Industrial Data and Intelligent Algorithms for Production Scheduling Optimization

Introduction – The presentation focuses on research driven by industrial data and intelligent algorithms to optimize production scheduling, applying operations research concepts.

Strategic Significance – Smart manufacturing evolves through three paradigms: digital manufacturing, digital‑networked manufacturing, and next‑generation intelligent manufacturing. The core is process re‑engineering using digital technologies combined with lean, theory of constraints, agile, and flexible manufacturing principles. Networked integration of collected data enables knowledge‑based decision making, while advanced AI (deep learning, reinforcement learning) enhances system cognition.

Chinese Electronics Manufacturing – In 2021 the sector generated 14 trillion CNY revenue, with the smart production system market projected to reach 465.8 billion CNY by 2025, indicating rapid growth.

Digital and Intelligent Transformation – Many firms are at digital or early intelligent stages (Industry 4.0). Most have ERP systems and are upgrading MES and APS to handle multi‑variant, short‑life‑cycle production, which creates scheduling challenges due to frequent line changes and complex PCB assemblies.

Challenges in Production Scheduling – Scheduling faces high uncertainty (order changes, material availability, machine failures) and is a strongly NP‑hard, multi‑objective, dynamic problem requiring on‑time order fulfillment, capacity maximization, line balancing, and minimal changeovers.

Data‑Driven Algorithmic Approaches

1. Predict‑then‑optimize : Use capacity assessment and MRP data to predict production times, then schedule orders accordingly.

2. Robust optimization : Model uncertainty sets (box, budget, data‑driven) to obtain solutions resilient to variability.

3. End‑to‑end data‑to‑decision : Directly map data to decisions without separate prediction stage, balancing cost trade‑offs.

Key Scientific Problems – How to model and characterize uncertainty using industrial data, solve the resulting NP‑hard dynamic scheduling problem with efficient algorithms (decomposition, heuristic, learning‑based, parallel computing), and integrate data with algorithm design.

Data‑Algorithm Relationship – Data are the ingredients, algorithms the chefs; effective integration is essential for intelligent manufacturing.

Future Trends – Adoption of APS will require alignment of technology, people, and organization. China’s complete industrial chain provides abundant data and scenarios, suggesting strong future growth for data‑driven scheduling solutions.

Conclusion – Despite difficulty, the underlying problem remains a linear optimization task; as Jack Dongarra said, “everything is linear algebra.”

Click the image below for the conference replay.