Smart Manufacturing: MLOps-Enabled Event-Driven Architecture for Enhanced Control in Steel Production

TL;DR

The paper tackles real-time adaptive control in steel forging by marrying a Digital Twin with edge-enabled, event-driven MLOps and Deep Reinforcement Learning to autonomously optimize induction-furnace heating. It introduces an EDMA-backed architecture that ingests sensor data, maintains a live digital twin, and trains DRL agents (DQN and PPO) to adjust power in response to state snapshots, validated in a steel-furnace DT environment and demonstrated in a real-world use case at Bharat Forge Kilsta. The study presents architectural patterns, data flows, and model-management practices that collectively reduce waste, improve energy efficiency, and raise product quality, while emphasizing edge computing and low-latency control suitable for industrial settings. The insights offer a scalable blueprint for applying DTs, MLOps, and DRL to other heavy industries, with concrete findings on performance, deployment, and generalizability, and a roadmap for future enhancements such as CI/CD, transfer learning, and explainability.

Abstract

We explore a Digital Twin-Based Approach for Smart Manufacturing to improve Sustainability, Efficiency, and Cost-Effectiveness for a steel production plant. Our system is based on a micro-service edge-compute platform that ingests real-time sensor data from the process into a digital twin over a converged network infrastructure. We implement agile machine learning-based control loops in the digital twin to optimize induction furnace heating, enhance operational quality, and reduce process waste. Key to our approach is a Deep Reinforcement learning-based agent used in our machine learning operation (MLOps) driven system to autonomously correlate the system state with its digital twin to identify correction actions that aim to optimize power settings for the plant. We present the theoretical basis, architectural details, and practical implications of our approach to reduce manufacturing waste and increase production quality. We design the system for flexibility so that our scalable event-driven architecture can be adapted to various industrial applications. With this research, we propose a pivotal step towards the transformation of traditional processes into intelligent systems, aligning with sustainability goals and emphasizing the role of MLOps in shaping the future of data-driven manufacturing.

Figures (14)

• Figure 1: A graphical and simplified representation of the furnace.
• Figure 2: Temperature ranges in the furnace.
• Figure 3: System architectural Overview
• Figure 4: The Telemetry parser’s structure and its main interconnections
• Figure 5: The Power control structure and its main interconnections