Chaos-X

Motivation

A Europe-wide study [1] from 2024 shows that nearly all companies consider resilience to be critical to their success. Despite its importance, technological approaches to enhancing resilience—such as digital twins or adaptive scheduling as part of production optimization—have so far been used only to a limited extent. Existing resilience tools often fail to adequately address internal and external disruptions on the shop floor, thereby compromising production targets. In information technology, however, chaos engineering has established itself as a proven approach for identifying vulnerabilities and improving system resilience through deliberately introduced disruptions. However, due to their complexity and potential risks to safety and, a direct application to industrial production systems is only possible to a limited extent. This creates a need for new resilience tools that safely and effectively apply the principles of chaos engineering to industrial production.

Objectives

The goal of the project is to experimentally evaluate and enhance the resilience of complex production systems at the shop floor level by applying chaos engineering to industrial production. To this end, an ICT-supported platform is being developed that integrates two complementary resilience tools: a digital twin for implementing the “daydreaming factory” as a model-based approach, and a connected adaptive, AI-based scheduling system that learns proactively. In addition, a standardized chaos generator is being developed that specifically introduces disruption scenarios into the simulation model and makes their effects analyzable. The digital twin serves as a virtual test environment for continuous resilience assessment, risk analysis, and validation of measures. Building on this, the AI-based scheduling is expanded to include resilience metrics, enabling agents to learn resilient decision-making strategies through the simulation of various disruption scenarios and allowing production plans to be adaptively adjusted. The developed methods and approaches will first be piloted in learning factories and subsequently validated in industrial application scenarios.

Approaches

First, the chaos engineering approach is applied to industrial production. To this end, specific disruption scenarios are developed and analyzed using a digital replica of the production system to identify vulnerabilities. Since directly introducing disruptions into the real production environment is not feasible due to potential risks to people, machinery, and processes, the analysis is conducted in a secure simulation environment known as the “daydreaming factory.” This environment combines real operational data with simulation-based scenarios and enables the systematic evaluation of system responses. The technological foundation is a digital twin, which is built using standardized data structures via OPC UA and the Asset Administration Shell (AAS). The insights gained from the simulated fault scenarios are then incorporated into production control. Using AI-based scheduling methods, production plans are adaptively adjusted to ensure, even in the event of disruptions, that production targets such as on-time delivery and throughput remain within an acceptable tolerance range to the greatest extent possible. This results in a data-driven, continuous resilience strategy for industrial production systems, utilizing resilience tools.

Sources

[1] Bentz Daniel, Doan Augustino, Meldt Leonie, Steinmeyer Maximilian, Metternich Joachim, Becker Martin. Resilienz in der industriellen Produktion: Eine Aufnahme der Ist-Situation. 2025. 10.26083/tuprints-00029006

Acknowledgement

This project is funded by the Federal Ministry of Research, Technology, and Space (Guidelines for Funding Research Projects to Improve the Exploration and Integration Phases of ICT Research). We are grateful for the opportunity to work on this project.

Funding source

Funding source