You are on page 1of 2

SMART PLANTS 1

Five steps to becoming a


fully-realised smart plant
In pursuit of Industry 4.0, more and more companies are gearing up to transform their
plants into smart plants. Smart plants combine physical processes, digital technologies,
machine learning and big data to create an optimally efficient and profitable system. While
transforming a facility into a smart plant is a complex process, this transformation can
easily be achieved with the right approach.
n by Martin Provencher, OSIsoft, USA

S mart plant technology has the power


to transform every aspect of an
industry. It can improve communications
Industry 4.0: the digital transformation journey

for remote monitoring, support health and


safety goals, improve energy efficiency,
and increase asset reliability – all of which
lead to smoother operations, lower costs
and higher profits.
Many companies struggle to see a big-
picture view of their plant because a plant
is not a single system – it is many different
assets and machines. Often these assets
have different control systems, from PLC
to SCADA and more, operating completely
independently of each other. For a plant
to function as a whole, integrated system,
all assets must first feed their data into performance baseline. A baseline can be Step 4: predict future anomalies
a single real-time operational data the expected performance of a machine Now that the company can see what is
structure, such as the PI System. Once the according to its operations manual or happening in real-time, engineers can
operational data has been collected, there data from previous years. This baseline begin to predict future anomalies. First,
are five vital steps that a company must comparison can help engineers identify all the data available must be modelled
follow to unleash the power of this data KPIs and create reports that reflect and analysed: equipment data, process
and transform the plant into an Industry how the plant is performing relative to data, energy consumption, maintenance
4.0-ready smart plant. expectations, and whether or not the plant status, staff availability and even weather
is meeting its goals. data. Analysing and modelling data can
Step 1: protect assets help create predictions about future
Once all of the company’s data is available Step 3: diagnose problems issues like equipment failures, process or
in a single place, operations staff, Next, the company can examine environmental deviations, and more. At
engineers and managers can identify operational data to identify the conditions this stage, staff are no longer performing
trends in real time to visualise what is truly that often result in failure or process the analysis – analytical tools can use
happening in the plant. These trends will deviations. When reliability and process historical failure data to create models that
allow individuals and teams to identify engineers analyse these conditions, they predict future problems.
abnormal situations and act on them to can start to understand why failures occur. However, for these predictions about
protect people, equipment and profit. They can begin to monitor asset conditions equipment failure or other production
Right from step one, this top-down view in real-time to spot deviations that might anomalies to be accurate, the operational
can save a company a tremendous amount indicate an approaching failure. It is at this data feeding these analytical tools must be
in damages and profit loss. stage that companies start to see a huge put into context. For example, sensor data
return on investment by avoiding serious can only be truly valuable when it is tied
Step 2: monitor performance problems before they even happen. This to a particular asset in a particular plant.
Now that a company has identified is also an important time to invest in data Asset Framework, the contextualisation
various operational trends in the data, visualisation tools such as PI Vision to layer of the PI System, can provide just this
this real-time data can be compared to a assist in this real-time analysis. sort of valuable context to a plant’s data.

© TRADESHIP PUBLICATIONS LTD, 2020 DECEMBER 2020 INTERNATIONAL CEMENT REVIEW


2 SMART PLANTS

Step 5: prescribe with machine “Data is the new gold. We the first time, operators were able to see
learning how their process was behaving and how
With a robust real-time data infrastructure had a gold mine sitting the AI model was telling them the system
and a full understanding of the conditions there and we were not was going to behave up to 25mins ahead
for success or failure for the plant as a of time.”
whole, it is time to implement an AI or tapping it.” Cemex opted to predict cooling
machine learning solution. Implementing performance within a 5-15min range to
machine learning is where a plant Rodrigo Javier Quintero, increase accuracy while giving operators
truly becomes smart. An AI or machine operations digital time to react to projections.
learning model can consider the entire Next, Cemex moved from a predictive
technologies manager, Cemex
picture: equipment data, financial goals to a prescriptive strategy. After organising
and information, parts availability, team the data in Asset Framework, operators
availability, weather conditions and more. used PI Vision to visualise a range of
AI and machine learning algorithms can Cemex’s cement production process and settings recommendations based on prior
then determine the best course of action is one of its most vital assets. This piece performance. During this phase, Cemex
to manage risks and optimise productivity of equipment is responsible for product relied heavily on operator expertise,
outcomes. quality and output but requires a massive allowing operators to offer direct feedback
For example, imagine a mining truck amount of energy to operate. As a vital and fine-tune the AI-recommended
is approaching its predicted maintenance but resource-heavy asset, the kiln was settings before deployment. Once it
period. The data confirms that the truck the natural starting point for deploying became clear that settings were on track, it
is ready for maintenance, but the staff or AI-driven optimisation technologies. was time to move to stage three: autopilot.
parts are not available to begin repairs. To start, Cemex isolated a small portion
The AI can determine how many more – the clinker cooling process – as a target Cooling on autopilot
days the truck can safely run and schedule for automation. “This is not a 100m dash. Having established confidence in the
a maintenance overhaul to coincide with This is a marathon,” Mr Quintero noted. AI-recommended settings, Cemex
parts and staff availability, keeping a plant During the clinker cooling process, cold deployed autopilot. Cemex modified
productive while ensuring safety and air is pushed into the kiln while heat from the OPC interface to accept PI System
reducing downtime. the raw material is recouped for process information. In conjunction with operator
efficiency. To optimise heat recovery, oversight, Petuum’s AI algorithms
Cemex leads the way Cemex operators must control the air input work together with PI System data
Few companies have demonstrated the cooling rates, all while ensuring process and the OPC interface to predict future
tremendous value of creating a smart stability and product quality. Though kiln performance. Those predictions
plant using this five-step process more Cemex operators are highly skilled, it is automatically feed proposed settings into
than Cemex. The company began its impossible to have the most-experienced the control system to optimise the cooling
journey towards becoming a smart plant team members available every day. process, promote energy efficiency and
with a decade of valuable data waiting Taking a phased approach, Cemex predict equipment failures – all while
to be analysed. It digitised operations opted to feed AI-driven recommendations maintaining quality standards. When
by installing equipment sensors and directly into the kiln control system via the autopilot is running, Cemex expects to
PI System technology in every facility, PI System to help all operators achieve improve energy usage by 2-7 per cent,
creating a stream of valuable real-time “golden run” capabilities. The company reduce process variability by half of the
data. “Data is the new gold. We had a partnered with Petuum, an AI solution standard deviation and increase overall
gold mine sitting there and we were not developer, for this project. yield by five per cent.
tapping it,” said Rodrigo Javier Quintero, Cemex unified its data into a single
operations digital technologies manager at From predictive to prescriptive real-time operational data management
Cemex, during the PI World San Francisco To test the possibility of achieving system, focussing on its most vital
conference in 2019. “golden run” status, Petuum and Cemex assets. The company then monitored
Cemex first tested the PI system on developed a “predict, prescribe, autosteer” performance and diagnosed issues in the
its rotary kiln. The kiln sits at the heart of plan. During the prediction phase, clinker cooling process, before moving
Petuum’s self-learning forward to predicting future anomalies and
Figure 1: phased development software analysed past using AI technology to prescribe automatic
performance data to settings for the kiln.
create output variables By following these five steps, Cemex has
that would predict cooler started down the path toward automating
behaviour. Operators processes and achieving previously
were then able to visualise impossible production goals with the help
projected behaviour in of Petuum and OSIsoft’s PI System. Since
PI Vision alongside real- Cemex already relies on the PI System in a
time performance data. number of its locations, the company now
“This was eye-opening plans to deploy AI technology alongside
for the kiln operators,” the PI System at plants in Europe and
said Mr Quintero. “For South America. n

INTERNATIONAL CEMENT REVIEW DECEMBER 2020

You might also like