Page 7 - EngineerIT July 2022

FRONT COVER STORY

External factors influence process change The challenges of an MPC project
In most companies, the reality of external influences on the day-to-day operations does MPC projects have many challenges.
not become apparent quickly, if ever. For instance, a 10% increase or reduction in the gold Unfortunately, most of these are only
price will typically not result in any change on the plant floor. The business will continue as reluctantly (if ever) accepted by business
usual. However, a long-term slump may lead to staff reductions as less profitable parts of leaders.
the mine are shut down. One of the reasons for this is that mines have one set of objectives
with a defined optimum operating envelope, without considering this external influence. These include (but are not limited to) the
Customers are looking for dynamic optimum operating envelopes that move when external following:
factors change. For instance, when the price is high, the envelope could move toward • There is no deep understanding of the
extending the life-of-mine, and when the price drops, the envelope could move to increase interrelationship between processes, process
efficiencies and cost reduction. variables, external influencing factors and
The effect of external influencers is not always understood on the plant floor. The operators’ actions.
optimum state can take many forms, depending on the strategy and state of the influencing • Most operators do not understand the
process variables at a specific time. The optimum state can change frequently and without correlation between external influencing
prior planning or warning. MPC can be used as a dynamic tool that makes provision for the factors, process variables and operator
changing nature of the process or changes to the business strategy and objectives. actions and only have a “gut feel” not
supported by data.
Where does machine learning fit in? • Lack of measurement data because the
Historical data, combined with deep process understanding and understanding of the actual instrumentation is “old school”, not digital,
cause/effect relationships, are used to create a model of the optimum operating envelope not connected, or not historised. This often
of a process. Machine learning tools, concepts and algorithms are extensively used during results in delays in collecting adequate data.
this process. This model is not only built to identify anomalies in process performance by • Instrument accuracy is a factor, but in many
comparing current operating conditions and variables with the variables as predicted by the cases, not as critical unless the instrument is
model, but is also used to process current process conditions through the model to predict changed. For example, suppose the model is
potential outcomes in the future. Based on this prediction, action can be taken to prevent built on first-principle design criteria. In that
adverse process conditions in the future. case, instrument accuracy will be a crucial
factor. Still, the effect will be less severe if the
An effective MPC implementation is… model is built using historical data, including
There are some critical prerequisites one needs to be aware of before implementation. an inaccurate instrument.
• Inconsistencies in process understanding and
The main requirements are: operator actions when process variations
• Deep process knowledge and understanding, from a first-principle and experiential occur. One of the biggest challenges is a
learning perspective. lack of understanding of the cause/effect
• Sufficient instrumentation and data points/variables in place for the machine or process relationships between process variables.
to build a relevant machine or process model.
• Sufficient historical data, ideally one year of data, but a minimum of six months, in order Customer value drives Iritron’s MPC
to understand the machine or process behaviour patterns. project methodology
• The ability to access the historical information to make it available to the MPC platform. Iritron follows a pragmatic approach to
assisting our clients. We believe value can be
Once these prerequisites are met, the design purpose of an MPC project can be fulfilled - derived even if everything is not “perfect” in
achieving rapid ROI. the client environment. We use engineering
principles in all we do, including analysis and
The benefits of an MPC project understanding of client operational processes,
When a process’s optimum operating envelope is understood, many benefits can be derived. before even contemplating implementing any
MPC model. We prefer more data points and
These include: better accuracy, but we live in an imperfect
• Reduction in process variability by keeping the process operating within the optimum world and provide for that. An imperfect model
envelope. that provides 60% of the potential benefits now
• Plant stability and fewer process fluctuations. is better than a perfect model that provides
• Managing the process within the optimum operating envelope considering internal and 100% of the potential benefits 24 months
external process influencing factors. after we have implemented the additional
• Operating closer to specifications and performance limits while maintaining safety instrumentation and collected adequate data.
margins. The future of business depends on increasing
• Increasing the lifespan of products or plant equipment. ROI from value-driven products and services.
• Reduction in maintenance costs, by either introducing a delay in planned maintenance The future of process control and automation
or planning equipment shut down, to prevent a critical equipment failure (even if the depends on model-predictive control supported
planned maintenance frequency indicates that the equipment should still be healthy). by AI and machine learning that continues to
• MPC helps to identify potential opportunities for improvement before implementation, shape the future of manufacturing, working
by simulating process performance. together in a complex dance. n

EngineerIT | July 2022 | 5