The future begins with industrial AI

“Domain knowledge is the secret sauce that distinguishes industrial AI from more generic AI approaches. Industrial AI will drive innovation and efficiency improvements in capital-intensive industries in the coming years, ”said Willie K Chan, CEO of AspenTech. Chan was one of the original members of the MIT ASPEN research program that became AspenTech in 1981, now celebrating 40 years of innovation. .

Incorporating this domain knowledge allows industry AI applications to understand the context, internal operation, and dependency between complex industrial processes and assets, and design features, capacity limitations, and safety and regulatory guidelines are critical. world industrial operations.

More generic approaches to AI can lead to accurate correlations between industrial processes and equipment, leading to inaccurate accuracies. Generic AI models are trained in large volumes of plant data, which typically do not cover the full range of potential operations. Because the plant can work under very tight and limited conditions for safety or design reasons. As a result, these generic AI models cannot be extrapolated to respond to market changes or business opportunities. This exacerbates production barriers to AI initiatives in the industrial sector.

In contrast, industrial AI leverages the expertise of industrial processes and real-world engineering domains based on the first principles taken into account by the laws of physics and chemistry (e.g., mass balance, energy balance) as protection to mitigate risks and meet all requirements. safety, operating and environmental regulations. This achieves a safe, sustainable and comprehensive decision-making process, creating comprehensive results and reliable approaches in the long run.

Digitization of industrial facilities is key to achieving new levels of security, sustainability, and profitability, and industrial AI is a key driver of this transformation.

• A process plant can deploy an advanced class enabled with industrial AI Hybrid models, a deeper collaboration between domain experts and data scientists, using the first principles of machine learning and more comprehensive, accurate and high-performance models. These hybrid models can be used to properly design, operate, and maintain plant assets throughout their life cycles. As they are important for the longer term, they also provide a better representation of the plant.

• A chemical plant could use industrial AI to get real-time statistics from integrated industry data from the edge to the cloud. Artificial Intelligence of Things (AIoT) to enable light decisions throughout the organization. Using dynamic and richer workflows, supply chain and operations technologies are linked together to detect changes in market conditions and to respond automatically to the operating plan and schedule.

• A refinery can use industrial AI to simultaneously evaluate thousands of oil production scenarios in multiple data source sets to quickly identify optimal crude oil processing slates. Combined with AI-rich capabilities, company-wide perspectives, and integrated workflows to improve executive decision making. This approach empowers employees to perform more strategic and strategic tasks to focus their time and effort on employees.

• A next-generation industrial facility can apply industrial AI as the plant’s “virtual assistant” to validate the quality and efficiency of a production plan in real time. AI-enabled cognitive guidance ultimately helps reduce confidence in individual domain experts to achieve complex decisions, and instead institutionalizes historical decisions and good practices to eliminate barriers to specialization.

These use cases are by no means specific, but are just a few examples: industry-wide AI capability, innovative, and generally applicable to the industry and laying the groundwork for the future digital workshop.

Industrial AI incorporates domain-specific knowledge along with the latest AI and machine learning capabilities in AI-enabled applications tailored for this purpose. This enables and accelerates the autonomous and semi-autonomous processes that carry out these operations — implementing the approach of Self-Optimizing Workflow.

The Self-Optimization Plant is a set of self-adaptive, self-learning, and self-sustaining industrial software technologies that work together to predict and act on future conditions by adapting operations within the digital enterprise. The combination of real-time data access and embedded industrial AI applications allows the Self-Optimizer Workshop to continually improve itself — using domain knowledge to optimize industrial processes, make easy-to-execute recommendations, and automate mission-critical flows.

This will have a number of positive effects on the business, including:

• Elimination of processes caused by process fluctuations and unplanned shutdowns or starts, achieving environmental, social and corporate governance goals. This reduces production waste and carbon footprint, fostering a new era of industrial sustainability.

• Promote overall safety by significantly reducing hazardous site conditions and relocating operations and production floor workers to safer roles.

• Unlocking new production efficiencies by optimizing margins and reaching new areas of production stability, even in declines, to achieve greater profitability.

It is a self-optimization plant that is not only industrial AI, but also the trajectory of the digital transformation of the industrial sector. By democratizing the application of industrial intelligence, the digital factory of the future promotes higher levels of security, sustainability and profitability and enables the next generation of digital workers – looking to the future in complex and volatile market conditions. This is the real potential of industrial AI.

To find out how industrial AI enables the digital workforce of the future and how to lay the foundations for a self-optimization workshop, visit
www.aspentech.com/selfoptimizingplant,
www.aspentech.com/accelerate, and www.aspentech.com/aiot.

This article was written by AspenTech. Not produced in the editorial board of the MIT Technology Review.