PlanningVis: A Visual Analytics Approach to Production Planning in Smart Factories

Dong Sun, Renfei Huang, Yuanzhe Chen, Yong Wang, Jia Zeng, Mingxuan Yuan, Ting-Chuen Pong, Huamin Qu

TL;DR

PlanningVis is a visual analytics system that bridges the gap between automated production planning and human expertise in smart factories. It allows planners to interactively optimize schedules and rapidly respond to unanticipated disruptions—like supply chain failures or natural disasters—through a multi-level, human-in-the-loop interface.

What Problem Does This Paper Address?

In the era of Industry 4.0, factories generate massive amounts of data, but turning that data into an efficient production plan remains a "wicked" problem. Planners must balance four competing objectives: minimizing order delays, reducing production costs, keeping inventory low, and maintaining a "smooth" (consistent) workload for machines.

Existing solutions usually fall into two camps:

  1. Purely Automated Algorithms: These are often "black boxes." They might find a mathematically optimal solution, but they can’t account for real-world nuances or react quickly when things go wrong (e.g., a shipment of parts is late).
  2. Manual Planning: Using spreadsheets to track thousands of products and complex Bills of Materials (BOM) is slow, error-prone, and makes it nearly impossible to compare different "what-if" scenarios.

The authors address the need for a system that combines the speed of optimization algorithms with the intuition and flexibility of human experts.

Methods

The researchers developed PlanningVis, which consists of a hybrid backend and a three-tiered visualization frontend:

  • The Hybrid Algorithm: To handle the scale of modern manufacturing, they use a combination of Linear Programming (for mathematical optimization) and heuristic algorithms (to keep the computation time fast enough for interactive use).
  • Multi-Level Visualization:
    • Plan Overview: Uses custom "plan glyphs" to show the "health" of different plan versions at a glance.
  • Product View: Employs a novel "Segmented Parallel Coordinates Plot." This design solves a common data viz problem: products with zero demand (outliers) usually squash the rest of the data. By segmenting the axes, they keep the visualization readable.
  • Production Detail View: Visualizes the Bill of Materials (BOM) as a "skewed tree" linked to daily production heatmaps. This allows planners to trace a finished product's delay back to a specific sub-component or raw material shortage.

Datasets

The system was evaluated using a sampled real-world dataset from a global telecommunications equipment manufacturer.

  • Scope: 30 days of planning data.
  • Scale: 1,038 distinct products or assembly items.
  • Content: Includes initial inventories, factory capacity sets, holiday schedules, and real/predicted order demands.

Key Insights and Ideas

  • Human-in-the-loop "What-If" Analysis: The core "aha" moment is that the system isn't just for making a plan; it's for testing them. Planners can drag-and-drop constraints—like adding a "holiday" to simulate a plant shutdown—and see the ripple effects across the entire supply chain in seconds.
  • Visualizing Variance: The product glyphs don't just show average performance; they use the angle of an arc to represent the variance of daily indicators. This helps planners identify products that are "unstable" even if their average looks fine.
  • The Power of Juxtaposition: By placing the BOM tree directly next to production heatmaps, the system reveals "hidden" constraints. For example, a router might be delayed not because of a lack of machine time, but because a specific shared raw material was prioritized for a different product.

Key Results

The authors conducted two case studies and interviews with six domain experts:

  • Optimization Success: In one case study, planners used the tool to identify a bottleneck in specific capacity sets. By increasing these by 50%, they significantly reduced order delays while maintaining a manageable production cost.
  • Crisis Management: The team simulated the impact of Typhoon Mangkhut. By adding a 2-day "holiday" (shutdown) to the affected plant, they used PlanningVis to quickly re-allocate resources and mitigate the resulting delays.
  • Expert Validation: Domain experts from Huawei noted that the system could "significantly improve production planning efficiency" compared to their traditional table-based workflows, specifically praising the ability to localize the root cause of problems.

Limitations and Future Work

  • Algorithm Latency: While fast, the algorithm still takes several seconds to run. For massive factories with tens of thousands of products, this could break the "flow" of interactive exploration.
  • Logistics Gaps: The current model does not account for transportation time or costs between different factory locations.
  • Granularity: The system currently looks at the factory level. The authors suggest future versions should drill down to specific production lines and individual machines to provide even more actionable insights.

Summary updated

Abstract

Production planning in the manufacturing industry is crucial for fully utilizing factory resources (e.g., machines, raw materials and workers) and reducing costs. With the advent of industry 4.0, plenty of data recording the status of factory resources have been collected and further involved in production planning, which brings an unprecedented opportunity to understand, evaluate and adjust complex production plans through a data-driven approach. However, developing a systematic analytics approach for production planning is challenging due to the large volume of production data, the complex dependency between products, and unexpected changes in the market and the plant. Previous studies only provide summarized results and fail to show details for comparative analysis of production plans. Besides, the rapid adjustment to the plan in the case of an unanticipated incident is also not supported. In this paper, we propose PlanningVis, a visual analytics system to support the exploration and comparison of production plans with three levels of details: a plan overview presenting the overall difference between plans, a product view visualizing various properties of individual products, and a production detail view displaying the product dependency and the daily production details in related factories. By integrating an automatic planning algorithm with interactive visual explorations, PlanningVis can facilitate the efficient optimization of daily production planning as well as support a quick response to unanticipated incidents in manufacturing. Two case studies with real-world data and carefully designed interviews with domain experts demonstrate the effectiveness and usability of PlanningVis.
BibTeX
@article{sun2020planningvis,
  author = {Dong Sun and Renfei Huang and Yuanzhe Chen and Yong Wang and Jia Zeng and Mingxuan Yuan and Ting-Chuen Pong and Huamin Qu},
  title = {{PlanningVis: A Visual Analytics Approach to Production Planning in Smart Factories}},
  journal = {IEEE Trans. Vis. Comput. Graph.},
  year = {2020},
  url = {https://openreview.net/forum?id=jkQPn6tIqN},
}

Loading related content...

Figure