FliPSi: Generating Data for the Training of Machine Learning Algorithms for CPPS
Anomaly detection, the recognition of faulty behavior, and diagnosis, the process of identification of the root cause of a fault, in Cyber-Physical Production Systems (CPPS) are complex tasks, due to the increasing complexity and modularity of modern production systems. But the increasing amount of data, generated by sensors, offers a solution: Machine Learning (ML) can be used to automatically generate models for anomaly detection and diagnosis based on data---at least in addition to manual models. While nowadays huge datasets for CPPS exist, they mainly cover repetitive OK situations and lack fault modes, which is problematic for the training of ML algorithms. In this work we present a possibility to overcome this problem by generating data with a simulation of modular CPPS. For simulation we use the game engine Unity, which has already been employed to generate data for the training of self-driving cars and the simulation of robots. Within Unity, modules of CPPS are designed and programmed. Afterwards they are combined into complex CPPS, which are then used for simulation using the built-in physics engine of Unity. The built-in physics engine is not constrained to certain applications, but can adequately simulate a wide variety of scenarios, allowing for the combination of different simulation types. The generated data can be exported for training purposes. In the simulation, faults can be inserted and data sets containing faults and normal behavior can be exported. This way, ML algorithms can be trained for fault detection and diagnosis. The data is especially useful for pre-training of ML algorithms, which can later be fine-tuned using data sets from the real CPPS. This way, the problem of repetitive data sets which lack data for fault cases can be overcome.
How to Cite
CPPS, Machine Learning, Simulation, Data Generation
This work is licensed under a Creative Commons Attribution 3.0 Unported License.
The Prognostic and Health Management Society advocates open-access to scientific data and uses a Creative Commons license for publishing and distributing any papers. A Creative Commons license does not relinquish the author’s copyright; rather it allows them to share some of their rights with any member of the public under certain conditions whilst enjoying full legal protection. By submitting an article to the International Conference of the Prognostics and Health Management Society, the authors agree to be bound by the associated terms and conditions including the following:
As the author, you retain the copyright to your Work. By submitting your Work, you are granting anybody the right to copy, distribute and transmit your Work and to adapt your Work with proper attribution under the terms of the Creative Commons Attribution 3.0 United States license. You assign rights to the Prognostics and Health Management Society to publish and disseminate your Work through electronic and print media if it is accepted for publication. A license note citing the Creative Commons Attribution 3.0 United States License as shown below needs to be placed in the footnote on the first page of the article.
First Author et al. This is an open-access article distributed under the terms of the Creative Commons Attribution 3.0 United States License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.