With Industry 4.0 comes the promise of leapfrog in productivity, quality, and the overall return on your manufacturing investment. A key driver of those improvements is the large amount of data gathered from the sensors, actuators, and other connected IoT devices deployed on the shop floor. The availability of IoT-driven, real-world data combined with the latest AI techniques opens the door for practical AI use cases to help solve real-world manufacturing problems.

The new techniques, specifically Deep Reinforcement Learning-based AI agents, allow AI solutions to move from theory (research labs and academia) to practical application on shop floors.

Where can AI help improve process manufacturing?

There are many challenges associated with optimizing process manufacturing. At a conceptual level, one can think about three main areas where each improvement can generate significant ROI.

Whether a process comprises steps such as extruding, melting, cutting, heating, cooling, forming, mixing, or more, typically, an AI agent can positively impact manufacturing by: 

1. 1. Minimizing start-up time and therefore reducing both lost material and production time loss
   2. Ensuring the end product stays within specifications and proactively adapt control system parameters (e.g., PLC set-points for various elements of the process) to avoid out-of-spec production drift
   3. Controlling overall quality to reduce defects in environments with changing conditions such as raw material variability, equipment wear and tear, and more

How can AI improve process manufacturing? 

Traditionally training an AI agent required advanced data science knowledge (i.e., machine learning algorithms from statistical to deep neural networks) and large amounts of human-verified training data. It is, in most cases, neither technically nor economically feasible in real life.

However, it is now possible to develop effective AI agents by combining internal process expertise with the newest simulation and Deep Reinforcement Learning (DRL) techniques.

Toolchains, such as Microsoft Project Bonsai, leverage the experience of subject matter experts (SMEs) to develop those AI agents without the need for in-depth data science. From operators to research engineers, these SMEs help define the best approaches to select the relevant data, simulate the process appropriately, and define the DRL training parameters.

DRL is a type of training that leverages simulators to train an AI agent that Project Bonsai will have automatically designed based on the system’s specificities: inputs, outputs, and complexity.

Who is already using this approach?

Although this approach is very new, working with the Microsoft Autonomous Systems engineering team, Neal Analytics has already helped several customers design, train and deploy DRL-trained AI agents using Project Bonsai.

One example of this process in action is PepsiCo’s Cheetos manufacturing process shown in the video below. In this example, the process consists of two stages: extrusion and baking.

The following video provides more high-level information about this project.

To start investigating whether your process is a likely candidate for optimization through a Project Bonsai AI Agent (aka “brain”), check out the resources below or contact us. 

Learn more: 

• • Deep Reinforcement Learning  
  • Simulation building strategies 
  • Step-by-step guide: Getting started with Project Bonsai
  • More details on the PepsiCo customer story