FAMILIAR Requirements Definition

UC1: Immersive Driving Data Simulation Addon

The main goal of the UC1 is to enhance ADAS developments and functions, by using FedML and XR Technology, based on decentralized system architecture. This is response to the objectives set by the EU, related to the Vision Zero accidents by 2030 mission, and the need of the automotive sector to collaborate effectively within value chain, especially in the field of data sharing between unites and entities involved, without losing privacy, control and security.

In ADAS function development, isolated approaches are made by Tier1 or internal engineering units. Combining FedML and XR engineering methods would allow a substantial improvement, including controlling data flows more efficiently. The test application of integrating a minimum of one vehicle test bench or other simulation devices would increase the number of human factor simulations, including a complete vehicle model.

Expected results are decentralized development architecture using FedML and XR technology to enhance vehicle and training data for ADAS functions. In addition, the setup of a driving simulator based on a test bench and a flexible integration of a laboratory XR environment is conducted.

UC2: Setup assistance for 3D printing industrial robots

Background and challenges:

Industrial robots and Wire Arc Additive Manufacturing (WAAM) can be used to produce individual parts decentrally. However, there are only a few experts who specialize in setting up such WAAM 3D printing processes and systems. Countless parameters or setup configurations have to be set up for each build part in order to achieve a good printing result. This currently requires a high level of know-how of the machine operator. An insufficient setup results in poor part quality and costly scrap production.

Technical approach in FAMILIAR:
The solution is an AI-based assistance to identify the setup parameters based on the CAD file and the environment. Initial setup configurations are displayed to the machine operator with the help of XR. During printing, geometric errors can occur due to poor setup. For localization of these anomalies, a geometrical platform monitoring system has to be integrated. The AI must identify the nature of the defect, while predicting how the setup parameters need to be adjusted. Possible adjustments of hardware-related parameters, such as clamping elements, are indicated to the machine operator with the help of XR. Finally, setup parameters are suggested, and the subsequent printing process is more promising.

UC3: Extended failure prediction in FEM simulation software

In use case 3, artificial neuronal networks (ANN) are trained by sources such as manufacturing processes, quality testing, advanced optic sensors, and head mounted display (HMD) via FedML. In particular, ANN will be used to extract helpful information from an uncorrelated data set by understanding the data and finding the patterns inside them, which are not possible by humans, determine the data related to production and model parameters that affect the part quality, mechanical performance, production quality. Furthermore, the ambiguities and assumptions involved in the material modelling and characterization using conventional methods are avoided; instead, a data-driven method based on machine learning is used for material modelling. This way, the margin of error in the simulation models decreases by joint effort. 

The ANN is supposed to decrease the margin of error in the FEM simulation models in the value chain and increase the accuracy of FEM simulations. Potential data source are the data taken from various manufacturing processes, tests, and advanced cameras. It was shown in previous research that implementation of an ANN in data characterization had been the efficient way in another context.