The average modern industrial plant uses less than 27% of the data it generates, according to industry experts at the ARC Advisory Group, Boston.
Typically, the remaining 73% of data – much of it produced by plant process-control systems as high-frequency operational control (OT) data – is seldom used.
Large volumes of other valuable functional data reside in a company’s general business or IT systems, and still more in the engineering systems (ET), covering specific design information for various assets. In addition to being rarely used, all this data is normally scattered about in separate silos and networks that support little or no cross-referencing.
“That is where the golden opportunity lies which we can now unlock with new software platforms that simplify better convergence and analysis of OT/IT/ET data,” says Charles Blackbeard, the business development manager for ABB’s Ability Digital. The benefits can be impressive, such as higher production rates from existing assets, less downtime because of predictive-maintenance practices, safer operation, reduced energy and other raw material inputs, as well as lower environmental impact.
Improved convergence of OT/IT/ET data means bringing together previously separate elements, which have now been streamlined and integrated. To achieve this, all OT, IT and ET data is accumulated in a data lake. Next, related data is contextualised and stored in an industry-specific data model, such as paper making or plastic extruding. Then advanced analytics and industrial artificial intelligence (AI) algorithms are applied to identify correlations not previously visible.
“Industrial AI can play a major role in identifying these patterns and making process predictions,” says Blackbeard. The terms AI and ML (machine learning) are often used interchangeably, which can be confusing at times. AI is the overarching science of making machines and physical systems smarter by embedding ‘artificial intelligence’ in them. ML is a subset of AI that involves systems gaining knowledge over time through ‘self-learning’ to become smarter and more predictable, without human intervention.
As an example, consider a motor, an essential and omnipresent asset in any plant. The motor generates a lot of operational data such as temperature, pressure and flow rate data from various stages of the production process. To acquire a holistic overview of the motor, we integrate information from all these systems and store the relevant pieces in a contextualised data model. This allows us to visualise and activate optimum equipment operation for the best overall process results, explains Blackbeard.
In a large plant, there can be hundreds of such assets performing many functions and running under different operating conditions with varied design parameters, all with data stored in various systems. Widespread OT/IT/ET integration and contextualisation is therefore critical to obtain a complete view of the plant and carry out valuable analytical tasks that improve operations, asset integrity and performance management, safety, sustainability, and supply chain functions. What emerges are patterns that accurately predict future behaviour, allowing improved process performance.
“We have been using AI/ML to deliver a higher degree of prediction accuracy and optimisation to operations, processes and assets. Combining AI with deep industrial domain expertise empowers operators to run their industrial processes safely, more effectively, and more sustainably,” notes Blackbeard.
He adds that there are several barriers, perceived and otherwise, that hinder the implementation of advanced analytics. The most common reason for hesitation is the perceived complexity. People mistakenly think it is much more difficult to achieve than it is. Another explanation is the incorrect belief that, to use big data, you must make massive capital expenditures, because it is an ‘all or nothing’ undertaking.
“But it is not. You can start with small steps,” points out Blackbeard. Other reasons might be lack of cooperation between OT, IT, and ET people, and just generally slow adoption of new digital tools in many industrial sectors. The fact is that it is easy to join this digital maturity journey, no matter where you are, using data and signals that are already available in your process control, business, and engineering systems,” concludes Blackbeard.