FliPSi: Generating Data for the Training of Machine Learning Algorithms for CPPS

Anomaly detection, the recognition of faulty behavior, and diagnosis, process identification root cause a fault, in Cyber-Physical Production Systems (CPPS) are complex tasks, due to increasing complexity modularity modern production systems. But amount data, generated by sensors, offers solution: Machine Learning (ML) can be used automatically generate models for anomaly detection diagnosis based on data---at least addition manual models. While nowadays huge datasets CPPS exist, they mainly cover repetitive OK situations lack fault modes, which is problematic training ML algorithms. In this work we present possibility overcome problem generating data with simulation modular CPPS. For use game engine Unity, has already been employed self-driving cars robots. Within modules designed programmed. Afterwards combined into CPPS, then using built-in physics Unity. The not constrained certain applications, but adequately simulate wide variety scenarios, allowing combination different types. exported purposes. simulation, faults inserted sets containing normal behavior exported. This way, algorithms trained diagnosis. especially useful pre-training algorithms, later fine-tuned from real cases overcome.