# Intervention of Artificial Neural Network with an Improved Activation Function to Predict the Performance and Emission Characteristics of a Biogas Powered Dual Fuel Engine

^{1}

^{2}

^{3}

^{4}

^{5}

^{6}

^{*}

## Abstract

**:**

_{x}). Estimated values and results of experiments were compared. The error analysis showed that the built model has quite accurately predicted the experimental results. This has been described by the value of Coefficient of determination (R

^{2}), which varies between 0.8493 and 0.9863 with the value of normalized mean square error (NMSE) between 0.0071 and 0.1182. The potency of the Nash-Sutcliffe coefficient of efficiency (NSCE) ranges from 0.821 to 0.8898 for BTE, HC, NO

_{x}and Smoke. This research has effectively emulated the on-board efficiency, emission, and combustion features of a dual-fuel biogas diesel engine taking the Swish activation mechanism in artificial neural network (ANN) model.

## 1. Introduction

^{2}value ranges from 0.99 and 1. Kakatia et al. [5] used log-sigmoid to forecast the output for Soot, HC, CO2, NO

_{x}, CO and BSFCeq, taking inputs as oxygen, methanol flow rate, diesel flow rate, and air flow rate etc. Hariharana et al. [6] carried out experiments to assess the effects of using hydrogen (H2) and Lemon Grass Oil (LGO) as a selective diesel replacement fuel, in a Compression Ignition (CI) engine with single-cylinder. The ANN model has been developed using a regular backpropagation algorithm to predict the association between engine performance responses and input factors (i.e., load, LGO and hydrogen). To forecast brake specific fuel consumption (BSFC), overall in-cylinder pressure and exhaust emissions, Agbulut et al. [7] used ANN. For BSFC, NO

_{x}, CO, HC, and CPmax the R

^{2}value obtained was 0.9995, 0.9999, 0.9902, 0.9990, and 0.9979 respectively.

^{2}values of 0.999978–0.999998. Tests conducted by Shukri et al. [11], indicated that the blend of diesel fuel with palm oil and methyl ester have improved the engine efficiency. For the in-cylinder pressure, heat release, thermal efficiency, and volume, the R

^{2}value of 0.996, 0.999, 0.989 and 0.998 was obtained respectively.

^{2}values of 0.99 were obtained; RMS values were lower than 0.02; and for test results, mean error percentage (MEP) values were lower than 2.7 percent. The sigmoid function was found to be the most commonly used activation function in models of the Artificial Neural Network in most of the studies so highlighted. The output varies from 0 to 1 for the sigmoid function, and from 0 to 0.25 for the derivatives of the sigmoid function. The Sigmoid is usually susceptible to the issue of vanishing gradient and method outcome is not zero-centered. In addition, the exponent and power operations make it costly to compute.

#### Motivating Factor for This Research Work

_{x}, Smoke. The researchers used the swish activation function to build a three layered ANN model.

## 2. Materials and Methods

^{3}capacity. Deenbandhu-based biogas plant has been traditionally used in provincial territories as cooking fuel. In a vault, the composite gas was stored and piped into the engine. To examine the physiochemical properties of fuels, the set standard of the American Society for Testing and Materials (ASTM) was used. Table 1 displays some associated fuel characteristics.

#### Experimental Setup

_{x}, and Smoke were measured using a Di-gas analyser (AVL 4000). The % volume has been used to note CO and Smoke, whereas, gm/kW.hr for both HC and NO

_{x}. In order to measure smoke exhalation, a diesel smoke metre (AVL 437) has been used; particularly the smoke opacity. In compliance with ASTM-D6522, the exhalations of gas are strictly regulated.

## 3. Application of Artificial Neural Network (ANN)

#### 3.1. Back Propagation and ANN for Current Study

_{x}and Smoke. Figure 2 demonstrates the ANN model used here in this study work and Table 4 lists the parameter values used in ANN.

#### 3.2. Swish Activation Function and Its Importance

#### 3.3. Count of Neurons for the Hidden Layer in ANN

#### 3.4. Selection among Sigmoid and Swish

_{x}, and Smoke, the value of RMSE was determined taking hidden neuron counts as 3, 5, and 7. ANN model employs sigmoid and swish activation function separately. Table 5 reveals that the RMSE values obtained for swish activation are lower relative to the commonly used sigmoid function, which proves swish, a better choice than the sigmoid. Also, when compared with rectified linear unit (ReLU) activation function, Swish is as effective as ReLU in computation, but demonstrates greater efficiency than ReLU. Swish values vary from infinity to infinity in the negative. The function curve is smooth and at all points the function is distinguishable, which is one of the reasons for outperforming swish from ReLU.

## 4. Model Evaluation

#### 4.1. Metrics for Evaluation

^{2}measure Equation (5), [25], which was found to be constrained by its intrinsic sensitivity to the expected and observed values of means and variances.

#### 4.2. Solver Architecture

^{2}, RMSE, NMSE, NRMSE, MSRE, NSCE, Theil, and KL values computed at 3, 5, and 7 hidden layer neurons using ANN with SWISH activation function.

_{x}and Smoke.

#### 4.3. Uncertainty Analysis for the Proposed Model

_{x}, and Smoke; considering load, bio-gas flow rate, n-butanol as data, which was obtained from the experimental results. In this study, the model’s predictability toward engine responsiveness showed good alignment with statistics of correlation. However, the complete uncertainty involved in measurement model derives from two distinct factors. One is the uncertainty of Theil that was considered in the development of the ANN model and the second is related to experimental tools. Total uncertainty estimation is seen in Table 7 using the Equation (12).

## 5. Results and Discussion

_{x}experimental outcomes. It exhibits remarkably low 0.0805 Theil uncertainty along with 0.0064 MSRE. Other statistical tests such as NMSE, NRMSE also displayed very low significance, i.e., 0.0071 and 0.0897. NSCE was observed as high as 88.98 percent in special error metrics, while KL-N was reported to be 0.080 which fulfilled the stronger compatibility of expected values with experimental findings. Figure 11 showed the similarity of forecasted smoke by the model presented with observed smoke through experiments. The values for Smoke MSRE, NMSE and NRMSE were 0.0625, 0.1182 and 0.1872, respectively. With Theil’s uncertainty as low as 0.1152, the model designed scored NSCE efficiency as high as 82.09 percent. In particular, a very low value of 0.0079 showed the KL-N divergence effectiveness, suggesting its good predictive accuracy.

## 6. Conclusions

^{2}as 0.8493 and maximum as 0.9863. Also the values for NMSE spans between 0.0071 to 0.1182. NSCE performance ranged from 0.821 to 0.8898 for BTE, HC, NO

_{x}and Smoke. The NSCE performance was found to range from 0.821 to 0.8898 for BTE, HC, NO

_{x}and Smoke. Therefore, it can be concluded that the on-board performance and exhaust characteristics of a dual-fuel biogas-diesel engine can be effectively simulated by the proven ANN model.

## Author Contributions

## Funding

## Acknowledgments

## Conflicts of Interest

## References

- Roy, S.; Banerjee, R.; Bose, P.K. Performance and exhaust emissions prediction of a CRDI assisted single cylinder diesel engine coupled with EGR using artificial neural network. Appl. Energy
**2014**, 119, 330–340. [Google Scholar] [CrossRef] - Banerjee, R.; Bose, P. Development of a neuro genetic algorithm based virtual sensing platform for the simultaneous prediction of NO
_{x}, opacity and BSFC in a diesel engine operated in dual fuel mode with hydrogen under varying EGR conditions. SAE Int. J. Engines**2012**, 5, 119–140. [Google Scholar] [CrossRef] - Chakraborty, A.; Roy, S.; Banerjee, R. An experimental based ANN approach in mapping performance-emission characteristics of a diesel engine operating in dual-fuel mode with LPG. J. Nat. Gas Sci. Eng.
**2016**, 28, 15–30. [Google Scholar] [CrossRef] - Akkouche, N.; Loubar, K.; Nepveu, F.; Kadi, M.E.A.; Tazerout, M. Micro-combined heat and power using dual fuel engine and biogas from discontinuous anaerobic digestion. Energy Convers. Manag.
**2020**, 205, 112407. [Google Scholar] [CrossRef] - Kakati, D.; Roy, S.; Banerjee, R. Development of an artificial neural network based virtual sensing platform for the simultaneous prediction of emission-performance-stability parameters of a diesel engine operating in dual fuel mode with port injected methanol. Energy Convers. Manag.
**2019**, 184, 488–509. [Google Scholar] [CrossRef] - Hariharan, N.; Senthil, V.; Krishnamoorthi, M.; Karthic, S. Application of artificial neural network and response surface methodology for predicting and optimizing dual-fuel CI engine characteristics using hydrogen and bio fuel with water injection. Fuel
**2020**, 270, 117576. [Google Scholar] [CrossRef] - Ağbulut, Ü.; Ayyıldız, M.; Sarıdemir, S. Prediction of performance, combustion and emission characteristics for a dual fuel diesel engine at varying injection pressures. Energy
**2020**, 197, 117257. [Google Scholar] [CrossRef] - Kurtgoz, Y.; Karagoz, M.; Deniz, E. Biogas engine performance estimation using ANN. Eng. Sci. Technol. Int. J.
**2017**, 20, 1563–1570. [Google Scholar] [CrossRef] - Leo, G.L.; Sekar, S.; Arivazhagan, S. Experimental investigation and ANN modelling of the effects of diesel/gasoline premixing in a waste cooking oil-fuelled HCCI-DI engine. J. Therm. Anal. Calorim.
**2020**, 141, 2311–2324. [Google Scholar] [CrossRef] - Shojaeefard, M.; Etghani, M.; Akbari, M.; Khalkhali, A.; Ghobadian, B. Artificial neural networks based prediction of performance and exhaust emissions in direct injection engine using castor oil biodiesel-diesel blends. J. Renew. Sustain. Energy
**2012**, 4, 063130. [Google Scholar] [CrossRef] - Shukri, M.R.; Rahman, M.; Ramasamy, D.; Kadirgama, K. Artificial Neural Network Optimization Modeling on Engine Performance of Diesel Engine Using Biodiesel Fuel. Int. J. Automot. Mech. Eng.
**2015**, 11, 2332–2347. [Google Scholar] [CrossRef] - Ashok Kumar, T.; Musthafa, M.M.; Chandramouli, R.; Kandavel, T.; Mohanraj, T.; Sridharan, G. Performance characteristics of a variable compression ratio CI engine simulation using artificial neural network. Energy Sources Part A Recover. Util. Environ. Eff.
**2019**, 23, 1–11. [Google Scholar] [CrossRef] - Yıldırım, S.; Tosun, E.; Çalık, A.; Uluocak, İ.; Avşar, E. Artificial intelligence techniques for the vibration, noise, and emission characteristics of a hydrogen-enriched diesel engine. Energy Sources Part A Recover. Util. Environ. Eff.
**2019**, 41, 2194–2206. [Google Scholar] [CrossRef] - Oğuz, H.; Sarıtas, I.; Baydan, H.E. Prediction of diesel engine performance using biofuels with artificial neural network. Expert Syst. Appl.
**2010**, 37, 6579–6586. [Google Scholar] [CrossRef] - Cay, Y. Prediction of a gasoline engine performance with artificial neural network. Fuel
**2013**, 111, 324–331. [Google Scholar] [CrossRef] - Shayler, P.; Goodman, M.; Ma, T. The exploitation of neural networks in automotive engine management systems. Eng. Appl. Artif. Intell.
**2000**, 13, 147–157. [Google Scholar] [CrossRef] - Papadimitriou, I.; Warner, M.; Silvestri, J.; Lennblad, J.; Tabar, S. Neural network based fast-running engine models for control-oriented applications. SAE Tech. Pap.
**2005**, 13, 214–233. [Google Scholar] - Mahla, S.K.; Ardebili, S.M.S.; Sharma, H.; Dhir, A.; Goga, G.; Solmaz, H. Determination and utilization of optimal diesel/n-butanol/biogas derivation for small utility dual fuel diesel engine. Fuel
**2021**, 289, 119913. [Google Scholar] [CrossRef] - Imtenan, S.; Masjuki, H.H.; Varman, M.; Fattah, I.M.R. Evaluation of n-butanol as an oxygenated additive to improve combustion-emission-performance characteristics of a diesel engine fuelled with a diesel-calophyllum inophyllum biodiesel blend. RSC Adv.
**2015**, 5, 17160–17170. [Google Scholar] [CrossRef] - Yang, F.; Cho, H.; Zhang, H.; Zhang, J.; Wu, Y. Artificial neural network (ANN) based prediction and optimization of an organic Rankine cycle (ORC) for diesel engine waste heat recovery. Energy Convers. Manag.
**2018**, 164, 15–26. [Google Scholar] [CrossRef] - Celebi, M.; Ceylan, M. The New Activation Function for Complex Valued Neural Networks: Complex Swish Function. In Proceedings of the 4th International Symposium on Innovative Approaches in Engineering and Natural Sciences, Samsun, Turkey, 22–24 November 2019. [Google Scholar]
- Hecht-Nielsen, R. Kolmogorov’s mapping neural network existence theorem. Proceedings of the international conference on Neural Networks. IEEE Press N. Y.
**1987**, 3, 11–14. [Google Scholar] - Belman-Flores, J.M.; Ledesma, S.; Garcia, M.; Ruiz, J.; Rodríguez-Muñoz, J.L. Analysis of a variable speed vapor compression system using artificial neural networks. Expert Syst. Appl.
**2013**, 40, 4362–4369. [Google Scholar] [CrossRef] - Gandomi, A.H.; Roke, D.A. Assessment of artificial neural network and genetic programming as predictive tools. Adv. Eng. Softw.
**2015**, 88, 63–72. [Google Scholar] [CrossRef] - Yassin, M.A.; Alazba, A.; Mattar, M.A. Artificial neural networks versus gene expression programming for estimating reference evapotranspiration in arid climate. Agric. Water Manag.
**2016**, 163, 110–124. [Google Scholar] [CrossRef] - Kshirsagar, C.M.; Anand, R. Artificial neural network applied forecast on a parametric study of Calophyllum inophyllum methyl ester-diesel engine out responses. Appl. Energy
**2017**, 189, 555–567. [Google Scholar] [CrossRef] - Chen, Z.; Yang, Y. Assessing Forecast Accuracy Measures; Technical Report 2004–10; Iowa State University, Department of Statistics & Statistical Laboratory: Ames, IA, USA, 2004. [Google Scholar]

**Figure 4.**Depiction for comparing NMSE, NRMSE and MSRE of the model taking (

**a**) 3 neurons; (

**b**) 5 neurons; and (

**c**) 7 neurons.

**Figure 5.**Illustration of R

^{2}and RMSE of the model for (

**a**) 3 neurons; (

**b**) 5 neurons; and (

**c**) 7 neurons.

**Figure 6.**Representation of values obtained with respect to Theil U2, NSE, and KL-divergence of the model for

**(a**) 3 neurons, (

**b**) 5 neurons, and (

**c**) 7 neurons.

**Figure 7.**(

**a**) Depiction of predicted and experimentally obtained BTE values taken for 20 test cases; (

**b**) Regression coefficient for the values predicted using ANN Vs experimental.

**Figure 8.**(

**a**) Depiction of predicted and experimentally obtained CO values taken for 20 test cases; (

**b**) Regression coefficient for the values predicted using ANN Vs experimental.

**Figure 9.**(

**a**) Depiction of predicted and experimentally obtained HC values taken for 20 test cases; (

**b**) Regression coefficient for the values predicted using ANN Vs experimental.

**Figure 10.**(

**a**) Depiction of predicted and experimentally obtained NO

_{x}values taken for 20 test cases; (

**b**) Regression coefficient for the values predicted using ANN Vs experimental.

**Figure 11.**(

**a**) Depiction of predicted and experimentally obtained Smoke values taken for 20 test cases; (

**b**) Regression coefficient for the values predicted using ANN Vs experimental.

**Table 1.**Various traits of the test fuel along with its ASTM based test scheme (# Not provided by the supplier).

Features | Diesel | n-Butanol | Test Scheme (ASTM) | Biogas | Test Scheme (ASTM) |
---|---|---|---|---|---|

Density (kg/m^{3})@15 °C | 840 | 810 | D4052 | 0.92 | D3588 |

Viscosity (mm^{2}/s)@40 °C | 2.72 | 3.64 | D445 | - | |

Heating Value (MJ/kg) | 42.6 | 33.2 | D4809 | 26.23 | D1945 |

Flash Point (°C) | 78 | 35 | D93 | - | |

Fire Point (°C) | 83 | 42 | D93 | ||

Cloud Point (°C) | −8 | # | DL500 | ||

Pour Point (°C) | −6 | −45 | D97 | ||

Cetane Number (CN) | 50 | 22 | D613 |

Chemical Formula | C_{4}H_{10}O |
---|---|

Boiling point | 117 (°C) |

Temperature for Auto ignition | 343 (°C) |

Octane number | 96 |

Oxygen | 21.62 (% by weight) |

Latent heat of vaporization at 25 °C | 626 (kJ/kg) |

Make | Kirloskar |
---|---|

Model Specifications | TV 1 |

System used for Cooling | Water Cooled |

Cylinder Count | 01 |

Rated Brake Power | 5.2 kW @ 1500 rpm |

Displacement volume | 661 (cc) |

Rated Speed | 1500 (rpm) |

Standard Fuel Injection Timing | 23° before TDC |

Bore × Stroke | 87.5 × 110 (mm) |

Compression Ratio | 17.5:1 |

Structure of the Network Used Here | 3 inputs, 01 hidden layer, and 5 outputs |

Percentage of Data used for training and testing | Training: 130 rows for training Testing: 20 rows for testing |

Type of the Network | Feed Forward Back Propagation |

Function used for Training | Backpropagation |

Optimization Function | Adam |

Transfer/Activation Function | Swish |

Criteria used to Stop | On-set of enhancement in validation error will results into breaking of training network |

**Table 5.**RMSE value for BTE, CO, HC, NO

_{x}, and Smoke computed using swish and sigmoid activation function taking number of neurons as 3, 5 and 7 at hidden layer of the ANN model.

O/P | Transfer Function (At Layer 1-2-3) | No. of Neurons | RMSE |
---|---|---|---|

BTE | swish | 3 | 1.741 |

5 | 3.927 | ||

7 | 3.288 | ||

sigmoid | 3 | 15.713 | |

5 | 15.693 | ||

7 | 15.691 | ||

CO | swish | 3 | 0.078 |

5 | 0.059 | ||

7 | 0.056 | ||

sigmoid | 3 | 0.08 | |

5 | 0.067 | ||

7 | 0.078 | ||

HC | swish | 3 | 0.286 |

5 | 0.304 | ||

7 | 0.307 | ||

sigmoid | 3 | 0.69 | |

5 | 0.61 | ||

7 | 0.57 | ||

NO_{x} | swish | 3 | 1.491 |

5 | 1.435 | ||

7 | 2.207 | ||

sigmoid | 3 | 16.94 | |

5 | 16.863 | ||

7 | 16.86 | ||

Smoke | swish | 3 | 2.448 |

5 | 3.154 | ||

7 | 3.170 | ||

sigmoid | 3 | 20.327 | |

5 | 20.326 | ||

7 | 20.31 |

**Table 6.**Computation of R

^{2}, RMSE, NMSE, NRMSE, MSRE, NSCE, Theil, KL using 3, 5, and 7 neurons at hidden layer with swish activation function while computing BTE, CO, HC, NO

_{x}, and Smoke.

O/P | TF (At Layer 1-2-3) | No. of Neurons | R^{2} | RMSE | NMSE | NRMSE | MSRE | NSCE | Theil | KL |
---|---|---|---|---|---|---|---|---|---|---|

BTE | swish + swish + swish | 3 | 0.9705 | 1.741 | 0.0119 | 0.1235 | 0.0193 | 0.8211 | 0.1047 | 0.0068 |

5 | 0.915 | 3.927 | 0.0716 | 0.2786 | 0.0405 | 0.0898 | 0.2361 | 0.008 | ||

7 | 0.943 | 3.288 | 0.0489 | 0.2332 | 0.0282 | 0.3618 | 0.1977 | 0.007 | ||

CO | 3 | 0.6576 | 0.078 | 0.2652 | 0.2609 | 0.1182 | 0.2497 | 0.3802 | 0.025 | |

5 | 0.8475 | 0.059 | 0.1332 | 0.199 | 0.0673 | 0.5635 | 0.29 | 0.018 | ||

7 | 0.9332 | 0.056 | 0.1182 | 0.1872 | 0.0625 | 0.6139 | 0.2727 | 0.013 | ||

HC | 3 | 0.9863 | 0.286 | 0.0938 | 0.1359 | 0.0673 | 0.8450 | 0.2294 | 0.016 | |

5 | 0.9038 | 0.304 | 0.1069 | 0.1441 | 0.0686 | 0.8256 | 0.2434 | 0.019 | ||

7 | 0.9006 | 0.307 | 0.1101 | 0.1457 | 0.0629 | 0.8218 | 0.246 | 0.016 | ||

NO_{x} | 3 | 0.9923 | 1.491 | 0.0077 | 0.0932 | 0.0067 | 0.8810 | 0.0837 | 0.027 | |

5 | 0.9288 | 1.435 | 0.0071 | 0.0897 | 0.0064 | 0.8898 | 0.0805 | 0.028 | ||

7 | 0.9813 | 2.207 | 0.0178 | 0.138 | 0.0129 | 0.7393 | 0.1239 | 0.04 | ||

Smoke | 3 | 0.8493 | 2.448 | 0.0147 | 0.1224 | 0.0125 | 0.8209 | 0.1152 | 0.0079 | |

5 | 0.9817 | 3.154 | 0.0216 | 0.1577 | 0.028 | 0.7027 | 0.1484 | 0.0057 | ||

7 | 0.9798 | 3.170 | 0.0299 | 0.1585 | 0.0211 | 0.6997 | 0.1491 | 0.0086 |

**Table 7.**Total uncertainty values obtained taking uncertainty values of the measuring instrument and the proposed ANN model.

Parameters | Values for the Uncertainties | Computation | Cumulative Uncertainties U _{T} (%) |
---|---|---|---|

BTE | 2.5, 0.11 | $\sqrt{{\left(2.5\right)}^{2}+{\left(0.11\right)}^{2}}$ | 2.5024 |

CO | 0.2, 0.27 | $\sqrt{{\left(0.2\right)}^{2}+{\left(0.27\right)}^{2}}$ | 0.3360 |

HC | 0.1, 0.23 | $\sqrt{{\left(0.1\right)}^{2}+{\left(0.23\right)}^{2}}$ | 0.2507 |

NO_{x} | 0.2, 0.08 | $\sqrt{{\left(0.2\right)}^{2}+{\left(0.08\right)}^{2}}$ | 0.2154 |

Smoke | 1, 0.11 | $\sqrt{{\left(1\right)}^{2}+{\left(0.11\right)}^{2}}$ | 1.0060 |

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |

© 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Arora, V.; Mahla, S.K.; Leekha, R.S.; Dhir, A.; Lee, K.; Ko, H.
Intervention of Artificial Neural Network with an Improved Activation Function to Predict the Performance and Emission Characteristics of a Biogas Powered Dual Fuel Engine. *Electronics* **2021**, *10*, 584.
https://doi.org/10.3390/electronics10050584

**AMA Style**

Arora V, Mahla SK, Leekha RS, Dhir A, Lee K, Ko H.
Intervention of Artificial Neural Network with an Improved Activation Function to Predict the Performance and Emission Characteristics of a Biogas Powered Dual Fuel Engine. *Electronics*. 2021; 10(5):584.
https://doi.org/10.3390/electronics10050584

**Chicago/Turabian Style**

Arora, Vinay, Sunil Kumar Mahla, Rohan Singh Leekha, Amit Dhir, Kyungroul Lee, and Hoon Ko.
2021. "Intervention of Artificial Neural Network with an Improved Activation Function to Predict the Performance and Emission Characteristics of a Biogas Powered Dual Fuel Engine" *Electronics* 10, no. 5: 584.
https://doi.org/10.3390/electronics10050584