Skip to Content

Intelligent control technologies impacting industrial operations

Capgemini
August 20, 2020

On average, 50% of companies that embrace artificial intelligence (AI) over the next 5–7 years may double their cash flow, and manufacturers that implement intelligent systems will achieve 17–20% productivity gains, according to the Microsoft whitepaper Bringing autonomy for the industrial control system [1]. In this article, we assess how Capgemini’s customers are responding to new technology and take a close look at Microsoft’s Project Bonsai solution as an enabler of autonomous systems.

Our work across the manufacturing sector has given us a perspective on how manufacturers are establishing connectivity between machines, production lines, and control systems. We can see that intelligent, connected systems are already guiding industry-wide digital transformation at several levels, including:

  • At the predictive level, autonomous systems are being used to leverage IIoT data to gain a greater understanding of the facility operation, using machine learning to predict key performance indicators (KPIs).
  • At the prescriptive level, adaptive self-optimizing technology and processes are being implemented in plant operations, including intelligent control systems that help equipment and machinery adapt in real time to changing inputs.

One solution set to transform industrial control systems is Microsoft’s AI development platform, Project Bonsai. This machine learning service creates and optimizes intelligence for autonomous control systems. Powered by AI, a Bonsai brain can adjust itself to changing environments and even optimizes towards multiple goals driving innovation and empowering employees to work efficiently in the transformative process.

Creating digital value

Digital has the potential to create new value for our customers through new connections via advisory, assistive, and autonomous capabilities, along with data and analytics across businesses are leading to network effects, scale, efficiency, and ecosystem-generating new opportunities. However, the reality today is that training autonomous systems requires deep expertise in AI, essentially making it unscalable. Traditional machine learning methodologies aren’t enough.

So, what’s needed? This is where Microsoft’s Bonsai steps into the ring. It responds to the need to infuse deep learning networks into discrete systems and processes across the digital landscape with easy access to AI for those who do not necessarily have advanced data science expertise. At Capgemini, we believe that bringing intelligence to autonomous systems at scale will require a unique combination of the new practice of machine teaching, advances in deep reinforcement learning and leveraging simulation for training.

Bonsai has fundamentally taken the above approach to building, training, and deploying AI models and empowering enterprises with the tools to solve the toughest optimization and automation problems. It allows subject matter experts to infuse their intelligence into machine learning models, resulting in more efficient, collaborative and accurate model development.

Bonsai uses simulations to train an AI model and the platform automatically selects the most appropriate deep reinforcement learning algorithm needed for training a specific model. This lays out the neural networks, tuning hyperparameters, decreasing model training time and allowing for reusability of each individual concept. Using Bonsai, we can focus on solving the business problem instead of constantly wrangling with low-level toolkits.

How does the Microsoft Bonsai platform work?

  • The first step is to build a model by leveraging the Bonsai machine teaching technique to deconstruct complex problems into the key concepts you want an AI model to learn.
  • These concepts are programmed into a model using special-purpose programming language Inkling.
  • The Inkling code then connects with the any existing or custom simulator including MATLAB Simulink (engineering, manufacturing), Transys (energy), Gazebo (robotics), and AnyLogic (supply chain).
  • Each machine teaching program, along with the specified simulation, is fed into the AI engine, where it is compiled to automatically generate and train the best machine learning model for a given problem.

The resulting high-level model can then be connected into your hardware or software application through Bonsai-provided libraries. As companies look for AI solutions, they often find that different tools are needed for different stages of the process. Bonsai’s end-to-end platform provides industrial operations owners with a complete set of tools for the programming, runtime, and deployment of deep reinforcement learning models for use in autonomous systems.

Looking ahead

Our next challenge is how to measure the value captured from AI in manufacturing and build a business case for autonomous systems. Through Capgemini’s partnership with Microsoft deploying autonomous systems in factories, we are exploring how new data can create more options for an organization to monetize the value created. Stay tuned for our next post…

Find out how Capgemini can accelerate the effectiveness of your digital manufacturing operations at scale with our “Factory of the Future” approach.

References:

[1] Bringing autonomy for the industrial control system


Authors

Srinivas Datla
Associate Vice President Digital Manufacturing
Email: srinivas.datla@us.sogeti.com
Subrato Dhar
Director, Digital Manufacturing
Email: subrato.dhar@us.sogeti.com