PSI - Issue 38

Florian Grober et al. / Procedia Structural Integrity 38 (2022) 352–361 Grober, Janßen, Küçükay / Structural Integrity Procedia 00 (2021) 000 – 000

353

2

1. Introduction The vehicle development of the 21st century is facing major challenges. On the one hand, the economic pressure requires a product design that is tailored precisely to the customer’s needs . On the other hand, modern trends, such as electromobility, car sharing or automatic driving, accelerate the change in user behavior. In consequence, two challenges arise for vehicle fatigue design: Firstly, the changes in user behavior must be continuously monitored in order to create precise load assumptions for subsequent development projects. This requires extensive data collection campaigns in the field which are fortunately supported by the trend towards digitization and data-driven product design. Secondly, the fatigue design of the vehicle components must match the customer requirements as accurately as possible, so that neither expensive over-dimensioning nor safety-threatening under-dimensioning takes place. In order to record the customer usage in the field, an approved method is the acquisition of data from vehicle internal sensors using durability counting functions, Pötter et al. (2010). Common examples for such methods are: ‘Range - Pair Counting’, ‘Level -Cros sing Counting’, ‘Rainflow Counting’ and ‘Time -at- Level Counting’ (see Köhler et al. (2017)). They enable a significant reduction in the required memory while maintaining meaningfulness. By collecting data from a large fleet of vehicles, a comprehensive customer knowledge arises, which enables the derivation of valid load spectra for component design. However, the inbuilt vehicle sensors only seldom measure specific load values such as forces or rotational moments. In most cases, general driving variables, as for example speed or acceleration, are recorded instead. Although these are not directly usable for the design and testing of components, they have the advantage that they are mostly vehicle type-independent and can therefore be better applied to subsequent projects. For this reason, the first challenge to be answered by this article is the transfer of general usage field data into concrete loads for customer-oriented testing of new vehicles under development. In general, the mere definition of testing requirements is not sufficient to guarantee ideal fatigue design because it is also necessary to ensure that the defined target load spectra are adequately realized in practice. Especially the durability road testing experiment, in which the vehicle prototype has to reach a compulsory mileage on special test courses on proving grounds or public roads damage-free, is usually not completely reproducible. This is mainly due to minor deviations from the driving instructions, for example as a result of bad weather, driver influence, street closures or changes in the road surface. As proven in Saathoff et al. (2004), these deviations reach a relevant extent even on well-defined test courses on proving grounds. The second challenge discussed in this article is therefore the implementation of a monitoring system for the achieved loads in the durability road test to ensure compliance with the testing requirements. In order to solve the described problems, a driver guidance system for intelligent route planning and optimal durability testing on the basis of target load spectra from customer field data is developed. 2. State of the art 2.1. Customer-oriented load spectra from field data While there have always been individual load operation measurements in customer vehicles for load spectra determination, the trend towards comprehensive mass data collection in the field has just arisen in the past few years due to an advancement in the technical possibilities. Therefore, the research question of a transformation of physical values that can be measured in the field using vehicle-internal sensors into concrete loads for the testing of subsequent projects has not yet been fully answered. On the one hand, there are methods for the implementation of ‘ virtual sensor s’ that perform an onboard calculation of load signals from the available signals using physical models or artificial intelligence in the vehicle control unit, Grupp et al. (2019). In a second step, these calculated signals can be processed by a durability counting algorithm. Since the results are related to the vehicle already in the field and cannot be easily transferred to other vehicle types, these methods are primarily used for load accumulation in scope of predictive maintenance approaches. For this reason, Grober et al. (2019) investigates the derivation of load spectra for vehicles to be developed from classified customer usage data. For this purpose, transfer functions between reference and load values from multibody simulation as well as adaptation methods for vehicle properties are created. This approach reaches its limits in complex multi-axial stress states, since the temporal reference between the input signals is blurred by the counting algorithms and cannot be clearly reconstructed. Thus, in Grober