SI Engine Modeling Using Neural Networks

Ayeb, Mohamed; Lichtenthäler, D.; Winsel, Thomas; Theuerkauf, H. J.

doi:10.4271/980790

Cited by 40 publications

(12 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A well known architecture is the multi layer perceptron architecture (MLP). The equations presenting the network are given by (1). Where y is the output of the network to behave like the process output and x is the input vector including all the process variables influencing the output y.…”

Section: A Diagonal Recurrent Neural Networkmentioning

confidence: 99%

“…The following variables have been used as inputs for the engine torque estimation based on the diagonal recurrent neural network architecture presented above: engine speed, manifold pressure, valve actuation, EGR (Exhaust Gas Recirculation) actuation, ignition timing and fuel injection [1], [2].…”

Section: Engine Torque Modelmentioning

confidence: 99%

“…Neural networks are universal approximators capable of presenting almost any static or dynamic behaviour [11], [12] in a compact form suitable for real time applications [1]. A well known architecture is the multi layer perceptron architecture (MLP).…”

Section: A Diagonal Recurrent Neural Networkmentioning

confidence: 99%

“…... u(k) .. .u(N)] and y s T =[y s(1) ... y s (k) ... y s (N)]. A neural model receives the same input data and produces the output serie {y M (1),..., y M (N)}.…”

mentioning

confidence: 99%

See 3 more Smart Citations

Robust identification of nonlinear dynamic systems using Design of Experiment

Ayeb

Theuerkauf

Wilhelm

et al. 2006

2006 IEEE Conference on Computer Aided Control System Design, 2006 IEEE International Conference on Control Applications, 2006

Self Cite

View full text Add to dashboard Cite

Modern combustion engines are becoming increasingly complex, with more control variables, to meet the newest emissions regulations and improve the dynamic torque characteristics. However, the large number of control variables presents a challenge for the calibration process. The effects of all control variables and their interactions on the engine characteristics should be quantified in an accurate way to be able to derive the best control strategies aiming to minimise fuel consumption and emissions while generating the desired torque within acceptable time delays. Design of Experiment DoE methods have been used successfully to cope with the difficulties related to the characterisation of the underlying process in the presence of a large number of control variables. DoE provides experiment design methods aiming to reduce the number of measurements needed while ensuring the best quality of information related to a given optimisation criterion.In this paper a procedure is proposed to derive global dynamic models based on dynamic neural networks. The selection of the training data is derived by a D-optimality criterion to enhance the model quality. This ensures a reduction of the amount of measurement data needed for the training and hence an acceleration of the training process. It also improves the model quality by reducing the uncertainty on the model parameters. 1 1 (6) The matrix J J T is usually bad conditioned and its inversion can lead to numeric problems. The iteration is then improved leading finally to the Levenberg-Marquardtiteration [7], [8]:

show abstract

Section: A Diagonal Recurrent Neural Networkmentioning

confidence: 99%

Section: Engine Torque Modelmentioning

confidence: 99%

Section: A Diagonal Recurrent Neural Networkmentioning

confidence: 99%

“…... u(k) .. .u(N)] and y s T =[y s(1) ... y s (k) ... y s (N)]. A neural model receives the same input data and produces the output serie {y M (1),..., y M (N)}.…”

mentioning

confidence: 99%

See 2 more Smart Citations

Robust identification of nonlinear dynamic systems using Design of Experiment

Ayeb

Theuerkauf

Wilhelm

et al. 2006

2006 IEEE Conference on Computer Aided Control System Design, 2006 IEEE International Conference on Control Applications, 2006

Self Cite

View full text Add to dashboard Cite

show abstract

“…Brahma et al [22] used the ANN models to optimize control variables of the diesel engine, although with some challenges regarding noise in optimization results. On-Board Diagnostics (OBD) [23,24] and Hardware-in-the-Loop (HIL) simulation [25][26][27] exploited ANNs computational efficiency. Using ANN's for estimating mass air flow rate through the VVT engine has been demonstrated by Wu et al in [28].…”

Section: Introductionmentioning

confidence: 99%

A Simulation-Based Approach for Developing Optimal Calibrations for Engines with Variable Valve Actuation

Filipi

Prucka

et al. 2007

Oil & Gas Science and Technology - Rev. IFP

View full text Add to dashboard Cite

Résumé -Une approche basée sur la simulation pour le développement de réglages optimaux des moteurs avec système d'actionnement variable de soupape -La technologie d'actionnement variable de soupape (Variable Valve Actuation, VVA) présente un fort potentiel pour augmenter les performances et réduire consommation et pollution. Les avantages du VVA proviennent d'une meilleure aération et de la possibilité de contrôler les résidus internes. Par contre, l'addition de variables de contrôle supplémentaires dans un moteur VVA augmente la complexité du système, ce qui nécessite la mise au point d'une stratégie de contrôle optimale afin d'en tirer tous les bénéfices. L'approche traditionnelle reposant sur l'expérimentation s'avère trop onéreuse dans ce cas du fait de l'augmentation exponentielle du nombre de tests. Ce projet propose comme alternative de définir le réglage des soupapes comme un problème d'optimisation. Il identifie les simulations nécessaires au développement d'une méthode générique capable d'intégrer un nombre croissant de degrés de liberté afin de développer un outil de grande fidélité pour la prédiction des effets de différents dispositifs. Puisque la résolution d'un problème d'optimisation requiert l'évaluation de centaines de fonctions, l'utilisation directe d'une simulation hautefidélité demande des temps de calculs beaucoup trop longs. En remplacement, un réseau artificiel de neurones est entraîné avec des résultats de simulation haute-fidélité et utilisé pour représenter la réponse du moteur à différentes combinaisons de variables de contrôle, et ce avec des temps de calcul beaucoup plus courts. Cet article décrit une méthode basée sur la simulation, apporte des détails sur les techniques de modélisation haute-fidélité et démontre une application à un prototype de moteur à essence à quatre cylindres et double arbre à cames indépendant. Abstract

show abstract