Using AI to accelerate process optimization:
Is your plant ready?
Sensor and process control upgrades are often necessary to enable the deployment of AI. A quantitative assessment of readiness can be an important first step.
By Nathan Flesher
With Aashay Jain, Otto van der Ende, and Raj Kumar Ray
Download the article
August 2024 | Infographic
AI can help increase process plant productivity, but many plant owners and operators are not yet ready to implement new technologies. A comprehensive digital-readiness assessment can help prioritize critical improvements and highlight further improvements to incorporate over time.
We published an article last year titled “AI: The next frontier of performance in industrial process plants,” and we have developed an approach to assess the digital and analytics readiness of process plants. This includes areas such as process control, advisory AI models, critical aspects of IT and operational technology, change management, and capability-building mindsets.
Process controls and automation: An overview
As explained in our article “The potential of advanced process controls in energy and materials,” the way process plants use control systems can be visualized as a pyramid. Advanced process controls (APCs) sit near the top. Base-layer controllers, such as programable logic controllers, proportional integrative derivatives, or distributed control systems,
sit below the APC and help execute set points. In this way, APCs facilitate the optimization of these systems, providing standalone strategies and helping to connect advisory systems at the top with the systems and subsystems below.
Advisory models
Capability and mindsets
Data
Supervisory controls (APC¹)
Base-layer controls (PLC,² PID,³ DCS⁴)
IT and operational technology
Change management
Sensors and instrumentation
Equipment
Agile
Are process plants ready for change?
All plants are at different levels of digital maturity. However, we have identified the following key insights from multiple digital-readiness assessments conducted across a variety of process plants and regions:
Advisory models
Only 10% of plants use AI to describe, predict, and inform process decisions; more plants could build data-driven advanced models to drive performance in complex systems
Supervisory controls
Only 34% of plants have installed APC¹ systems for crucial unit operations, and use of those APCs¹ is only 63%; more plants would benefit from installing APCs¹ and using them for better process control
Sensor and instrumentation
75% of plants have instrumentation in place that controls critical process variables (density, mill power, etc), and only 70% of these instruments are properly calibrated and catalogued
IT and operational technology
Only 18% of plants have dedicated IT teams that support deploying and scaling AI solutions; all plants would benefit by having a full-time dedicated team within the IT department to drive the deployment of AI solutions
Change management
60% of organizations have a change management strategy in place to enforce or embed
AI solutions in their current operations routine
Assessing the digital readiness of process plants
Many plant owners and operators can’t justify adding additional sensors or APCs because they don’t understand
the value they would create. At the same time, their data or process automation teams want to install sensors indiscriminately, without a plan to link to value creation.
This in mind, a qualitative assessment of readiness could be an important first step in deploying AI-based process optimization. To improve digital maturity for deploying AI, plant owners and operators can take the following five steps:
1
2
3
4
5
Increase sensor coverage, and focus on calibrating and cataloging installed instruments
Improve use of installed APCs,¹
and increase coverage of APCs¹ for crucial unit operations
Use advisory models to describe, predict, and inform process decisions
Build dedicated IT and operational technology teams for AI deployment and upskilling on agile, digital, and analytical skills
Incorporate a change management strategy to embed
AI solutions, and invest in capability building and talent development
Note: Advisory models are machine learning–based models that enable plants to identify and optimize operating recipes under different process conditions. Supervisory controls such as model predictive controls, fuzzy logic, and advanced process controllers supervise base-layer controls. Base-layer controls attempt to bring a manipulated variable to the set point at which it has been prescribed.
1 Advanced process controller.
2 Programmable logic controller.
3 Proportional integrative derivative.
4 Distributed control system.
Source: “The potential of advanced process controls in energy and materials,” McKinsey, November 23, 2020; “AI: The next frontier of performance in industrial processing plants,” McKinsey, September 19, 2023
ABOUT THE AUTHORS
Nathan Flesher is a partner in McKinsey’s San Francisco office, Aashay Jain and Raj Kumar Ray are consultants in the Gurugram office, and Otto van der Ende is an associate partner in the Brussels office.
The authors wish to thank Gervasio Briceno and Sean Buckley for their contributions to this article.
RELATED ARTICLES
Article
Al: The next frontier of performance in industrial processing plants
Article
Advanced tech elevates process modeling in batch-processing industries
Article
The potential of advanced process controls in energy and materials
Article
Al: The next frontier of performance in industrial processing plants
Article
Advanced tech elevates process modeling in batch-processing industries
Article
The potential of advanced process controls in energy and materials