Hybrid Machine Learning Approaches and a Systematic Model Selection Process for Predicting Soot Emissions in Compression Ignition Engines

Shahpouri, Saeid; Norouzi, Armin; Hayduk, Christopher; Rezaei, Reza; Shahbakhti, Mahdi; Koch, Charles Robert

doi:10.3390/en14237865

Cited by 20 publications

(16 citation statements)

References 44 publications

(42 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Nonetheless, this concept was previously applied especially in technical sciences to solve problems. For instance, vector methods were applied in: nanostructure analysis (Zhao et al 2021 ), forecasting in the energy sector (Yousaf et al 2021 ; Aslam et al 2021 ], e-business modelling (Sun et al 2021 ), machine learning (Shahpouri et al 2021 ), or issues related to the COVID-19 pandemic (Alghazzawi et al 2021 ; Scuttari et al 2021 ; Ting et al 2020 ).…”

Section: Discussionmentioning

confidence: 99%

Systemic analysis of a manufacturing process based on a small scale bakery

2022

View full text Add to dashboard Cite

The main aim of the article is to present two new innovative concepts of reliability of a functioning manufacturing system in the process of making bread in small-scale bakeries. Reliability is understood as one of the representations of an operator acting on specific streams in time t o to t. One of these represents the global reliability of a system as a function of parallel action of all the streams of the system in time t o to t and is denoted as P g (t) . The second representation of reliability is a scalar value, P ss It shows a new function of global reliability of a manufacturing process as a product of system stream reliability. In order to plot the flow of the manufacturing process’s global reliability function, we need to perform detailed calculations, computations, and analysis of the differences of individual values in real time, as well as plan an algorithm of the flow of system streams. This needs a lot of effort, translating however, to a detailed picture of the process. In the analysed example, measurements and research revealed an important increase of the value of reliability in a transition from a traditional to a robotised bakery. The article also presents a new concept of the reliability of a technological process, based on the analysis of relations of elements of the following streams: energy, matter, information, time, and finances. It shows the method of specifying streams and the method for defining the reliability of important and supportive relations. Important relations between stream elements are defined as having the reliability value of one in time. Supportive relations bear the reliability within a continuum between zero and one. Important relations are designated based on research, experience, and knowledge. Stream systemic reliability P ss is a scalar value, i.e. a number from the continuum between zero and one. The P ss value characterises failure-free operation of the whole system. Its average value in the normative time t n expresses the efficiency of the manufacturing system. The value P ss is a quotient of the number of important relation and the sum of important and supportive relations. The formula for P ss shows the method of optimising the process through the increasing of the number of important relations between the input stream components. The concept has been applied to study the efficiency of operation of a small-scale bakery. Systemic analysis of a bakery allows for important increase in the reliability of baking bread if robotisation has been implemented. The concept of systemic-stream reliability P ss may be applied to analyse the efficiency of any technological process and optimisation of any manufacturing process.

show abstract

Section: Discussionmentioning

confidence: 99%

Systemic analysis of a manufacturing process based on a small scale bakery

2022

View full text Add to dashboard Cite

show abstract

“…This is the main reason that physics-based based models, such as detailed 3D combustion models, which offer a high prediction accuracy, have seen limited MPC implementation due to high computational times. These models first need to be simplified or linearized to be feasible for MPC implementation (Shahpouri et al, 2021a;Norouzi et al, 2019).…”

Section: Introductionmentioning

confidence: 99%

Machine Learning Integrated with Model Predictive Control for Imitative Optimal Control of Compression Ignition Engines

Norouzi¹,

Shahpouri²,

Gordon³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

The high thermal efficiency and reliability of the compression-ignition engine makes it the first choice for many applications. For this to continue, a reduction of the pollutant emissions is needed. One solution is the use of machine learning (ML) and model predictive control (MPC) to minimize emissions and fuel consumption, without adding substantial computational cost to the engine controller. ML is developed in this paper for both modeling engine performance and emissions and for imitating the behaviour of an Linear Parameter Varying (LPV) MPC. Using a support vector machine-based linear parameter varying model of the engine performance and emissions, a model predictive controller is implemented for a 4.5 Cummins diesel engine. This online optimized MPC solution offers advantages in minimizing the NO x emissions and fuel consumption compared to the baseline feedforward production controller. To reduce the computational cost of this MPC, a deep learning scheme is designed to mimic the behavior of the developed controller. The performance in reducing NOx emissions at a constant load by the imitative controller is similar to that of the online optimized MPC compared to the Cummins production controller. In addition, the imitative controller requires 50 times less computation time compared to that of the online MPC optimization.

show abstract

“…Physics-based modeling provides accurate models based on physical insight into the system, but engine combustion model and emissions detailed 3D model are computationally expensive [34,35]. This makes physical models impractical for real-time model-based control.…”

Section: Introductionmentioning

confidence: 99%

“…These models can be as accurate as physical models while requiring significantly less computational time. This is desired for implementation of the model-based controllers in ECUs [34]. Using ML in modeling has been successfully implemented for diesel engine using support vector machine (SVM) [11,12,36], neural network [14,37], and Extreme Learning Machine (ELM) [38,39].…”

Section: Introductionmentioning

confidence: 99%

“…The main contributions of this paper are The structure of the paper is shown in Figure 1 and it includes logging data, data-driven model development, and controller implementation using ESM co-simulation. First step including experimental data collection and detailed physical modeling GT-power was done in our previous papers [47,34]. Then, randomly generated inputs are fed into the GT-power model and the output data is recorded for modeling.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning based Model Predictive Control for Compression Ignition Engines

Norouzi¹,

Shahpouri²,

Gordon³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Machine learning (ML) and nonlinear model predictive control (NMPC) are used in this paper to minimize the emissions and fuel consumption of a compression ignition engine. In this work machine learning is used for two applications. In the first application, ML is to identify a model for implementation in model predictive control optimization problems. In the second application, ML is used as a replacement of a model predictive controller where the ML controller learns the optimal control action by imitating or mimicking the behavior of the model predictive controller. In this study, a deep recurrent neural network including long-short term memory (LSTM) layers are used to model the emission and performance of a 4.5 liter 4-cylinder Cummins compression ignition engine. This model is then used for model predictive controller implementation. Then, a deep learning scheme is deployed to clone the behavior of the developed controller. In the LSTM integration, a novel scheme is used by augmenting hidden and cell states of the network in an NMPC optimization problem. The developed LSTM-NMPC and the imitative NMPC are compared with the calibrated Cummins Engine Control Unit (ECU) model in an experimentally validated engine simulation platform. Results show a significant reduction in Nitrogen Oxides (NO x ) emissions and a slight decrease in the injected fuel quantity while maintaining the same load. In addition, the imitative NMPC has a similar performance as the NMPC but with two orders of magnitude less computation time.

show abstract

Hybrid Machine Learning Approaches and a Systematic Model Selection Process for Predicting Soot Emissions in Compression Ignition Engines

Cited by 20 publications

References 44 publications

Systemic analysis of a manufacturing process based on a small scale bakery

Systemic analysis of a manufacturing process based on a small scale bakery

Machine Learning Integrated with Model Predictive Control for Imitative Optimal Control of Compression Ignition Engines

Deep Learning based Model Predictive Control for Compression Ignition Engines

Contact Info

Product

Resources

About