Production Yield Optimization

Production Yield Optimization (PYO) is a pragmatic and proven solution that enables manufacturers to optimize extruders, robotic, chemical, and other production processes using advanced AI trained using deep reinforcement learning (DRL) techniques. 

This solution integrates best practices and frameworks from design to training to deployment: Machine Teaching, process simulation, DRL, and IoT & Edge computing. 

From food to plastic and metal, extruders are used across industries for countless manufacturing processes. Traditional control systems, such as PLCs, do an excellent job at controlling extruders when all parameters are relatively stable. However, extruders age, raw material characteristics vary depending on the batch or supplier, and other environmental variables change. Therefore, it is not always possible for operators to consistently adapt control parameters and do it in a timely fashion to ensure optimum production yield, quality, and waste reduction.

With our PYO solution for extruders, an AI agent built using DRL will act as a meta-controller for your process. Our specialists design the AI by including your best subject matter experts’ knowledge and relevant system and environmental variables. This AI agent will control or propose in real-time new parameters and setpoints for your process controller and will either let your operator supervise or have them perform the actions themselves.

Learn more about how we implemented an extruder product yield optimization for PepsiCo’s Cheetos production.

Robotic control enables manufacturers to accelerate and normalize production. However, programming them is time-consuming and, once deployed, these programs can only support limited environmental changes. A misplaced component or a misaligned receptacle can be sufficient to throw the robot off its course and require human intervention.  

With our DRL solution for robotic control, manufacturers can deploy AI agents that can make the robotic process more effective and resilient to changing environmental configurations. 

See how we leveraged Microsoft Project Bonsai to build an AI-controlled robotic arm for handling coin bags.

Chemical process outputs are often extremely sensitive to the slightest change in raw material characteristics or equipment behavior. Using AI agents, chemical processes can become more robust these changes. 

Traditional AI would require an extensive data set of labeled data as well as AI system design expertise. This is hardly possible on a live chemical process and without hiring expert data scientists.  

We offer end-to-end services to design, train and deploy a yield optimization AI agent trained using Deep Reinforcement Learning algorithms.  

Learn more about the different frameworks and technologies we will use for your chemical process yield optimization process here.