Table 4 - uploaded by Amin Paykani
Content may be subject to copyright.
Source publication
Nondominated sorting genetic algorithm II (NSGA-II) is well known for engine optimization problem. Artificial neural networks (ANNs) followed by multi-objective optimization including a NSGA-II and strength pareto evolutionary algorithm (SPEA2) were used to optimize the operating parameters of a compression ignition (CI) heavy-duty diesel engine. F...
Context in source publication
Context 1
... 5−7 show the network outputs versus experimental values for test data points for three different engine outputs. The R, RMSE and MRE values for the NO x , smoke and BSFC are listed in Table 4. Consistency between the predicted values and the measured values implies that the 48 experimental values for every output parameter are large enough to develop accurate ANNs when the engine speed, load and injection timing are selected as input parameters and tan-sigmoid/linear as transfer functions. ...
Similar publications
The performance and emission of a single cylinder four stroke variable compression multi fuel engines when fueled with 20%, 25% and 30% of Karanja blended with diesel are investigated and compared with standard diesel. Experiment has been conducted at compression ratios of 15:1, 16:1, 17:1, and 18:1. The impact of compression ratio on fuel consumpt...
Citations
... In the work [58], to solve the engine optimization problem, a multi-layer perception (MLP) neural network followed by multi-objective optimization including a non-dominated sorting genetic algorithm II (NSGA-II) and strength Pareto evolutionary algorithm (SPEA2) were used. This study allowed the authors to decide which algorithm is preferable for optimizing engine emissions and fuel consumption. ...
An approach based on an artificial neural network (ANN) for the prediction of NOx emissions from underground load-haul-dumping (LHD) vehicles powered by diesel engines is proposed. A Feed-Forward Neural Network, the Multi-Layer Perceptron (MLP), is used to establish a nonlinear relationship between input and output layers. The predicted values of NOx emissions have less than 15% error compared to the real values measured by the LHD onboard monitoring system by the standard sensor. This is considered quite good efficiency for dynamic behaviour prediction of extremely complex systems. The achieved accuracy of NOx prediction allows the application of the ANN-based "soft sensor" in environmental impact estimation and ventilation system demand planning, which depends on the number of working LHDs in the underground mine. The proposed solution to model NOx concentrations from mining machines will help to provide a better understanding of the atmosphere of the working environment and will also contribute to improving the safety of underground crews.
... For the first model (denoted by ANN1), the utilized algorithm and the functions suitable for creating Shallow Neural Networks are used. The number of neurons and hidden layers started from low and increased one by one to find the best value according to the procedure in the previous study [49]. Thirty nodes in each hidden layer have been found to yield the best results. ...
Flamelet Generated Manifold (FGM) is an example of a chemistry tabulation or a flamelet method that is under attention because of its accuracy and speed in predicting combustion characteristics. However, the main problem in applying the model is a large amount of memory required. One way to solve this problem is to apply machine learning (ML) to replace the stored tabulated data. Four different machine learning methods, including two Artificial Neural Networks (ANNs), a Random Forest (RF), and a Gradient Boosted Trees (GBT), are trained, validated, and compared in terms of various performance measures. The progress variable source term and transport properties are replaced with the ML models. Particular attention was paid to the progress variable source term due to its high gradient and wide range of its value in the control variables space. Data preprocessing is shown to play an essential role in improving the performance of the models. Two ensemble models, namely RF and GBT, exhibit high training efficiency and acceptable accuracy. On the other hand, the ANN models have lower training errors and take longer to train. The four models are then combined with a one-dimensional combustion code to simulate a counterflow non-premixed diffusion flame in engine-relevant conditions. The predictions of the ML-FGM models are compared with detailed chemical simulations and the original FGM model for key combustion properties and representative species profiles.
... The advantage of these methods is that the final solution cannot easily become entrapped in the local optimum solution due to the random and stochastic nature of their operators. Therefore they are very appropriate in non-linear combustion systems (as it was used in [52]) in which the solution is prone to be entrapped in the local optimum. The GA starts with generating a random population. ...
In the Computational Fluid Dynamics (CFD) simulation of advanced combustion systems, the chemical kinetics must be examined in detail to predict the emissions and performance characteristics accurately. Nevertheless, the combustion simulation with detailed chemical kinetics is complicated because of the number of equations and a broad timescale spectrum. The Flamelet-Generated Manifold (FGM) is one of the examples of tabulation methods that has received much attention in recent years due to its fast and accurate prediction of combustion characteristics. The Progress Variable (PV) definition in FGM and other PV-based tabulated approaches is often selected randomly or depending on the user's experience. When complicated combustion systems are involved, such choices can become extremely difficult. In the current work, a generic approach for formulating a global PV is developed and tested in various operating conditions relevant to combustion engines. The method is based on a genetic algorithm optimization to maximize the monotonicity of PV, ensuring that for each value of PV, the dependent thermophysical properties have unique values. The FGM model's ability to reproduce the detailed kinetics evolution of the essential combustion and emission parameters of a non-premixed diffusion flame in Spray A configuration is evaluated in both one-dimensional counterflow and CFD simulation. It is concluded that with the use of the current approach, important combustion characteristics can be predicted much better compared to non-optimized PV while eliminating the manual selection of PV definition by the user. Since the algorithm needs to be executed before the chemistry tabulation in the pre-processing step, it does not increase the runtime of the FGM simulation. The algorithm only needs a few minutes to be finished on a standard desktop. The improvement in the results and the distribution of the values of important species in the computational domain is examined.
... WER can be implemented through thermomechanical regeneration (no chemical reactions), such as organic Rankine cycle (ORC) [39,40], Brayton cycle [41], turbocharger [42], etc. WER also can be employed using thermochemical regeneration, such as fuel reforming (with chemical reactions). This wasted thermal energy (exergy), can be reduced in advanced combustion strategies like homogeneous charge compression ignition (HCCI) [43] or reactivity-controlled compression ignition (RCCI) [44,45] (see Fig. 1), since these combustion strategies have lower exergy destruction than conventional diesel combustion due to higher proportions of premixed combustion with more homogeneous in-cylinder distributions of equivalence ratio and gas temperature [46,47]. The exergy destruction related to the combustion occurring inside the engine cylinder will be discussed in the following paragraph. ...
There has been a growing demand to develop new energy conversion devices with high efficiency and very low emissions for both power and propulsion applications in response to the net zero-carbon emission targets by 2050. Among these technologies, solid oxide fuel cells (SOFCs) have received attention due to their high electrical efficiency (above 60%), fuel flexibility, low-emission, and high-grade waste heat, which makes them particularly suitable for a large number of applications for power and propulsion systems. The higher operating temperatures make SOFCs suitable candidates for integration with an additional power generation device such as an internal combustion engine (ICE) by (a) using the residual fuel of the anode off-gas in the engine, which further increases overall system efficiency to values exceeding 70%, (b) decreasing combustion inefficiencies and (c) increasing waste heat recovery. This paper reviews the published work on hybrid SOFC-ICE systems considering various configurations. It has been found that integrated SOFC-ICE systems are promising candidates over conventional engines and stand-alone SOFCs to be used in stationary power generation and heavy-duty applications (e.g., marine and locomotive propulsion systems). The discussion of the present review paper provides useful insights for future research on hybrid electrochemical-combustion processes for power and propulsion systems.
... Genetic Algorithms (GA) have found applications over a diverse range of industrial and scientific fields, due to their ability to accurately approximate the set of solutions on multimodal problems with multi-objectives and large search spaces. Examples include: medical and biological data classification for Parkinson, diabetes and cancer sicknesses diagnosis [5,6]; adjustment of beam-angle in radiotherapy [7]; DNA sequencing [8] machine learning and artificial intelligence for intelligent stock trading and market forecasting [9]; control systems for greenhouse gases and fuel consumption reduction [10]; design of engineering artefacts, such as space satellite antenna [1];training of neural network for feature selection [11]; and job shop scheduling [12]. Utilisation of genetic algorithms often leads to unusual, but more effective solutions that do not follow the field-related experience, with example in Fig. 1.1, further showing their usefulness. ...
... A recent review by Zhou et al [2] lists 50 major applications for evolutionary algorithms. This is augmented by recent high profile examples that utilise these algorithms from a number of different fields of study, including architecture, bioinformatics, computational science, evolutionary theory, environmental science and materials engineering ( [3][4][5][6][7][8][9][10][11][12][13][14]). The problems being solved are becoming more complex and therefore promising methods for the improvement of genetic algorithms should be further explored and documented. ...
... NSGA-II has also been improved since the CEC ′ 09 competition, therefore the new results have been run and the tables have been updated to reflect this. NSGA-II is preferred over NSGA-III [24] or U-NSGA-III [10] in this comparison as it has been shown to exhibit better performance on bi-objective problems. Results for MOEA/D on the constrained problems are not included in the original CEC ′ 09 benchmark and so it is benchmarked on these problems and the tables are updated. ...
The current state-of-the-art of genetic algorithms is dominated by high-performing specialistsolvers with fast convergence. These algorithms require prior knowledge about the characteristics of the optimised problem to operate effectively, although such information is not available in most real cases. Most of these algorithms are only tested on a narrow range of similar benchmarking problems with lower complexity than the real cases. This leads to the promotion of high-performing strategies for these cases, but which might prove to be ineffective on practical applications. This hypothesis is supported by a low uptake of the current specialist/convergence algorithms on real-world cases; NSGA-II remains the most popular algorithm despite being developed in 2002. It is suggested that this is due to its uniformly good performance across a wide range of problems with distinct characteristics, indicating high generality, and a high diversity retention across iterations. To assess if increasing the generality and diversity of the search improves the performance on real problems, the Multi-Level Selection Genetic Algorithm (MLSGA) is extended to develop a “diversity-first” general-solver genetic algorithm. It is selected as it shows high promise for the diversity-oriented methodology. Firstly, the reasons behind why it exhibits high diversity are investigated, as it is shown that the collective-level mechanisms create additional evolutionary pressure, while the fitness separation approach leads to collectives targeting different regions of the search space. This creates unique region-based search which leads to retention of higher diversity of solutions between generations. Secondly, as the MLSGA is exhibiting a poor convergence, the algorithm is combined with the current state-of-the-art algorithms in the hybrid approach (MLSGA-hybrid) to offset this problem. MLSGA-hybrid focuses on increasing the convergence of the search, over the original MLSGA algorithm, while retaining its emphasis on the diversity. The results demonstrate that this improvement leads to top performance on a range of problems. This is particularly the case on constrained problems indicating that the diversity has been retained. Thirdly, the co-evolutionary variant is introduced and tested (cMLSGA), which combines multiple evolutionary algorithms to improve the generality of the method. To validate the performance of the “diversity-first” general-solver approach, the algorithm is tested on 100 benchmarking problems and compared with top algorithms from the current state-of-the-art. It is shown that cMLSGA is the best iv general-solver, due to the most robust performance across the evaluated cases, while maintaining a higher focus on problems where elevated diversity of the search is preferred, such as discontinuous, constrained and biased cases. Finally, the cMLSGA approach is benchmarked on 3 engineering cases with a wide range of diverse characteristics and compared with other leading genetic algorithms. It is shown that the convergence-oriented solvers are ineffective for real-world applications due to higher complexity of practical problems, whereas performance of specialist-solvers is low due differences between real-world cases and benchmarking functions they are adjusted to. According to that, the “diversity-first” genetic algorithms with a high generality are preferred and there should be more focus on algorithms with these characteristics in the future.
... In particular, hybrid MOEAs have been trained and optimized structures from introduced recurrent neural networks. Kakaee et al. 17 published a method of using ANNs followed by multi-objective optimization using NSGA-II evolution algorithm and SPEA2 optimization algorithm to optimize the operating parameters of a compression ignition heavy-duty diesel engine. Vieira and Tome 18 published two different methods to increase the search speed of the multi-objective evolution algorithm (MOEA) using ANNs. ...
In this study, a new methodology, hybrid Strength Pareto Evolutionary Algorithm Reference Direction (SPEA/R) with Deep Neural Network (HDNN&SPEA/R), has been developed to achieve cost optimization of stiffness parameter for powertrain mount systems. This problem is formalized as a multi-objective optimization problem involving six optimization objectives: mean square acceleration of a rear engine mount, mean square displacement of a rear engine mount, mean square acceleration of a front left engine mount, mean square displacement of a front left engine mount, mean square acceleration of a front right engine mount, and mean square displacement of a front right engine mount. A hybrid HDNN&SPEA/R is proposed with the integration of genetic algorithm, deep neural network, and a Strength Pareto evolutionary algorithm based on reference direction for multi-objective SPEA/R. Several benchmark functions are tested, and results reveal that the HDNN&SPEA/R is more efficient than the typical deep neural network. stiffness parameter for powertrain mount systems optimization with HDNN&SPEA/R is simulated, respectively. It proved the potential of the HDNN&SPEA/R for stiffness parameter for powertrain mount systems optimization problem.
... After training the neural network, the network has the ability to be applied in the genetic optimization algorithm, and then used to predict specific fuel consumption, RI, and maximum pressure at different combinations of initial pressure and injection timing. It should be noted that, out of the 70 data sets generated by the CFD model, 15% are used as validation data to prevent the networks from being overtrained [43,44]. Also, another 15% are kept away as test data and are not included in the training process in order to measure the accuracy of the generated networks. ...
... But the algorithm using ANNs can be highly effective with chaotic components. Recently, some researchers like Shen et al. (2018), Smith et al. (2014), Kakaee et al. (2015), and Vieira and Tome (2005) used ANNs to predict optimal targets in different areas. ...
This article presents a new control method based on fuzzy controller, time delay estimation, deep learning, and non-dominated sorting genetic algorithm-III for the nonlinear active mount systems. The proposed method, intelligent adapter fractions proportional–integral–derivative controller, is a smart combination of the time delay estimation control and intelligent fractions proportional–integral–derivative with adaptive control parameters following the speed range of engine rotation via the deep neural network with the optimal non-dominated sorting genetic algorithm-III deep learning algorithm. Besides, we proposed optimal fuzzy logic controller with optimal parameters via particle swarm optimization algorithm to control reciprocal compensation to eliminate errors for intelligent adapter fractions proportional–integral–derivative controller. The control objective is to deal with the classical conflict between minimizing engine vibration impacts on the chassis to increase the ride comfort and keeping the dynamic wheel load small to ensure the ride safety. The results of this control method are compared with that of traditional proportional–integral–derivative controller systems, optimal proportional–integral–derivative controller parameter adjustment using genetic algorithms, linear–quadratic regulator control algorithms, and passive drive system mounts. The results are tested in both time and frequency domains to verify the success of the proposed optimal fuzzy logic controller–intelligent adapter fractions proportional–integral–derivative control system. The results show that the proposed optimal fuzzy logic controller–intelligent adapter fractions proportional–integral–derivative control system of the active engine mount system gives very good results in comfort and softness when riding compared with other controllers.
... Genetic Algorithms (GA) are utilised across a range of different fields of study: reduction in greenhouse gases and fuel consumption [1], discovering rules from large biological datasets [2], beam-angle-optimisation in radiotherapy [3], DNA sequencing [4], training of neural network for feature selection [5], control systems for hydraulic wind turbines [6] and job shop scheduling [7]. The demand has led to the development of a diverse range of genetic algorithm methodologies with varying performance across different types of search spaces [8]. ...
p>There is a vibrant community devoted to developing novel Genetic Algorithms each year. The effectiveness of these algorithms is normally evaluated based upon a selected set of artificial functions self-selected by the same community. These sets are generated to reduce the complexity seen in real applications so they can be run in a computationally short period of time, so that the final answer is known and so that certain characteristics can be isolated, giving more information about the strengths and weaknesses of a given methodology. In particular, the literature is dominated by a wide range of continuous and unconstrained problems, which are dominated by only one characteristic. This leads to a bias in the current Genetic Algorithms towards a set of specialist solvers for these problems, dominated by convergence enhancing mechanisms, as the success of the current state-of-the-art is directly linked to them. In this paper, the success of these specialist solvers is determined on two different engineering problems to verify the performance of a number of specialist and general genetic algorithms. The results show that the general-solvers exhibit better performance on the engineering problems with and without constraints. It is concluded that more emphasis should be given to general solvers development and the development of new benchmarking problems should be broader in scope and contain problems with many different characteristics.</p
... Among all the technological advancements, CRDI system has spearheaded the technological renaissance [3][4][5]. However, while CRDI system showed a significant improvement of fuel consumption and soot emission, a propensity of increased nitrogen oxide (NOx) formation was always accompanied [4][5][6], which was considered as a performance and emissions trade-off situation. ...
... In contrast, when dealing with the multi-objective optimization problems, Non-dominated sorting genetic algorithm II (NSGA-II), proposed by Deb [24], can find a set of non-dominated solutions (Pareto-optimal set), where the non-dominated solutions always perform better on at least on criterion than others. In addition, optimization approach using NSGA-II has also considered the optimality and diversity of the optimal solutions [6]. For example, NSGA-II was implemented for reducing CO and NOx emission from a direct injection dual-fuel engine by Lotfan et al. [25], and Etghani et al. developed a hybrid method of modified NSGA-II and TOPSIS to optimize performance and emissions of a diesel engine using biodiesel [26]. ...
... In the optimization process of previous multi-objective optimization approaches, including GA, PSO and NSGA-II, there are always many operating points of engines need to be evaluated, which is impossible to evaluate from the engine experimental test. Thus, the engine models have to be employed, in the previous studies, GT-Power models [21,27] which based on the physical modeling, and ANN [6] which based on the empirical modeling, have been used for the engine optimization. However, when the physical models are used for the optimization process, due to the larger amount of computations, the process usually takes much time. ...
The multi-objective optimization problems of diesel engines are always challenging for the engineers, especially with the application of new technologies that aim at improving the engine performance and emissions. This paper proposed a novel online optimization approach using NSGA-II coupled with a machine learning method (SVM). The proposed online optimization approach was conducted based on an engine physical model, which was calibrated and validated carefully using experimental data. In the optimization process, the engine physical model is used as a substitute of real engine to generate training data for the SVM and validate the accuracy of the optimization results; SVM, with fast computing speed, undertakes the massive calculating workloads of fitness evaluation on searching the Pareto optimal solutions. Moreover, this paper proposed an enhancing training method to guarantee the accuracy of SVM model. When applying on a marine diesel engine, the proposed online optimization approach has demonstrated its reliability and high efficiency. In addition, with fast computing speed, the well trained SVM model can develop the engine responses maps rapidly. Eventually, based on the Pareto-optimal solutions obtained by the proposed optimization approach, combining with the maps, the solving of multi-objective optimization problems will be significantly facilitated.
