Naslov
Application of Artificial Neural Network (ANN) modeling for efficient assessment of average Feret diameter of oil-in-water emulsions prepared using continuously operated microfluidic device

Autori
Grgić, Filip ; Benković, Maja ; Valinger, Davor ; Jurina, Tamara ; Gajdoš Kljusurić, Jasenka ; Jurinjak Tušek, Ana

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, prošireni sažetak, znanstveni

Izvornik
Book of extended abstracts 6th International Conference The Implementation of Microreactor Technology in Biotechnology (IMTB) / Šalić, Anita ; Seručnik, Mojca ; Jurinjak Tušek, Ana ; Zelić, Bruno ; Žnidaršič Plazl, Polona - Ljubljana : Fakulteta za kemijo in kemijsko tehnologijo Univerze v Ljubljani, 2022, 109-110

ISBN
978-961-7078-24-4

Skup
6th International Conference on Implementation of Microreactor Technology in Biotechnology (IMTB)

Mjesto i datum
Portorož, Slovenija, 05.06.2022. - 08.06.2022

Vrsta sudjelovanja
Poster

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
oil-in-water emulsions ; NIR spectroscopy ; microfluidic emulsification ; ANN modeling ; average Feret diameter of the droplets

Sažetak
Emulsions are used as carriers of hydrophilic and lipophilic substances in order to improve the bioavailability and stability of biologically active compounds. Therefore, emulsions have found a wide range of applications.1 Conventionally, emulsions are prepared based on shear or impact loads generated by manual or mechanical mixing.2 Due to non-uniform loads throughout the system, emulsions prepared in the described manner consist of droplets of different sizes. An alternative to the classical methods of producing emulsions are microfluidic systems.3, 4 By reducing the dimensions of the devices by several orders of magnitude, significant energy savings and a reduction in the negative impact on the environment can be achieved. Compared to traditional emulsion production methods, microfluidic systems ensure that only one droplet is produced at a time, and thus the process results in a highly monodisperse emulsion. The droplet size distribution of emulsions is affected by a variety of factors, including mixing speed and time, mixer type, oil composition, oil-to-water ratio, and surfactant type and concentration.5 Therefore, a fast and non-destructive analytical tool is critical to monitor the process, and near-infrared (NIR) spectroscopy has proven beneficial in emulsion analysis.6 Analysis of NIR spectra requires the acquisition of a large amount of data. To reduce the complexity of the acquired data and identify significant patterns in NIR spectra, various mathematical and statistical methods (such as principal component analysis (PCA), partial least square regression (PLSR), canonical correlation analysis (CCA), and principal Component Regression (PCR)) have been used.7 However, the nonlinearity of the data is a major problem in multivariate approaches. Modeling with artificial neural networks (ANNs) can be effectively used to overcome difficult nonlinear interactions. By training for numerous input-output systems, ANN provides a model of biological network structures that can identify and replicate cause-effect relationships. In this work, oil-in-water emulsions were prepared in microfluidic systems using polyethylene glycol (PEG1500, 6000 and 20000) as emulsifiers. The microfluidic systems consisted of glass microchips with laser-engraved microchannels (width: height: length = 250 μm: 150 μm: 55.3 mm) equipped with static teardrop micromixers (Micronit Microfluidics B.V., Enschede, The Netherlands). Analysis of the flow profile of oil-water two-phase systems was performed. The effect of emulsifier concentration, oil concentration and total flow rate on the average Feret diameter of the prepared emulsions was analyzed. The NIR spectra of the prepared emulsions were recorded using a NIR spectrometer (NIR-128-1.7-USB/6.25/50 µm Control Development Inc., South Bend, USA) in the wavelength range from 904 nm to 1699 nm. Principal component analysis (PCA) was used as a qualitative tool to identify similarities and differences between samples based on the raw NIR spectra in Statistica v.13.0 software (Tibco Software, Palo Alto, USA). Artificial neural network (ANN) modeling was used to predict the average Feret diameter of oil-in-water emulsions based on the recorded row NIR spectra. The input layer consisted of 5 neurons representing the coordinates of the first five factors from PCA. The applicability of the ANN models was evaluated by the coefficient of determination for training, testing, and validation, and by the mean square error also for training, testing, and validation. When choosing the optimal architecture of an artificial neural network, the number of neurons in the hidden layer is also considered as a criterion. The obtained results show that the average Feret diameter of emulsions with water, as a continuous phase, increases with increasing molecular weight of the emulsifier used. Moreover, the developed ANN models have high values of the determination coefficients for training, test and validation with the associated small errors and can be reliably used for predicting the size of the droplet diameter of the dispersed phase in the process of producing oil-in-water emulsions.

Izvorni jezik
Engleski

Znanstvena područja
Biotehnologija

POVEZANOST RADA