Search Results

Due to stricter emission regulations and more environmental awareness, the powertrain systems are moving toward higher fuel efficiency and lower emissions. In response to these pressing needs, new technologies have been designed and implemented by manufacturers. As a result of increasing complexity of the powertrain systems, their control and optimization become more and more challenging. Virtual powertrain calibration, also known as model-based calibration, has been introduced to transfer a part of test bench testing into a virtual environment, and hence considerably reduce time and cost of product development process while increasing the product quality. Nevertheless, virtual calibration has not yet reached its full potential in industrial applications. Volvo Penta has recently developed a virtual test cell named VIRTEC, which is used in an ongoing pilot project to meet the Stage V emission standards.

The present paper aims at developing an innovative procedure to create a one-dimensional (1D) real-time capable simulation model for a heavy-duty diesel engine. The novelty of this approach is the use of the top-level engine configuration, test cell measurement data, and manufacturer maps as opposite to common practice of utilizing a detailed 1D engine model. The objective is to facilitate effective model adjustments and hence further increase the application of Hardware-in-the-Loop (HiL) simulations in powertrain development. This work describes the development of Fast Running Model (FRM) in GT-SUITE simulation software. The cylinder and gas-path modeling and calibration are described in detail. The results for engine performance and exhaust emissions produced satisfactory agreement with both steady-state and transient experimental data.

In order to meet emissions and power requirements, modern engine design has evolved in complexity and control. The cost and time restraints of calibration and testing of various control strategies have made virtual testing environments increasingly popular. Using Hardware-in-the-Loop (HiL), Volvo Penta has built a virtual test rig named VIRTEC for efficient engine testing, using a model simulating a fully instrumented engine. This paper presents an innovative Artificial Neural Network (ANN) based model for engine simulations in HiL environment. The engine model, herein called Artificial Neural Network Engine (ANN-E), was built for D8-600 hp Volvo Penta engine, and directly implemented in the VIRTEC system. ANN-E uses a combination of feedforward and recursive ANNs, processing 7 actuator signals from the engine management system (EMS) to provide 30 output signals.

The automotive industry makes extensive use of virtual models to increase efficiency during the development stage. The complexity of such virtual models increases with the complexity of the process that they describe, and thus new methods for their development are constantly evaluated. Among many others, data-driven techniques and machine learning offer promising results, creating deep neural networks that map complex input-output relations. This work aims at comparing the performance of two different neural network architectures for the estimation of the engine state and emissions (flow fuel, NOx and soot). More specifically, Convolutional Neural Network (CNN) and Long-Short Term Memory (LSTM) will be evaluated in terms of performance, using different techniques to increase the model generalization. During the learning stage data from different engine cycles are fed to the neural networks.

In engine research and development there are often different engine parameters that produce similar effects on the end-point results. When calibrating modern engines, a huge number of parameters needs to be set, which also includes compensation parameters for model imperfections. In this context, simpler, more robust, and physically based models should be beneficial both for calibration work load and powertrain performance. In this study, we present an experimental methodology that uses intermediate (“intrinsic”) variables instead of engine parameters. By using simple thermodynamic models, the engine parameters EGR, IVC, and PBoost could be translated into oxygen concentration, temperature and gas density at the start of injection. The reason for this transformation of data is to “move” the Design of Experiment (DoE) closer to the situation of interest (i.e. the combustion) and to be able to construct simpler and more physically based models.