ViPrIA – Virtual Product Development Using Intelligent Assistance Systems

The research project ViPrIA – Virtual Product Development Using Intelligent Assistance Systems – has been running since 1 October 2019. The project is funded by the Federal Ministry of Education and Research (BMBF) for a period of three years. Partners of the research project are SCALE, SIDACT and Fraunhofer SCAI. Associated industrial partners are AUDI, Porsche and Volkswagen.

The goal of ViPrIA is the development of intelligent assistance systems based on artificial intelligence (AI) and machine learning approaches (ML) to support engineers in simulation-based virtual product development. With the help of intelligent assistance functions, calculation engineers are to be supported in the development process with complex decisions and relieved of routine tasks.

Motivation

In the research and development process, complex and cost-intensive physical experiments are increasingly being replaced by computer-based simulations. In automotive manufacturing, virtual crash tests replace real tests with complex prototypes. Current challenges in the field of numerical simulation in product development are  

• to provide intelligent assistance systems for the development engineer and
  
• to make the handling of complex simulation data and their analysis easy.

Focal Points and Objectives

Numerical simulations are indispensable in modern product development. The engineer models the product with the help of a finite element grid with a very high degree of detail. In the case of crash tests in the automotive industry, the input data of a simulation model reflects information about various design configurations of the product, including material properties, geometry, connection components or operating ranges. Furthermore, different load cases are investigated, e.g. frontal, rear or side crash.

The results of ViPrIA will support calculation engineers in vehicle development in two essential stages of the development process

• On the one hand, an intelligent assistance system will be developed that supports the user already during the setup of a new numerical simulation, e.g. by indicating meaningful modifications and result prognoses.
  
• On the other hand, the goal is a successive automation of the analysis of simulation results, which generates a direct feedback to each simulation, if this shows an unexpected behavior. For this purpose, analysis methods – especially for anomaly detection – are carried out after simulation; in addition, new in-situ concepts for analysis during simulation are also considered.

Thus software components and method modules are developed in ViPrIA, which support the developer permanently in the research and development process with the help of an intelligent assistant and relieve him of routine tasks. In the future, design decisions will also be derived from historical data; the developments in the project represent the first important steps in this direction.

Scientific and Technical Objectives of the Project

With the help of the analysis of the existing database of simulations and experiments, an automatic plausibility check of product modifications of the development engineer shall be made possible during the design of a new numerical simulation run. This avoids complex and cost-intensive simulation runs or test series without usable results, and the entire development process becomes more efficient and cost-effective. In addition, new assistance systems will enable more agile development and qualitatively better results.

Another project goal is the automated analysis of simulation data during and after the simulation. Although different analysis steps already take place in the calculation workflow after each simulation, these are typically concentrated on scalar values. The project partners develop methods that enable a detailed comparison of simulation results at component level. These methods are then used in an extended and automated context. For this purpose, they are to be integrated into the simulation workflow, automated and used for anomaly detection. This allows searching for outliers in deformation behavior and automatically displaying unexpected behavior.

The results of the project are integrated into the development process with CAD and CAE. The designer, for example, makes modifications to the geometry and, if necessary, to the material used in the CAD program during the work process. Previously, he received feedback – typically several days later – from different CAE departments. The new developments in ViPrIA enable an immediate qualitative forecast of the effects of CAD design changes. For example, the data analysis of previous similar configurations of the model can show that occupant safety is deteriorating. In this way, the designer can perform a plausibility test with the help of an automated analysis of similar or comparable data. This helps to reduce the number of design iterations required, simplifying and accelerating the entire development process.

SCAI Work in ViPrIA

SCAI's work focuses on the search for and comparison of simulation configurations including anomaly detection as well as scalable and automated data analysis of simulation results. For this purpose, methods of machine learning will be used.