Although conditions in the process industry are changing undeniably fast and competitive pressure is increasing, adaptive systems offer an opportunity to make production processes more efficient, explains Andreas Eschbach, CEO and founder of eschbach
Software designed for plant process management using artificial intelligence (AI) is steadily gaining recognition as a way to help manufacturers meet these changes. Integrating AI and tribal knowledge into enterprise platforms can provide a powerful assistant in 24/7 operations.
Companies in the process industry must respond to rapidly changing conditions. Natural gas prices show, for example, the enormous pressure to which the process industry is exposed and, at the same time, how volatile and unstable the framework conditions have become.
In addition, complexity in production has increased steadily in recent years owing to constantly changing market requirements and technological innovations. Competitive pressures present process engineers and operations management decision makers with uncertainties and new challenges.
For production teams, this means that processes must be optimally controlled and malfunctions must be eliminated as quickly as possible. After all, unplanned downtime is very costly. At the same time, every shift team knows how difficult it can be to make the right decisions or how time-consuming it can be to resolve incidents.
In many cases, similar incidents or conditions have occurred in the past … and corresponding solutions to those problems have been documented. Having existing knowledge, operational incidents can be avoided or quickly eliminated. But when problems arise in production, employees lack the time to sift through written documentation and records.
Naturally, it’s frustration when digitally documented solutions are available but cannot be found. More seriously, if the workforce is unable to quickly access process-relevant information, the efficiency of the plant declines with every passing minute — and time and revenue are lost as a result. Adaptive systems that learn from past data and provide solutions are therefore in demand.
A focus on human-centric AI
With the goal of making plant process management (PPM) sustainably more efficient, powerful systems, machines and AI are becoming increasingly important in production. But the digital transformation in the process industry must centre on people. Machines or algorithms cannot react to changes and take responsibility for the safety of processes and plant; only people can do this.
The goal is to create a process landscape in which humans remain in control while being supported in the best possible way by machines, digital communication and cognitive assistance systems: machine-assisted humans.
The future of the process industry requires scenarios in which machines and AI support employees so that maximum efficiencies (and safety) can be achieved in 24/7 shift operations. AI acts like an assistant at the side of production workers. Ultimately, decisions in production facilities are made by humans … but can be supported by AI.
A case in point
To create a decision-making process that takes a holistic view of process disturbances, a well-thought-out algorithm is needed. Nowadays, it’s simply no longer sufficient to rely on simple “IF–THEN” programming. The amount of data, the flow of information and the process landscape are too complex for this.
When a piece of equipment fails, it’s common for shift teams to perform standard reset and troubleshooting procedures to no avail. When more experienced production staff arrive, their knowledge provides context: “The last time it was freezing outside, the liquid line partially froze. This changed the preproduct, which eventually caused the line to stop.”
This example illustrates that human language can decisively provide context in the operational process.
The value of smart searches
There are now enterprise platforms on the market that integrate the newest AI-supported features to precisely address this challenge. AI capabilities such as Smart Search provide production teams with process-relevant knowledge in a user-friendly format — and as spontaneously as needed.
The aim of the AI function is to provide employees with the best possible support in the form of relevant information and suggested solutions from the past.
A major challenge is that two different workers’ descriptions of the chain of events described earlier would look completely different.
If we now consider that thousands of entries occur in operational communication, the problem becomes clear.
With conventional options, searching for an existing event is like searching for a needle in a haystack. This is when AI algorithms excel at finding solution descriptions. It also allows shift personnel to more efficiently and, most importantly, more quickly use historical knowledge to make informed decisions, thereby avoiding or rapidly resolving disruptions.
In classical programming, the human defines the algorithm and tells it exactly what criteria should be used to perform actions. By contrast, AI can learn these criteria itself by recognising patterns in the data. The algorithms can then provide predictions of estimated outcomes with certain probabilities, whereas humans provide the framework for the data search.
The algorithm adjusts the search criteria accordingly based on perceived intelligence. From this, correlations in the data can be determined faster than a human can work out that logic. AI provides production teams with desired results and picks up on a variety of details in addition to the problem. In this way, AI enables well-founded approaches to solutions.
Enhancing interactive operational communication
Data is a fundamental prerequisite of AI. This is important both in terms of quantity and quality. Only when sufficiently good data is available can AI algorithms be meaningfully trained.
Currently available enterprise platforms have been developed specifically for the process industry and enable a solid database: in addition to plant performance, work orders and incidents are displayed transparently in digital overviews. In this way, interactive operational communication in shift operations is ensured.
At the same time, shift logs are recorded in an audit-compliant manner and can be retrieved at any time. The data basis — the documented wealth of experience of the employees — expands with each new entry and even can be used across different locations. In this way, the real-time information helps to further increase efficiency and safety in plant operation.
Plant malfunctions often cannot be rectified during the ongoing shift but need to be communicated to the following team. Here, too, the advantages of multifunctional enterprise platforms are evident.
Shift handovers can now be used efficiently without process-relevant knowledge being lost. Smart Search capabilities show the production teams which solution approaches are target-oriented for certain malfunctions and which are not.
Delivering AI-enabled insights
Suggested solutions or recommendations are only part of the potential of Smart Search capabilities. In addition to filtering stored information, Smart Search can process communications by cataloguing and indexing topics, keywords, phrases and more.
In this way, Smart Search provides insights into the overall picture of the situation and, in the event of disruptions, delivers solutions that have already been proven to work. Ultimately, standardised processes and workflows are essential in the process industry; however, different processes, languages and even communication capabilities create enormous diversity in the process landscape.
When the previous programming is no longer able to handle information inflows that do not conform to a precise format with prescribed form fields for data entry, new and smarter technologies are taking hold. The overarching goal is to take human-to-human communication and knowledge management to a new level.