees quick return on investment related to Proficy CSense. The team
has already gotten valuable insights so far in a short time.

“The more we use the Proficy CSense software,” Møller concludes, “the more ways we
keep finding to apply it. With AI, we can take production to the next level.”
Additional Examples
Exxaro and Wabtec Bump Up
Throughput at Coal Operation

10%
Increase in throughput

+98%
Utilization since commissioning

6 month budgeted cycle

Successful POC
Introduction
About Exxaro
Since their formation in 2006, Exxaro Resources Limited has grown to become one of the largest
and foremost black-empowered coal and heavy mineral companies in South Africa, with other
business interests around the world in Europe, the United States of America, and Australia. Our asset
portfolio includes coal operations and investments in iron ore, pigment manufacturing, renewable
energy (wind), and residual base metals. In 2019, Exxaro produced 45.6 million tonnes (Mt), reflecting
contributions from our expanded flagship Grootegeluk mine. At 31 December 2019, the group had
assets of R68.8 billion and a market capitalization of R47 billion (US$3.3 billion).

Watch Video

Challenge: Like many industries, coal processing Action

is complex, and digital transformation provides With research showing that Exxaro could expect a 5% gain in
throughput, the company worked with Wabtec to implement
an opportunity for improvement. The absence of Advanced Process Control, leveraging GE Digital’s Proficy
advanced process controllers (APCs) in the coal CSense and process digital twin technology, at its GG45
process plant. Following success, the team implemented two
processing industry led Exxaro to explore research more APC projects and explored a skills transfer process to
into opportunities for improvement, particularly in Exxaro engineers to maximize use of the solution.

the area of throughput.

 About the Speakers
Hennie Engelbrecht, Principal Engineering – Process
Improvement & Performance, Metallurgy, Exxaro
Resources

Hennie Engelbrecht is the Principal Engineering for Process

Improvement & Performance, Metallurgy for Exxaro
Resources. With nearly 20 years in mining and metallurgy
including managing the operations of the largest coal
beneficiation complex in the world, he has extensive
experience with process and performance improvement.
Prior to his current role, Hennie served as the metallurgy
manager, coal beneficiation manager, and chief engineer of
process development. He holds a Bachelor of Engineering
Outcomes (Chemical Engineering & Extractive Metallurgy) from
• Nearly 10% gain in throughput Stellenbosch University.

• Utilization higher than 98% (from commissioning to date) Marnus Olivier, Lead Project Manager and Advanced
Process Control/Data Science Engineer, Wabtec
• Proof of concept completed within 6-month
budgeted cycle Marnus Olivier is a Lead Project Manager and Advanced
Process Control/Data Science Engineer for Wabtec. In this
• APC solution is now solely utilized to control the GG5 plant, role, Marnus develops and implements advanced process
including start-up, shutdown, and steady state conditions control solutions for global mining companies – with
• Continued collaboration and continuous improvement outcomes ranging from stabilizing pH and temperature to
optimizing throughput and reducing waste. His expertise in
applying advanced process control, predictive models and
analytics spans 10 years at a diverse set of mines. Marnus
holds a MScEng in Chemical Engineering from University
of Stellenbosch along with certifications in data science in
machine learning.
Herning Vand gathers valuable
knowledge with their historical data
Based in Denmark, Herning Vand strives to efficiently deliver clean water to its customers

Going Beyond Big Data

”Big Data” has become a buzzword and everyone

agrees that it is worthwhile to explore the large
volumes of data. Herning Vand has invested in an
advanced calculator that can help to interpret their
process data and already in the first project they came
across particularly valuable knowledge.
 “We have achieved some excellent results
with our first project and we see a great future
optimization potential with the tool.”
—Jan Ravn, Chief Operating Officer at Herning Vand

For years, Herning Vand has recorded and saved a large However, Proficy CSense shall, just like other simulation
amount of process data that is collected online from Herning tools, be used shrewdly and Herning Vand, together with
Vand’s 14 purification plants. process consultants from COWI who were responsible for
the initial consultancy, gained some valuable experience.
The data is logged with the clear intention to make Herning
COWI assists Novotek, a GE Digital partner, with process
Vand wiser and therefore better equipped to optimize
technical advice in connection with the use of
processes and the overall operations. However, for that
Proficy CSense.
to happen the large volumes of data must first be sorted,
analyzed, processed, evaluated and thoroughly compared so
that the important correlations and trends can be localized.

“A spreadsheet can be used to handle relatively large amounts of data, but it will not work when there
are too many parameters in play simultaneously. Therefore, last spring we invested in Proficy CSense
software solution, which is dedicated to finding and using mathematical correlations
of large amounts of data,” — Jan Ravn
Is there a connection?
“I see Proficy CSense as a closed box that is filled to the brim
with advanced calculation routines. We just feed the box
with the data we want it to compute and then we tell it what
parameters we want to observe.

Proficy CSense finds the right mathematical description of the

data stream and can then show two curves of the same data
stream. One curve contains the actual data and illustrates the
process as it was in reality. The other curve is generated by the
mathematical model that Proficy CSense has set and therefore
shows a simulated process.

When the program is fed with multiple data streams at once,

it automatically locates the possible relationship that exist
between the different data and that is exactly what we
are after.”

— John Sorensen, Senior Project Manager for water

Figure 1: Modeling of variations in
and wastewater at COWI.
ammonium concentration. Proficy CSense
has generated a model where the result is
shown by the blue line. The red line is the
measured values.

Figure 2: Simple comparison of trend

When CSense has found the mathematical models and curves to get a quick overview of the
correlations, the user can then determine which parameters
situation. The figure shows the ammonium
to focus on and what to look out for. Just like when you insert
concentration in four process lines.
different values into an equation. Here it is just an automated
solution to an almost unlimited number of equations.
”We can find correlations that we did not
know existed and our theories can be
confirmed or dismissed. Once we have
located the significant correlations, we
can begin to optimize the process based
on this new knowledge,”

“If, for example, I want to have output

A as high as possible, how should I then
set inputs B, C and D?”
—Jan Ravn

Figure 3: Proficy CSense has identified the parameters that have

the greatest influence on a given run-off result.
A savings of DKK 500,000
The first project, which was the optimization of the gas and too many parameters in play can make it difficult to
production that is based on sludge from the purification plants, understand the results.
gave an output improvement of as much as 20%.
As Jan Ravn explained, "Our advice to new users is that you
The gas is used for electricity production and according to start simple and carefully consider both what it is you want
Jan Ravn the increase corresponds to an annual additional to have answers to and also what parameters are likely
production of approximately 400,000 kilowatt hours, which to affect these answers. If you do not know exactly which
earns Herning Vand approximately DKK 500,000. parameters affect your focus area, you can gradually reduce
the number using Proficy CSense as you test each one. When
Herning Vand is now engaged specifically in a series of small
the data streams are recorded and analyzed they fit together
defined projects that will ultimately achieve the goal of
mathematically, so when you adjust each parameter up and
making Herning Vand energy neutral.
down with the mouse, you see how the other parameters are
“We knew roughly which buttons we had to press to achieve affected to either go up, down or remain unchanged. When Jan Ravn and John Sorensen
this gain and so CSense should not take all the glory. you have isolated the relevant parameters there is the option assess the collected information.
However, the tool makes it much easier for us to check the to activate the tool to simulate towards an optimum process
accuracy of our assumptions and it can also show us the within a given framework."
way to the process-related correlations that we cannot find
ourselves.

— Jan Ravn
The first project is only the tip of the iceberg in relation to what we expect to
The Art of Definition achieve with Proficy CSense. We have an ambition that the tool will be used
The advice from Jan Ravn and John Sorensen is that you regularly for small and large projects, and therefore assist us to pick all the
have to be good at defining the amount of different data that
low-hanging fruit that would otherwise be missed during a busy working day,”
CSense works with and you get the best and fastest results if
you have relatively robust process knowledge.

The more data you put in means you get more answers out The findings are instantaneous and the application potential is great.”
at the other end. However, there is of course an upper limit,
— Jan Ravn
Positive changes in Rockwool
Proficy CSense's Troubleshooter advances Rockwool's capacity for process analysis

About Rockwell

The ROCKWOOL Group is the world’s leading supplier of

innovative products and systems based on stone wool,
improving the environment and the quality of life for
millions of people.
“A year ago, one of my employees New Usage

enthusiastically told me about We expect to make a great use of Troubleshooter to a

whole new production process in our new factories, while
a software system for process at same use it in our existing environments. We also have
a large ongoing project for automatic control of a complex
analysis, Proficy CSense's production process, including environmental facilities.
Troubleshooter, which he An unexpected and very interesting use, currently being
investigated together with Novotek, of Proficy is the
discovered at a seminar. possibility to simulate a production process in a simulator.

It was his opinion that it was This can be used for training our operators before they
control the actual process. The “proof of concept” has been
exactly something like that, we
needed in our department.”
“There is no doubt that this software has Rockwool, A Global Leader
— Mogens Heine Larsen, a quick return on investment due to better
Section Manager, Process & Quality Service,
The Group is amongst the global leaders within
analysis and optimization of processes.”
Rockwool International A/S the insulation industry. Together with other
— Mogens Heine Larsen building-related products such as acoustic
Our division is called Process & Quality Service, POP, and ceilings, cladding boards and consultancy
consists of eleven highly experienced process experts, all
business, the Group ensures energy efficient
working for Rockwool International in Hedehusene. Half of
and fire-safe buildings with good acoustics and
them are based at our factories around Europe, travelling to completed and we will continue by presenting the idea
to the Directors who have the overall responsibility for a comfortable indoor climate. We create green
our various production lines in the world. The experts’ tasks
include helping our subsidiaries to solve problems with raw training in correlation to introduction of new factories and solutions for the horticultural industry, inventive
materials, processes and quality controls. Our consultants production lines. special fibres for industrial use, effective
are all highly experienced, almost as if they have a built-in insulation for the process industry and marine
Proficy CSense, which decreases the daily needs. However, Solutions: and offshore as well as noise and vibration
there is a need for new analytical tools that can provide Process optimization systems for modern infrastructure.
new insights and help to “adjust myths”. The tool should not
require deeper mathematical insights or to be difficult to use. Benefits: Better analysis
We have now found all that we look for in Proficy CSense's and optimization of
Troubleshooter, working with GE Digital partner, Novotek.
processes
The software has paid for itself
We have chosen a setting where Proficy CSense's Troubleshooter is on a server, accessible
to all consultants. However, only one user at a time can access the information, on the one
license available. Getting all data was challenging at first, but once we got everything in
order; it was easy to transfer between our systems. There is no doubt that the software has
paid for itself as we got better analysis and optimization from our processes.

The possibility to analyze data from production has evolved from

simple trending to a complete overview of the significance and
correlation of these data. Proficy CSense is a tool that through the
application of advanced mathematical algorithms, at the same time
being easy for the user to use, gives an insight into the process that
previously has not been available.

Mogens Heine Larsen, Section

Manager, Process & Quality Service,
Rockwool International A/S
Use your industrial data in an easy way
Proficy CSense enables engineers and operators to use the data
from production to avoid mistakes and create simulations.
The result is knowledge of how and what is needed to improve the quality and yield of the production process,
by using historical data, hence knowledge, intelligently to make it possible to identify the causes of the
problems and variations in the process.

Prevent Recurring Problems

With Proficy CSense, you can now prevent future problems from happening by using modeling and thereby the
ability to predict how the process will be developed. Proficy CSense can be used in wholesale, manufacturing
and continuous process.

Numerous Usages
With the use of Proficy CSense and Troubleshooter, you can visualize process problems and their causes
through modeling and simulation based on historical data. Proficy CSense becomes a digital Process Twin
facilitating process intelligence and optimization. It requires knowledge of the actual process, but no knowledge
about modeling, PCA etc., as this is not visible to the user.

Proficy CSense works with most historians on the market. With Proficy Troubleshooter you can: After the collected data is transformed into intelligent
Even data from relational databases and text files can be • Identify the likely causes of process problems and variations knowledge, it is used to find causes of process problems and
used. This gives the users more data to work with in the variations, and to take real time corrective actions.
multivariate analysis of historical data. • Identify process rules automatically based on historical data

• Identify correlations between variables

• Estimate the benefits of an online real-time solution

• Build and simulate the real-time solution

Decision support
for operators
Profice CSense can, by using the models, provide feedback in
real time on process events and thereby avoid discrepancies
and errors. You can, step by step, expand the decision support
for operators—directly integrated in the iFIX and CIMPLICITY
HMI/SCADA screens or other HMI/SCADA software. As
these decisions and guidance have proven its value, you can
continue to expand with real-time control.

With Proficy CSense, we can ensure that operators and

process engineers get immediate insight into the process
status, the rules that drives the process in its current status,
as well as identifying the process values that have the
greatest impact right now. This is possible through a specially
developed real-time component and provides a unique
opportunity for operators and process engineers to constantly
be alert and to avoid errors. The loop has been closed and we
have now enhanced the intelligent HMI/SCADA system.
Impala Platinum reduces process variation
by 40% with Proficy CSense

Background
Impala Platinum’s Base Metals Refinery (BMR) in Springs, South Africa, receives
its raw material from the company’s mining, concentrating, smelting, and
converting facilities in nearby Rustenburg.

The BMR then removes as much of the base metals as possible and sends the
Platinum Group Metals (PGM) concentrate to the Precious Metals Refinery
(PMR) plant for further processing. The base metals are refined and sold
separately to maximize the conversion of raw material into revenue.
The Challenges
Poor pH control, complex,
multi-variable environment

Impala Platinum identified difficulties in controlling

the pH in the first stage autoclaves, and poor pH
control has a serious and negative impact on the PMR
processes. It had to decrease the base metals content
(thereby increasing the precious metal content in the
concentrate), which could be done by controlling the
pH and consequently improving the nickel and iron
extraction efficiencies.

Up to this point, the pH had been controlled through a

cumbersome process of manual sampling and operator
intervention. Responses to changes were delayed and
future values could not be predicted automatically.

As shown in Figure 1, the first stage leaching process

involved operators having to deal with numerous
interdependent variables (pH, temperature, levels,
pressures, etc.).

Figure 1:
An overview of the process
 With operators having to juggle all these variables, it was easy for temperature and pH
The processing of profiles, for example, to go beyond their set limits. This affected Impala’s ability to remove
as much as possible of the nickel from the solids, and it degraded the quality of all products
platinum is a complex downstream.

and costly affair. The Another important metal to remove (leach) in this process was iron, which had to be kept in
solution. To do that, active pH control needed to be applied to the autoclave. Iron should be
quality of the upstream discarded with the nickel solution otherwise it would contaminate the PGM solids, which
would impact downstream processes negatively and can even lead to costly reworking of
process known as the concentrate.

Base Metal Refining In addition, ore grade is gradually dropping and there are large variations in concentrations
in the matte supplied by the smelter. All of this contributes to an environment where
(BMR) can have serious manual process optimization is extremely challenging.

financial consequences
on the profitability
of the downstream
Precious Metals
Refining (PMR) process.
The Solution
Advanced pH controller provides real-time optimization
Impala Platinum turned to GE to help address its challenges. Over a period of six weeks, GE installed its
Proficy CSense solution, providing Impala BMR with control and optimization of its nickel-leaching process.
The software relieved operators of many decisions that were previously difficult given the complex, variable,
and real-time nature of the processes in their charge. “Some systems are just
too complex and have
too many variables
An advanced controller based on a hybrid of analytics technology and
fundamental principles was implemented to provide: for operators to run
• Stabilization of the pH to improve the leaching efficiency of iron and nickel. This provides a
optimally. That’s when
better quality of PGM concentrate for PMR and other processes. we need the help of
• On-line and real-time pH sampling. This provides the degree of control required to help predictive solutions
optimize the process. Another consideration was that pH cannot be measured inside the
autoclave, so a special sampling pot was designed by GE to address this sampling problem.
that can deal with the
• Simplification of the control approach by manipulating acid addition to compensate for
complexities of our
pulp density and product feed rate changes, as well as changes in pH measurements, while unique problems.”
taking into account that high-pressure acid leach has complex reaction dynamics.
— Tim Spandiel,
Impala BMR Manager
 GE’s Proficy CSense solution helps to increase
process efficiency
Since the solution was implemented, it has greatly reduced the level of impurities in the product from the BMR, thus increasing
the concentration of PGMs. This ultimately leads to more platinum being extracted in the PMR downstream.

Sello Semosa, Impala’s BMR Manager, said that operators have embraced the solution, which has led to increased productivity.
She explained, “With Proficy CSense, I know I have a dedicated solution that is running real time with no disruptions. It has
freed up time for me and my staff, and time is important in terms of production.” Enabling effective information sharing across
the company, the solution automatically provides performance reports. This enables management to have a near real-time
view of the efficiency of the BMR nickel leaching process on which so many other processes rely.
“Although these
numbers may seem Solution results:
small, they are
extremely significant 40% 0.5% 3.3% 1.4%
improvement increase increase increase
to the financial in pH stability whereby in nickel extraction in iron extraction increase in the PGM
benefit of all our variation was reduced efficiency efficiency grade (concentration)
from 1.2 to 0.7
refining processes.”
— Tim Spandiel,
Impala BMR Manager Other benefits included:
• Reduction in pH peaks in the autoclave, which can lead to certain elements becoming very difficult to leach.
These elements go right through the process and end up contaminating the PGM solids, with the result that
the entire batch has to be recycled through a lengthy and costly processing pipeline.

• Acceptance of the system by operating staff

• Stable process control in spite of variations

• Indirectly limits environmental emissions

Lonmin increases smelter throughput
by 10% while improving recovery
When platinum prices were at an all-time high, Lonmin
wanted to maximize efficiency and unlock any hidden capacity
in its process. GE previously helped the Lonmin concentrating
section to solve some tough challenges, so GE was called
in again to help with a bottleneck in the smelter area. This
would prove to be the start of a new partnership in which
GE would help Lonmin with its continuous-improvement
journey through optimization and operational transparency.

10% increase in throughput in the

filter and drying section

Lonmin is the third-largest producer

of PGM (Platinum Group Metals) in the
25 % decrease in variation in
feed into the slag mill

world. The company’s operations are

located in the Bushveld Complex near
Rustenberg in South Africa. 45% decrease in variation in
cyclone feed pressure
Solution Overview
The mine performance solution
using GE Digital's Proficy CSense
analytics software was initially
implemented at the filter and
drying section of the Lonmin
smelter operation to help remove
bottlenecks and increase efficiency.
Results were so positive that
the solution was expanded to
include several other sections in
the smelter, and it was further
expanded to include monitoring of
the control systems.

As a next step, Lonmin plans to utilize the

predictive analytic capabilities of Proficy
CSense to increase uptime and decrease
maintenance costs on its fans and blowers.
Filter and The first section where the Proficy CSense
solution was implemented was the filter and
After GE implemented the process optimization
solution on this circuit, throughput increased
by more than 10%, consistently reaching
drying section
drying process.
maximum design capacity.
At the time, concentrate could not be dried
Temperatures stabilized and costly trips were
fast enough to meet demand from the
eliminated. Even in recent times with depressed
furnaces.
metal prices, when throughput is not the main
Operators struggled to juggle the multiple priority, Lonmin continuously uses the solution
interacting process variables. Excessive to ensure that equipment damage and costly
process instability due to varying filter cake inefficiencies do not occur.
feed and moisture content in the feed, as well
as frequent trips were encountered, increasing
the wear-and tear on the equipment. A holistic
process control and optimization approach was
needed that could not be provided by normal
regulatory control.

Slag concentrator Another problem area in the smelter was the

slag plant, where material from the furnaces are
variation in the cyclone feed pressure. Overflows
are now a thing of the past.
concentrated and recycled back into the process.
The overall impact on the stability of the
Design limitations in this circuit caused concentrator ultimately led to greater
major instabilities in the milling section, recovery, which is up 1.5%.
which impacted the whole concentrator.
Although other factors also played a role, the
Spillages were frequent events, which further process-optimization module was a major
wasted production time. contributor. Every percent extra recovery in the
smelter area has a direct impact on Lonmin’s
GE’s Proficy CSense stabilized the in-mill
bottom line—saving millions of dollars in metal
density and float feed rate and ensured optimal
that would otherwise have been lost.
cyclone operation. A performance evaluation
post-implementation showed a 25% decrease
in variation in the feed and a 45% decrease in
Off-gas handling The third area where process optimization
was applied was the gas cleaning plant,
The advanced controller stabilized the
pH control in the absorber, which brought

plant
where the offgas from the furnaces and the emissions back into range. The GE process
converters is treated. experts also made some recommendations
during the investigation phase that improved
At this stage, Lonmin had some issues with
overall stability of the plant.
sulfur dioxide emissions, which exceeded
the allowable limit.
The process dynamics were complicated,
and there was large variability in the sulfur
dioxide concentrations in the feed that the
conventional controllers could not handle. This
caused large disturbances in the process, which
led to high emissions.

Process monitoring Lonmin’s smelter also uses the monitoring tools

and services to optimize the performance of
The Lonmin plant has seen unprecedented
stability in the years since monitoring
its base-layer (PID) control. The automation began.
manager and his staff can immediately see
where the control inefficiencies lie and apply
resources where it matters most. They are
managing the process proactively, receiving
early warnings of changing process and asset
conditions. GE process experts also have weekly
calls with Lonmin control experts, augmenting
the team onsite with GE expertise developed by
monitoring a multitude of plants and processes.
“Without the GE solution, we would incur
additional cost for inefficiencies and we will
definitely have equipment damage due to our
inability to control the process in the same way
as an analytically-driven system.”
— Percy French
Automation Manager, Smelter, Lonmin
Success on many levels
Although there were some normal change-management
challenges with the real-time optimization modules, the
software now is an integral part of the Lonmin smelter
operation, with operators being heavily reliant on the solution
to run the process optimally. The solution has been able to
provide significant value throughout the Lonmin smelter:

10% 25% 45%

increase in throughput decrease in variation in decrease in variation in
in the filter and drying feed in the slag mill cyclone feed pressure
section

OPTIMAL

Geater recovery in the Stabilization of pH Optimization of

slag plant control in the absorber in performance of baselayer
the gas cleaning plant (PID) control
Plans for continued
improvement
According to Percy French, Automation Manager at the smelter at Lonmin,
Proficy CSense has helped him to “highlight more and more potential areas
for improvement.” As a next step, he plans to implement the predictive
monitoring capability of the solution on some of the smelter’s critical
equipment, including fans and blowers, to help reduce failures and prioritize
maintenance.

“The software is running and the people are

used to it, it’s very difficult to manage
without it.”

— Percy French
Automation Manager, Smelter, Lonmin
About GE
GE (NYSE: GE) is the world’s Digital Industrial Company, transforming industry with software-defined machines
and solutions that are connected, responsive and predictive. GE is organized around a global exchange of
knowledge, the “GE Store,” through which each business shares and accesses the same technology, markets,
structure and intellect. Each invention further fuels innovation and application across our industrial sectors.
With people, services, technology and scale, GE delivers better outcomes for customers by speaking the
language of industry.

Contact Information

www.ge.com/digital

©2021 General Electric. All rights reserved. *Trademark of General Electric. All other brands or names are property of their respective holders.
Specifications are subject to change without notice. 09 2021