Optimize system-level efficiency of your extruder with AI
As discussed in previous articles, here and here in particular, Deep Reinforcement Learning (DRL) trained AI agents can help optimize manufacturing processes across different manufacturing challenges aspects.
In addition to being able to help with those challenges at a specific piece of equipment level, those AI agents can also help at a broader system level.
When training those AI Agents, aka “brains,” if using the Microsoft Project Bonsai toolchain, what is essential is that the process simulator inputs and outputs accurately represent the behavior of the equipment the agent will control.
But who said it had to be “a machine,” e.g., an extruder, and not a system made of multiple pieces of equipment?
Conceptually, from an AI agent perspective, whether the controlled process comprises one or multiple pieces of equipment does not matter. It is the system’s overall behavior that the agent will optimize. It does not need to know whether this is an extruder, an extruder plus an oven, or any other combination of process manufacturing steps.
This approach is crucial because measuring output quality at every step is not always possible. For instance, until a plastic or metal object is cut, it might not be measurable whether the melted material used was correctly extruded or not.
The AI agent will be an AI model in a Docker container running for instance on an Azure IoT Edge device. It will physically be deployed close to the process wherever it makes the most sense and to whatever edge topology is being used at the system, line, and plant-level.
Let’s now look at an example of an AI agent taking this system-level view of a production line with the specific example of the PepsiCo Cheetos extrusion project. For this project, it was not possible to correctly assess whether the extrusion process performed properly or not until the (delicious)snack was baked.
Therefore, as illustrated below, the AI agent monitors and controls not the extruder alone, but the system made of the extruder plus the oven.
We can extrapolate this approach to many types of extrusion or non-extrusion-based manufacturing processes.
Granted, not every DRL project in process manufacturing will use this approach. However, when needed, it is possible to take a higher-level look at a process to decide the right granularity level for this AI agent.
Please refer to this customer story to learn more about the PepsiCo Cheetos project. You can also learn more about this project in 10-minutes video.
- About Autonomous Systems
- Designing an AI agent to optimize extruder operations: A real-life example
- Step-by-step guide: Getting started with Project Bonsai
- Optimizing plastic extrusion with AI: Self-service demo
- Top reasons for using AI-based simulator vs. a physics based simulator for your process manufacturing RL projects
- 5 points to consider before starting your DRL autonomous system project in process manufacturing optimization