Ultrasound assisted synthesis of water-in-oil nanoemulsions: Parametric optimization using hybrid ANN-GA approach

https://doi.org/10.1016/j.cep.2019.107649Get rights and content

Highlights

  • Novel approach adopted for the formulation and optimization of water-in-diesel oil nanoemulsion.

  • A collaborated modeling approach introduced FFBPANN-GA coupled with RCCD-RSM.

  • Multi objective hybrid GA ascertained robust than conventional methods.

  • Enhancement in nanoemulsion properties incorporated by optimizing dzavg. and kinematic viscosity.

  • Proposed scheme reduced water and surfactant fraction, power requirement and cost.

Abstract

Present study deals with the newly investigated CEMNSE (combined energy mixed surfactant nanoemulsion) method for optimizing the operating parameters concerned with the formation of water-in-oil nanoemulsions (W/O NE). The formulation process was intensified by optimizing the operating parameters of CEMNSE method by minimizing functions of two response variables viz. avg. droplet size (nm) and kinematic viscosity (mm2.s−1). A combined approach of ultrasonic cavitation and isothermal dilution method is used in formulating W/O NE. Optimization is carried out with an integral hybrid genetic algorithm (GA) with back propagation artificial neural network (BPANN) and response surface methodology (RSM) based on rotatable central composite design (RCCD). Combined approach process parameters as input to the proposed models are water fraction (0.05-0.11, w/w), surfactant fraction (0.10-0.020, w/w), power density (21.25–46.75, W. cm-2), and ultrasonication time (4–10, min.). Hybrid GA model predicted optimum values of avg. droplet size and kinematic viscosity as 53.54 nm and 1.459 mm2.s−1, respectively, with errors < 2.2%. However, optimized process parameters predicted as water fraction-0.052 (w/w), surfactant fraction-0.105 (w/w), power density-29.94 (W.cm-2), and ultrasonication time-9.7 min. Multi objective hybrid GA ascertained robust than conventional methods in this study.

Introduction

Global energy demand is expected to increase over the coming decades. A primary cause accredited to this upsurge is a result of increase in economic growth measured by global GDP. According to the IEO2017 (International energy outlook) report energy consumption is expected to increase from 575 quadrillion Btu in 2015 to 736 quadrillion Btu in 2040, a 28% increase. Regardless of the increase in use of renewable sources, only a 2–3% increase is expected in year 2040 compared to year 2015. Whereas, 77% of energy demand still be depending on fossil fuels. Liquid fuel consumption however, is expected to decrease from 33% to 31% in 2040. To match the energy demand, overuse of liquid fuels may lead to energy crisis. New technology is therefore desired capable of reducing the liquid fuel consumption and pollution, simultaneously. Diesel fuels are used in compression ignition engines of heavy-duty vehicles or machines with the advantage of high compression ratio that produces high turbo boost pressure even at low loads with the limitations of COx, NOx and particulate matters (PMs) emissions. To cope up the current scenario, different options like blending of oxygenated fuels with diesel and biodiesel etc. are available with their own pros and cons. One of the recent research areas in this reference is water emulsified diesel fuels that not only work without modifications in the engine but also reduces the pollutant labels (NOx, PMs) drastically by bringing out micro-explosion followed by secondary atomization. Moreover, combustion properties specifically brake specific fuel consumption also improves.

Water-in-oil nanoemulsions are kinetically stable multiphase colloidal dispersions in the size range of 20–200 nm [[1], [2], [3], [4], [5]] and one of the conditions for emulsified fuels. Smaller droplet size of nanoemulsions make them inimitable in the context of stable size, large surface area per unit volume and optical transparency [[6], [7], [8]]. However, their small size negates separation driven by gravitation force and avoids the phenomenon of creaming/sedimentation and flocculation [[8], [9], [10]]. Emulsification phenomenon being inherently non-spontaneous (ΔG>0) has tendency to separate the immiscible phases. An external force therefore is required to stabilize the system either in the form of change in chemical potential through phase inversion (low energy method, LEM) or by application of high shear/pressure through homogenizer, ultrasonic cavitation etc. (high energy method, HEM) [[11], [12], [13]]. In case of LEM droplet deformation takes place in viscous shear flow under action of rotating flow. Droplet deforms to prolate ellipsoidal shape and breaks when Weber number (Wb=ηcVr/γ) exceeds critical Weber number (Wb,crit=f[flow type,ηd/ηc]) [6,14]. Whereas, critical Weber number is a function of flow type and ratio of viscosities of dispersed phase to continuous phase (ηdc) [15].

Nanoemulsion formation through HEM, however, exploits intensive shear force to generate nano-drops. An example of such type is ultrasonic cavitation wherein formation and growth of microbubble/cavity takes place when ultrasonic waves are introduced to the liquid sample. Microbubble grows until a critical size is obtained, thereafter, implosion results in high shear rate or microjet causing breakup and dispersion of droplets. Moreover, implosion of cavitation bubbles also generates high temperature (≈5000 K) and pressure (≈1000 bar) locally [16,17]. Formation of nano-drops ≤100 nm by ultrasonic cavitation has been reported by many researchers [18,19]. Whereas in case of water-in-diesel oil, lower droplet size not only improves the combustion and emission standards but also imparts enhanced stability.

Optimization of process parameters for the formation of emulsions plays a vital role in efficient selection of variables. In recent years, response surface methodology (RSM) for efficient optimization of independent variables has been extensively tested and reported in literature specifically in emulsion formations [[20], [21], [22], [23]]. However, in optimization of nanoemulsions formation, experimental designs have been extensively used to find out significant variables by implementing low and high energy methods [3,20,21,24].

Artificial neural network (ANN) has been effectively used as a tool for determining the factors controlling the particle size [25], predicting phase behavior [26], stability [27], rheological behavior of emulsions [28] and surface tension of pure organic compounds [29]. Recently, central composite design (CCD) combined with the ANN has been reported in the literature pertaining to food, landfill leachate treatment and other different fields [30,31] like loading of nanoparticles on nanotubes for dispersion and extraction of dye [32,33]. Whereas, ANN with genetic algorithm (GA) has also been used for identifying the phase behavior of colloids and biosorptive remediation [34,35].

The most recurrent aim for optimization is to exploit the benefits of nanoemulsions compared with the conventional macro-emulsions with respect to their smaller size and low polydispersity index (PI). Parameters of heuristic and metaheuristic algorithms have been tuned by using different approaches by several researchers [[36], [37], [38]]. In a metaheuristic approach, researchers have evaluated guidelines to select parameters for particle swamp optimization and validated them by using experimental design [39]. Whereas in other studies parameters of Tabu search algorithm were tuned by factorial design [40]. Several studies have used experimental design (RSM) for getting the optimized combination of parameters that would help in fine tuning the ANN [41,42]. Furthermore, many studies involved the use of different optimization techniques like Taguchi method, Evolutionary Algorithms, etc., for improvements in performance of ANN [[43], [44], [45], [46], [47], [48]]. The lack of knowledge and information of these tools in analyzing the results poses the difficulty in working in this area [49].

Present work is focused on the process intensification of W/O NE formulation by optimizing the process parameters concerned with CEMSNE (combined energy mixed surfactant nanoemulsion) method wherein, ultrasonication in combination with isothermal dilution method was implemented by adopting the heuristic approach. However, proposed algorithm included first the parametric optimization through RSM, thereafter ANN is applied on the RSM generated design points. Further, genetic algorithm finally (GA) tuned the results by making use of hybrid GA technique wherein, RSM generated fitness function and ANN generated data points were used to initialize the algorithm. Therefore this work applies both models viz. CCD-RSM and ANN-GA for efficiently optimizing the formation of W/O NE that comprises of minimum droplet with lower kinematic viscosity simultaneously [50].

Section snippets

Materials

Non-ionic surfactants used in the study viz. 4-Octylphenol polyethoxylate (HLB 13.5) and Sorbitanmonooleate (HLB 4.3) were acquired from two different vendors, Molychem, Mumbai, India and Sigma Aldrich, USA, respectively. Diesel oil was procured from the Hindustan Petroleum Corporation Limited, Roorkee, Uttarakhand, India retail outlet pump. Surfactants used in the present work were not purified further. Double-distilled water was used in the preparation of W/O NEs.

Pre-emulsification unit

A pre-emulsification unit

Experimental design

RSM is a mathematical and statistical analytical tool based on developing model that inspects effect of multiple independent variables and their interaction with response variables. RSM was first developed in 1950s by Box and collaborators [52]. Box-Behnken (BBD) and central composite (CCD) designs are mostly used response surface designs applicable for optimizing the emulsification process. Both designs can accommodate second order terms, however, BBD is less expensive in experimentation than

Results and discussions

Formation of W/O NEs was attained by finding optimal values of response variables like droplet size (YDia) of dispersed phase and kinematic viscosity (YKV). Combination of low and high energy methods was used in the NE formation by application of set of experiments using statistical technique like RCCD-RSM. Set of experiments in the normalized form planned by the RSM were utilized in ANN model for simulating the experimental results. Data obtained from the ANN model and RCCD-RSM generated

Conclusions

A novel intensified approach (CEMNSE method) was adopted for the formulation of water-in-diesel nanoemulsion by optimizing and predicting the process parameters responsible for minimizing response variables viz. dispersed phase droplet size (YDia) and nanoemulsion kinematic viscosity (YKV). Ultrasonic cavitation and isothermal dilution method in combination was implemented in the nanoemulsion formation as a part of CEMNSE method. A collaborated modeling approach introduced FFBPANN-GA coupled

Acknowledgement

One of the author, H.K., acknowledges the financial assistance in the form of research assistantship provided by the Ministry of Human Resource and Development (MHRD), Government of India.

References (62)

  • P. Walstra

    Principles of emulsion formation

    Chem. Eng. Sci. Science

    (1993)
  • M. Sivakumar et al.

    Cavitation technology - A greener processing technique for the generation of pharmaceutical nanoemulsions

    Ultrason. Sonochem.

    (2014)
  • T.S.H. Leong et al.

    Minimising oil droplet size using ultrasonic emulsification

    Ultrason. Sonochem.

    (2009)
  • J. Carpenter et al.

    Ultrasonic assisted formation and stability of mustard oil in water nanoemulsion: effect of process parameters and their optimization

    Ultrason. Sonochem.

    (2017)
  • I. Alzorqi et al.

    Optimization of ultrasound induced emulsification on the formulation of palm-olein based nanoemulsions for the incorporation of antioxidant β-D-glucan polysaccharides

    Ultrason. Sonochem.

    (2016)
  • T. Mehmood et al.

    Optimization of mixed surfactants-based β-carotene nanoemulsions using response surface methodology: an ultrasonic homogenization approach

    Food Chem.

    (2018)
  • M. Katsouli et al.

    Optimization of water in olive oil nano-emulsions composition with bioactive compounds by response surface methodology

    Lwt - Food Sci. Technol.

    (2018)
  • S. Saravanan et al.

    Optimization of DI diesel engine parameters fueled with iso-butanol/diesel blends – response surface methodology approach

    Fuel

    (2017)
  • C.M. Pey et al.

    Optimization of nano-emulsions prepared by low-energy emulsification methods at constant temperature using a factorial design study

    Colloids Surf. A Physicochem. Eng. Asp.

    (2006)
  • A. Amani et al.

    Determination of factors controlling the particle size in nanoemulsions using Artificial Neural Networks

    Eur. J. Pharm. Sci.

    (2008)
  • R.G. Alany et al.

    Use of artificial neural networks to predict quaternery phase systems from limited experimental data

    J. Pharm. Biomed. Anal.

    (1999)
  • T.O. Fasuan et al.

    Application of osmotic pressure in modification of Amaranthus viridis starch

    Lwt.

    (2018)
  • M.R. Sabour et al.

    Comparative study of ANN and RSM for simultaneous optimization of multiple targets in Fenton treatment of landfill leachate

    Waste Manag.

    (2017)
  • M. Moghaddari et al.

    A simple approach for the sonochemical loading of Au, Ag and Pd nanoparticle on functionalized MWCNT and subsequent dispersion studies for removal of organic dyes: artificial neural network and response surface methodology studies

    Ultrason. Sonochem.

    (2018)
  • N. Vedaraman et al.

    Ultrasonic extraction of natural dye from Rubia cordifolia, optimisation using response surface methodology (RSM) &amp; comparison with artificial neural network (ANN) model and its dyeing properties on different substrates

    Chem. Eng. Process. Process. Intensif.

    (2017)
  • P. Kundu et al.

    Formulation development, modeling and optimization of emulsification process using evolving RSM coupled hybrid ANN-GA framework

    Chem. Eng. Res. Des.

    (2015)
  • S. Jacob et al.

    Modeling and optimization of anaerobic codigestion of potato waste and aquatic weed by response surface methodology and artificial neural network coupled genetic algorithm

    Bioresour. Technol.

    (2016)
  • E. Taillard

    Robust tabu search for the quadratic assignment problem

    Parallel Comput.

    (1991)
  • S. Dharma et al.

    Optimization of biodiesel production process for mixed Jatropha curcas-Ceiba pentandra biodiesel using response surface methodology

    Energy Convers. Manage.

    (2016)
  • M. Mäkelä

    Experimental design and response surface methodology in energy applications: a tutorial review

    Energy Convers. Manage.

    (2017)
  • J.F.C. Khaw et al.

    Optimal design of neural networks using the Taguchi method

    Neurocomputing

    (1995)
  • Cited by (15)

    • Recent advances in delivery systems optimization using machine learning approaches

      2023, Chemical Engineering and Processing - Process Intensification
    • Soft-computing modeling and multiresponse optimization for nutrient removal process from municipal wastewater using microalgae

      2022, Journal of Water Process Engineering
      Citation Excerpt :

      Several well-known techniques are utilized for global optimization of process variables, viz., the genetic algorithm (GA), harmony searches (HS), particle swarm optimization (PSO), and ant colony optimization (CO) [51–53]. However, GA is very popular and widely used for global optimization [54]. Thus, the best predictive models were hybridized with GA to optimize the nutrient removal process operational parameters.

    • Energy consumption, flow characteristics and energy-efficient design of cup-shape blade stirred tank reactors: Computational fluid dynamics and artificial neural network investigation

      2022, Energy
      Citation Excerpt :

      It should be noted that relative to the explainable machine learning (ML) technologies such as decision tree (DT) and regularized linear models, the absence of explanation of ANN prevents end-users from understanding its underlying mechanics or formalization the input-output dependence [17] due to its “black-box” nature. However, ANN owns competitive advantages on self-learning, parallel computing and storage, better robustness and generalization ability [18–20], and was reported to embody better prediction or classification capability on many occasions [21–23], especially for handling nonlinear multiple input-multiple output systems, “big data”, and classification problems with more categorical variables [23,24]. For instance, Ibrahim et al. [21] proved that ANN showed a better prediction accuracy than DT and linear regression.

    View all citing articles on Scopus
    View full text