Focus on Second Open Call winners: GreenChemAI: AI-Enabled Sustainability for Rchemie’s Chemical Manufacturing

WHAT ABOUT THE EXPERIMENT?

Our experiment aims to optimize the viscosity control process in the motor oil production line using a Generative AI-based system. The key challenge addressed is the variability in the viscosity of the final product, which is currently controlled through manual adjustments and laboratory testing. This process is prone to delays and inefficiencies, as the production cycle requires repeated testing and manual intervention. Our goal is to integrate Generative AI with existing automation and SAP ERP systems to predict viscosity values in real-time and automatically adjust the formulation of the motor oil, reducing the need for manual interventions.

Operating within the automotive lubricant manufacturing sector, our experiment will leverage the following technologies: Open-source Large Language Models (LLMs) such as Llama 3, Mamba-7B, and Mixtral8x22B as the foundational AI technology. Customized AI Models for predicting viscosity levels based on real-time data inputs. Edge computing with GPU capabilities for efficient AI model execution and iterative optimization with Nvidia Jetson. Cloud-based model training for computationally intensive processes and seamless cloud- edge integration for updates. SAP ERP system integration for managing raw material inputs and production schedules. Automation systems integration for adjusting the formulation of the oil based on AI recommendations with Siemens S7-1500 PLC.

The innovative aspects of our project include: i.Enabling Gen AI to dynamically optimize manufacturing processes based on simple user-defined constraints. ii.Providing a user-friendly interface, empowering users of all technical levels to interact with the system effectively. iii.Ensuring real-time optimization by deploying Gen AI on GPUs for accelerated computations.

The system will operate in an edge-to-cloud continuum, where the base Gen AI model is trained in a cloud environment but customized and deployed on the edge for enhanced data security and process-specific performance. Our experiment will take place at the RChemie production facility, located in Tuzla/Istanbul/Türkiye. This initiative represents a novel application of Generative AI in industrial manufacturing, aligning with Industry 5.0 principles by enhancing automation and human decision-making while ensuring data privacy and operational efficiency.

WHAT IS THE EXPECTED IMPACT?

This experiment will significantly contribute to the twin transition efforts of our company. By eliminating the need for manual intervention and laboratory testing delays, our production process will become more responsive, data-driven, and intelligent. This transition from a traditional, labor-intensive approach to a smart manufacturing system powered by AI will be a pivotal step in our digital transformation journey. The proposed experiment will contribute digitalization level of our company, particularly in three key dimensions:

Vertical Integration, Shop Floor Connectivity, and Shop Floor Intelligence.

1. Vertical Integration: One of the primary contributions of this project is the enhancement of vertical integration within our operations. By integrating shop floor operations with our ERP system (SAP), we will create a seamless data flow from the shop floor to the top floor. This integration will allow real-time data from the production line, including machine status, raw material levels, and viscosity predictions, to be linked directly with our ERP system. This will not only streamline decision-making at all levels but also provide higher visibility into production schedules, inventory management, and resource allocation. The system will enable us to operate with much greater efficiency, with the ability to adjust based on real-time insights rather than relying on manual intervention or post-production reports.

2. Shop Floor Connectivity: This project will also greatly improve shop floor connectivity by ensuring that our shop floor equipment—such as the Siemens S7-1500 PLC, production machines, and other sensors— are interconnected with AI-driven decision systems. The introduction of a Generative AI model that communicates in real-time with the automation system will provide dynamic control over the production process. Through this connectivity, data from the machines and sensors will be continuously transmitted and processed, enabling the automation system to adjust production parameters based on AI predictions without human input. This will minimize delays, reduce errors, and increase overall equipment efficiency, creating a more connected and intelligent manufacturing environment.

3. Shop Floor Intelligence: Finally, this project will contribute to the development of shop floor intelligence by significantly improving our ability to collect, process, and act on shop floor data. Currently, data from the production line is largely used for visibility and basic reporting. Through the integration of Generative AI, we will elevate our data processing from a visibility level to an adaptive level, where the system will not only monitor operations but also adapt and optimize processes in real-time. The AutoML system will continuously improve the AI model by learning from the laboratory results and process data, ensuring that the system becomes more precise over time. This intelligent system will allow us to predict and control production outcomes, leading to more efficient processes, higher-quality products, and reduced waste. The transition to adaptive intelligence will transform our approach to manufacturing, making our operations more proactive and data-driven.

In summary, this project will enhance our digital maturity by improving integration, connectivity, and intelligence on the shop floor, leading to faster decision-making, more efficient operations, and higher- quality products.