Modeling the Gas Exchange Processes of a Modern Diesel Engine With an Integrated Physics-Based and Data-Driven Approach

Gonzalez, Jorge Pulpeiro; Ankobea-Ansah, King; Milian, Elena Escuder; Hall, Carrie M.

doi:10.1115/dscc2019-9226

Cited by 4 publications

(2 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…When there are only a small number of samples as encountered for the steady state ANN models, single-layer models could be adequate. This matches with previous research in Pulpeiro González et al, 19,23 where a simple single-layer ANN model fits the complex turbocharger model well. However, when there is a large number of samples as in the transient ANN models, multi-layer ANNs will be necessary to capture the nonlinear characteristics during the transient processes as shown in Figure A3.…”

Section: Resultssupporting

confidence: 91%

“…In Schaberg and Atkinson, 17 a data-driven, model-based calibration technique utilizing artificial neural networks (ANNs) was used to develop optimized transient and steady-state calibrations with both conventional diesel fuel, and neat gas-to-liquid diesel fuel. In Filipi et al, 18 ,Pulpeiro González et al, 19 and Kushwaha and Saraswati, 20 ANN models were used for VGT modeling in order to represent the VGT characteristics for any vane position. In these studies, ANNs are investigated and appear to be promising not only for modeling complicated air handling systems of modern diesel engines, but also for control and the improvement of efficiency and emissions.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Neural network-based air handling control for modern diesel engines

Qian

Huo

Hall

2022

Proceedings of the Institution of Mechanical Engineers, Part D:

View full text Add to dashboard Cite

Complex air handling systems, featuring technologies such as exhaust gas recirculation (EGR) and variable geometry turbochargers (VGTs), are commonly used in modern diesel engines to meet stringent emissions and fuel economy requirements. The control of diesel air handling systems with EGR and VGTs is challenging because of their nonlinearity and coupled dynamics. In this paper, artificial neural networks (ANNs) and recurrent neural networks (RNNs) are applied to control the low pressure (LP) EGR valve position and VGT vane position simultaneously on a light-duty multi-cylinder diesel engine. Intake manifold pressure (IMP) and air-fuel equivalence ratio, or [Formula: see text], are selected as the control objectives since they directly impact engine emissions and cylinder power output. Meanwhile both signals are available on production engines, so no additional hardware costs for measurement systems will be introduced. Both transient and steady state experimental data are separately used to train the two categories of neural networks (NNs). The NNs with minimum mean square error (MSE) for the training data sets are compared to conventional proportional-integral (PI) control and model predictive control (MPC). The most accurate NN controller has almost no overshoot during the transient process while the steady state error [Formula: see text] and IMP are at most 7.0% and 3.8%, respectively, under a wide range of engine speeds from 2500 to 4000 rpm, thus showing the potential of this approach.

show abstract

Section: Resultssupporting

confidence: 91%

Section: Introductionmentioning

confidence: 99%

Neural network-based air handling control for modern diesel engines

Qian

Huo

Hall

2022

Proceedings of the Institution of Mechanical Engineers, Part D:

View full text Add to dashboard Cite

show abstract

Hybrid Physics-Based and Data-Driven Modeling for Improved Standpipe Pressure Prediction

Erge

Oort

2021

Day 5 Fri, March 12, 2021

View full text Add to dashboard Cite

During drilling operations, it is common to see pump pressure spikes when flow is initiated, including after a connection or after a prolonged break in drilling operations. It is important to be able to predict the magnitude of such pressure spikes to avoid compromising wellbore integrity. This study shows how a hybrid approach using data-driven machine learning coupled with physics-based modeling can be used to accurately predict the magnitude of pressure spikes. To model standpipe pressure behavior, machine learning techniques were combined with physics-based models via a rule-based, stochastic decision-making algorithm. To start, neural networks and deep learning models were trained using time-series drilling data. From there, physics-based equations that model the pressure required to break the mud's gel strength as well as the flow of non-Newtonian fluids through the entire circulation system were used to simulate standpipe pressure. Then, these two highly different methods for predicting/modeling standpipe pressure were combined by a hidden Markov model using a set of rules and transition probabilities. By combining machine learning and physics-based approaches, the best features of each model are leveraged by the hidden Markov model, yielding a more accurate and robust prediction of pressure. A similar result is not achievable with a purely data-driven black-box model, because it lacks a connection to the underlying physics. Our study highlights how drilling data analysis can be optimally leveraged. The overarching conclusion: hybrid modeling can more accurately predict pump pressure spikes and capture the transient events at flow initiation when compared to physics-based or machine learning models used in isolation. Moreover, the approach is not limited to pressure behavior but can be applied to a wide range of well construction operations. The proposed approach is easy to implement and the details of implementation are presented in this study. Being able to accurately model and manage the pressure response during drilling operations is essential, especially for wells drilled in narrow-margin environments. Pressure can be more accurately predicted through our proposed hybrid modeling, leading to safer, more optimized operations.

show abstract