Search Results

Computational models directly derived from data gained increased interest in recent years. Data-driven approaches have brought breakthroughs in different research areas such as image-, video- and audio-processing. Often denoted as Machine Learning (ML), today these approaches are not widely applied in the field of vehicle Noise, Vibration and Harshness (NVH). Works combining ML and NVH mainly discuss the topic with respect to psychoacoustics, traffic noise, structural health monitoring and as improvement to existing numerical simulation methods. Vehicle interior noise is a major quality criterion for today’s automotive customers. To estimate noise levels early in the development process, deterministic system descriptions are created by utilizing time-consuming measurement techniques. This paper examines whether pattern-recognizing algorithms are suitable to conduct the prediction process for a steering system.

Existing electronic combustion engine control systems only guarantee a desired air-to-fuel-ratio λ in stationary operation. In order to achieve the desired λ also in in-stationary use of the engine, it is necessary to use new-technology-based control systems. Artificial Intelligence provides methods to cope with difficulties like wide operation range, unknown nonlinearities and time delay. We will propose a strategy for control of a Spark Ignition Engine to determine the mass of air inside the combustion chambers with the highest accuracy. Since Neural Networks are universal approximators for multidimensional nonlinear static functions they can be used effectively for identification and compensation purposes of unknown nonlinearities in closed control loops.

In order to reduce emissions of spare ignition engines using a three way catalyst, a stoichiometric air-fuel ratio must be guaranteed in stationary and transient operation of the engine. This aim can be reached by using a specific feed-forward structure for the control of the paths of air and fuel based on identification abilities of Artificial Intelligence. As approximators for multidimensional nonlinear static functions we will use specific Neural Networks (NN) together with sophisticated stability-proven learning structures. The acquired knowledge within the NN determines our control action mainly through using feed-forward structures. Our investigations are based on the so-called mean-value-modelling approach of SI engines; it is our aim to present this strategy.

In motorsports, aerodynamic development processes target to achieve gains in performance. This requires a comprehensive understanding of the prevailing aerodynamics and the capability of analysing large quantities of numerical data. However, manual analysis of a significant amount of Computational Fluid Dynamics (CFD) data is time consuming and complex. The motivation is to optimize the aerodynamic analysis workflow with the use of deep learning architectures. In this research, variants of 3D deep learning models (3D-DL) such as Convolutional Autoencoder (CAE) and U-Net frameworks are applied to flow fields obtained from Reynolds Averaged Navier Stokes (RANS) simulations to transform the high-dimensional CFD domain into a low-dimensional embedding. Consequently, model order reduction enables the identification of inherent flow structures represented by the latent space of the models.

Aging effects such as coking or cavitation in the nozzle of common rail (CR) diesel injectors deteriorate combustion performance. This is of particular relevance when it comes to complying with emission legislation and demonstrates the need for detecting and compensating aging effects during operation. The first objective of this paper is to analyze the influence of worn nozzles on the injection rate. Therefore, measurements of commercial solenoid common rail diesel injectors with different nozzles are carried out using an injection rate analyzer of the Bosch type. Furthermore, a fault model for typical aging effects in the nozzle of the injector is presented together with two methods to detect and identify these effects. Both methods are based on a multi-domain simulation model of the injector. The needle lift, the control piston lift and the pressure in the lower feed line are used for the fault diagnosis.

Deep neural network models have been widely used for environment perception of intelligent vehicles. However, due to models’ innate probabilistic property, the lack of transparency, and sensitivity to data, perception results have inevitable uncertainties. To compensate for the weakness of probabilistic models, many pieces of research have been proposed to analyze and quantify such uncertainties. For safety-critical intelligent vehicles, the uncertainty analysis of data and models for environment perception is especially important. Uncertainty estimation can be a way to quantify the risk of environment perception. In this regard, it is essential to deliver a comprehensive survey. This work presents a comprehensive overview of uncertainty estimation in deep neural networks for environment perception of intelligent vehicles.

In order to predict reality as accurately as possible leads to the fact that numerical models in automotive vibroacoustic problems become increasingly high dimensional. This makes applications with a large number of model evaluations, e.g. optimization tasks or uncertainty quantification hard to solve, as they become computationally very expensive. Engineers are thus faced with the challenge of making decisions based on a limited number of model evaluations, which increases the need for data-efficient methods and reduced order models. In this contribution, variational autoencoders (VAEs) are used to reduce the dimensionality of the vibroacoustic model of a vehicle body and to find a low-dimensional latent representation of the system.

Electrified vehicles are particularly quiet, especially at low speeds due to the absence of combustion noises. This is why there are laws worldwide for artificial driving sounds to warn pedestrians. These sounds are generated using a so-called Acoustic Vehicle Alerting System (AVAS) which must maintain certain minimum sound pressure levels in specific frequency ranges at low speeds. The creation of the sound currently involves an iterative and sometimes time-consuming process that combines composing the sound on a computer with measuring the levels with a car on an outside noise test track. This continues until both the legal requirements and the subjective demands of vehicle manufacturers are met. To optimize this process and reduce the measurement effort on the outside noise test track, the goal is to replace the measurement with a simulation for a significant portion of the development.