The Use of Neural Nets for Matching Fixed or Variable Geometry Compressors With Diesel Engines

Nelson, Sigurd A.; Filipi, Zoran; Assanis, Dennis N.

doi:10.1115/1.1563239

Cited by 22 publications

(9 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A neural network as used by Nelson et al 13 and determined to be ill-suited to compressor modeling due to the large number of required coefficients and the limited number of data points available to train the neural network.…”

Section: Scalable Modeling Approachmentioning

confidence: 99%

Scalable turbocharger performance maps for dynamic state-based engine models

Bell

Zimmerle

Bradley

et al. 2016

International Journal of Engine Research

View full text Add to dashboard Cite

Adapting turbocharger performance maps to a form suitable for dynamic simulations is challenging for the following reasons: (1) the amount of available data is typically limited, (2) data are typically not provided for the entire operating range of the compressor and turbine and (3) the performance data are non-linear. To overcome these challenges, curve fits are typically generated using the performance data individually for each device. The process, however, can take uneconomical amounts of effort to implement for a range of compressors and turbines. This article introduces a method to implement non-dimensional performance maps thereby allowing a range of turbochargers to be modeled from the same performance data, reducing the effort required to implement models of different sizes. The non-dimensional maps seek to model the performance of compressor and turbine families in which the geometry of the rotor and housing are similar and allow the turbocharger to be scaled for simulation in much the same way used to design customized sizes of turbochargers. A method to match the non-dimensional compressor map to engine performance targets by selecting the compressor diameter is presented, as well as a method to match the turbine to the selected compressor.

show abstract

Section: Scalable Modeling Approachmentioning

confidence: 99%

Scalable turbocharger performance maps for dynamic state-based engine models

Bell

Zimmerle

Bradley

et al. 2016

International Journal of Engine Research

View full text Add to dashboard Cite

show abstract

“…Other modeling solutions that use the PR instead of ψ for estimating Φ were also investigated, in the form of parabolic functions 8 or piece-wise functions. 9 Apart from Φ -based models, a neural network approach 10 was also investigated, but was later shown in Paul and Kolmanovsky 11 to be not very effective in predicting mass flow rate. For better estimation over a wider PR range, Oskar and Eriksson 12 proposed an elliptic function–based piece-wise model.…”

Section: Introductionmentioning

confidence: 99%

A physics-based turbocharger model for automotive diesel engine control applications

Song

Upadhyay

Xie

2018

Proceedings of the Institution of Mechanical Engineers, Part D:

View full text Add to dashboard Cite

Control-oriented models of turbocharger processes such as the compressor mass flow rate, the compressor power, and the variable geometry turbine power are presented. In a departure from approaches that rely on ad hoc empirical relationships and/or supplier provided performance maps, models based on turbomachinery physics and known geometries are attempted. The compressor power model is developed using Euler’s equations of turbomachinery, where the gas velocity exiting the rotor is estimated from an empirically identified correlation for the ratio between the radial and tangential components of the gas velocity. The compressor mass flow rate is modeled based on mass conservation, by approximating the compressor as an adiabatic converging-diverging nozzle with compressible fluid driven by external work input from the compressor wheel. The variable geometry turbine power is developed with Euler’s equations, where the turbine exit swirl and the gas acceleration in the vaneless space are neglected. The gas flow direction into the turbine rotor is assumed to align with the orientation of the variable geometry turbine vane. The gas exit velocity is calculated, similar to the compressor, based on an empirical model for the ratio between the turbine rotor inlet and exit velocities. A power loss model is also proposed that allows proper accounting of power transfer between the turbine and compressor. Model validation against experimental data is presented.

show abstract

“…Canova et al 7 proposed a relationship between the JK model empirical coefficients and compressor geometry parameters, such as rotor diameter, Trim, and area/radius (A/R) ratio. Apart from c-based models, a neural network approach 8 was also investigated, which was later shown by Paul and Kolmanovsky 9 to be not very effective in predicting mass flow rate. For better estimation over a wider pressure ratio range, Leufven and Eriksson 10 proposed an elliptic function-based piecewise model.…”

Section: Introductionmentioning

confidence: 99%

A physics-based zero-dimensional model for the mass flow rate of a turbocharger compressor with uniform/distorted inlet condition

Song

Zhao

Sun

et al. 2018

International Journal of Engine Research

View full text Add to dashboard Cite

Turbocharger compressor, when fitted to a vehicle, usually operates with a curved inlet pipe which leads to distorted inlet flow field, hence deteriorating compressor flow capability. During the measurement of compressor performance, turbocharger-engine matching and controller design, the inlet flow field is, however, assumed to be uniform, which deviates from the real-world conditions. Consequently, the overall system performance could be compromised if the inlet distortion effect is ignored. To address this issue, in this article, a turbomachinery physics-based zero-dimensional model was proposed for the mass flow rate of a compressor with distorted inlet flow field due to 90° and 180° bent inlet pipe. The non-uniform flow is approximated as two-zone flow field, similar to parallel compressors, with the total pressure deviation between two zones modeled as a function of the flow velocity and pipe geometry. For each flow zone, the corresponding mass flow rate is estimated by approximating each sub-compressor as an adiabatic nozzle, where the fluid is driven by external work delivered by a compressor wheel governed by Euler’s turbomachinery equation. By including turbomachinery physics and compressor geometry information into the modeling, the model achieves high fidelity in compressor map interpretation and extrapolation, which is validated in experiments and the three-dimensional computational fluid dynamic simulation.

show abstract

The Use of Neural Nets for Matching Fixed or Variable Geometry Compressors With Diesel Engines

Cited by 22 publications

References 18 publications

Scalable turbocharger performance maps for dynamic state-based engine models

Scalable turbocharger performance maps for dynamic state-based engine models

A physics-based turbocharger model for automotive diesel engine control applications

A physics-based zero-dimensional model for the mass flow rate of a turbocharger compressor with uniform/distorted inlet condition

Contact Info

Product

Resources

About