Getting ‘Intelligent’ about Operating Chemical Plants

Finding and creating value from digital transformation in an existing chemical company is like trying to find a great white whale: it’s anxiously searched for, dimly perceived, enormous, yet elusive. But why is that?

Unlike a “Digital First” mindset in a new venture, digital transformation in an existing company is a very different challenge. More so, digital transformation in chemical companies often means overhauling legacy systems that have been the backbones for decades. It requires a delicate balance between maintaining operational integrity and embracing innovation. Moreover, the culture within these traditionally conservative entities must shift to one that values agility and continuous learning. We must acknowledge that this journey is a marathon, not a sprint – patience and persistence are valuable virtues. Chemical manufacturing companies that see real increases in productivity are not simply grafting technology onto existing work processes, in the name of innovation. Instead, they recognize that technology has changed fundamental assumptions about what operations can (and should) achieve, in much the same way that lean manufacturing and agile operating models challenged how leaders thought about waste, variability, and flexibility. Even at the highest levels of corporate performance, businesses that have successfully brought intelligent technology into their operations often exist in states of continuous experimentation and iterative improvements, simply to keep pace. Digital Twins and Artificial Intelligence is proving to be the unlock for many, as it helps automate digital transformation and reduces human intervention. Consequently, they must also prioritize data security while pushing boundaries; after all, they operate critical infrastructure.

Role of Artificial Intelligence in becoming Intelligent about Chemical Manufacturing

?Artificial intelligence (AI) and Machine Learning (ML) are revolutionizing market sectors from aerospace to oil and gas to chemical manufacturing. As more industries integrate AI and ML into day-to-day operations, there is increasing evidence of how these technologies improve and enhance systems. For example, AI and ML improve safety, increase efficiency, profitability, optimize supply chain management, enable proactive maintenance, enhance quality control, and simplify regulatory compliance. Gains made possible via AI and ML are delivering value across the board, but they are particularly significant in industries that are resource constrained, as well as those that are finding it increasingly difficult to attract and maintain skilled workers.

?Capitalizing on digital twin technology

Digital twins are versatile tools for improving supply chain complexities, process design, control, and automation as well as mechanical integrity. Designed for flexibility, a digital twin can be data driven, physics based, or a combination of both and is applicable in multiple use cases. A digital twin provides a virtual environment for designing and validating large process plants via scenario analysis in steady state and dynamic conditions. Using a digital twin expedites the design process, generating better results and eliminating the need to spend time working with complex spreadsheets. In the same way, a digital twin can be applied to processes and control systems. Decoupling hardware and software engineering, a digital twin enables parallel testing in a safe, secure environment that facilitates late-stage configuration changes without extensive reengineering.

By using curated, real-time data to predict process behavior, a digital twin also delivers insights for adjusting operator actions to enhance throughput, always ensuring the best performance for ongoing plant-wide optimization.

Finally, a digital twin is an excellent training tool, permitting operators to hone their skills in a virtual environment that does not require working equipment. This tremendously fast-tracks competency and lowers the learning curve for reaching fully qualified operator status by exposing workers to abnormal situations and allowing them to resolve issues virtually before encountering them in a working plant.

Transforming chemical process facilities

In the chemical process industry, AI and ML are making operations safer, more sustainable, and more efficient. Pattern recognition is currently one of the most prevalent intelligent technologies that is improving safety. ML tools like clustering on real-time process and alarm data, make it possible to identify performance patterns that can lead to abnormal - and sometimes dangerous - conditions. Using heuristics and an ML-based classification algorithm to diagnose faults based on real-time symptoms and alarms can improve asset reliability, making it easier and faster to correct malfunctions.

Leak detection is another prevalent use case where intelligent tools are being employed in chemical process facilities. In facilities around the world, AI is applied to classify thousands of pixels of hyperspectral images to detect and identify gas leaks and issue alerts, when concentrations approach dangerous levels.

Simultaneously, efficiency may be boosted with the convergence of physics-based and data-driven ML models that might predict quality of product streams, in the reactor, that otherwise would be dependent on intermittent lab readings. Not knowing the quality of product streams can mask signs of runaway reactions or the formation of dangerous byproducts, which have repercussions for safety, efficiency, and profitability. ML can help identify the occurrence of incomplete reactions or a prevalence of side reactions; either of which can negatively impact production yields. With the ability to recognize conditions that can affect quality, operators and engineers can quickly take corrective action, making decisions based on an approximation of conditions instead of waiting for time consuming lab analysis results.

Processes also are being optimized using hybrid models—ensembles of first principles physics-based models and ML models—to identify multi-variability in a complex unit operation with intermittent nonlinearity to execute predictive control via a model-based controller.

In addition to streamlining processes, advanced technology may enable plants to allay off-spec production to not only reduce material waste and rework, but also help identify situations that could lead to inferior products that could potentially need to be recalled.

Expanding Intelligent Operations

Uptake of Intelligent Operations is occurring rapidly in chemical plant operations, and more applications will be in place in the next year as the proliferation of powerful edge devices and AI/ML deployment accelerates. The visual inspection processes may be improved by the convergence of computer vision, drone technology and AI, which can be particularly impactful in confined spaces where inspection and monitoring are extremely dangerous for humans.

The use of ML algorithms in conjunction with knowledge graphs is likely to expand soon. That is because of the ability of knowledge graphs to help machines reason the information stored and understand connections among pieces of data. Used together, they autonomously contextualize data for root cause analysis—for example, to identify causes for variances in key performance indicators—generate actionable insights and prioritize major causes for future reference.

Historical plant performance information can be used with heuristics and sophisticated ML to predict future performance and guide operators through difficult operating conditions for better results. ML algorithms may simplify emissions management to calculate emissions on the fly at the equipment level in real time by triangulating several types of discrete data.

Generative AI Advancements

Advancements over traditional natural language processing (NLP), which relied on keyboard matching and had difficulty grasping emotions and language subtleties, are changing the scope of NLP applications. Employing large language models (LLM) with sentiment analysis alters the landscape for applying generative AI (GenAI) in the chemical processing industry.

Today, LLM-enhanced NLP is driving some advisory service-based applications. Using GenAI to provide a conversational user interface in the operator station via a feature like Apple’s digital assistant, Siri, improves the operator’s ability to react quickly to irregular situations. And when corrective action is taken, GenAI (with support from other systems) can generate an unbiased report, providing a clearer picture of the event. This capability means workers may devote time to other activities instead of collating papers and organizing data to generate a traditional shift report.

NLP is also being used as an intelligent search tool, saving thousands of man-hours for operators and engineers who otherwise would be spending time analyzing information to improve decision-making. A GenAI algorithm can be trained on multiple documents and systems to find answers to questions that historically required significant human input. This capability allows functional teams to quickly access information and coordinate activities with the plant.

Lastly, GenAI, coupled with reinforced learning (which focuses on training AI agents to make decisions and take actions in an environment to achieve a specific goal), is driving workflows that used to require a human to make decisions, ensuring human-like factors like culture and emotion are captured in the model as it learns from predictions.

Overcoming obstacles

Implementation of digital transformation programs using Intelligent and Autonomous applications in multiple industries has revealed several hurdles to deploying and scaling large initiatives, especially when they include AI and ML, which require a large volume and variety of data. It is extremely critical to ensure the data is clean and accessible at the right time with the correct time stamp, and data must be contextualized to allow actionable insights. The speed with which data is transferred is another consideration because signal latency—even if the delay is only a matter of milliseconds—can have repercussions.

And data security continues to be a concern. Most communication tools used in commercial environments are not appropriate in an industrial context. Technology providers must demonstrate that their solutions deliver the security necessary to safeguard data, which is the backbone of digital transformation initiatives.

For some organizations, bias and causality are the biggest hurdles to overcome in deploying AI-based solutions because prediction models are often considered a “black box.” It is extremely difficult to demystify the process and ascertain that there was no bias or overfitting of the model.

All these issues are amplified when a company begins scaling its digital solutions because of the inherent complexity of change management and adoption. In fact, the biggest challenge in executing a digital transformation is not technical difficulties but issues with change management and executive engagement. Success depends on establishing key criteria, setting adoption milestones, driving a change program, and ensuring regular progress measurement and reporting.

The face of digital transformation in near term

Chemical process industries will continue to see development to expedite digital transformation. One of these is virtualization and cloud computing. As more analytics and AI algorithms are deployed in digital twins, there may be an architectural tradeoff among physical, virtual and cloud computing risks and opportunities. Although virtualization of hardware has become a norm, some companies are not yet comfortable with cloud computing for certain mission-critical applications. Analytics on the edge is another area that may receive more attention. More smart edge devices are expected to be deployed with powerful analytics to address some of the latency issues of deploying mission critical applications on the cloud. More virtual and cloud adoption are anticipated in the medium term as trust is developed and latency issues are addressed.

Cybersecurity is a constant concern, and while open systems and cloud infrastructure introduce operational advantages, they also carry the risk of malicious cyberattacks. That has led many companies to look for help in securing their infrastructure by designing, deploying, and testing industrial grade cyber secure hardware and software, a trend that is likely to increase at a very fast pace.

As the use of digital twins expands, more segments of the plant and more processes may be integrated, and as that happens, communication among digital twins will enable even greater efficiencies and economies. For this to occur, there should be a convergence of individual technologies that can be used in combination to create an immersive three-dimensional virtual or virtual/physical industrial environment. This “industrial metaverse” of digital twins may be able to autonomously orchestrate many workflows that are currently prone to human errors but also has the potential to unlock value from digital twin technology that currently is unexplored.

Future Applications

The emerging space of AI-enabled Advanced Robotics may be the next frontier to enable Remote, Autonomous and Intelligent Operations. Today Robotics is applied primarily in inspection and monitoring but will expand to other functions. Autonomous Mobile Robots are being tested for use in more complex tasks, including operations and maintenance rounds.

Moreover, as OpenAI gains a firm foothold, there will be a proliferation of generative AI-based applications in the industrial world that can revolutionize the industry by accelerating material discovery, improving sustainability, enhancing product development, and expediting innovations in material science, streamlining process optimization, and reducing costs and time to market.

In conclusion, intersection of deep process knowledge, ubiquitous sensing, virtual technologies and Artificial Intelligence is likely to intrinsically change the trajectory of digital transformation in chemical process industries, by providing operators and engineers with tools that enhance and accelerate the pace of advancement and significantly impact safety, reliability, efficiency, sustainability, and agility of their supply chains.

Written By

Tathagata Basu, BE (Chem Eng.), MS ?(Data Analytics), MBA

Global General Manager, Sustainable Fuels & Chemicals

Honeywell Process Solutions

?

PS: As a citizen data scientist, the views expressed here are my own and do not necessarily reflect the views of my employer