Simultaneous UV-Vis spectrophotometric quantification of ternary basic dye mixtures by partial least squares and artificial neural networks

2016 ◽  
Vol 74 (10) ◽  
pp. 2497-2504 ◽  
Author(s):  
Seyed Karim Hassaninejad-Darzi ◽  
Mohammad Torkamanzadeh

One of the main difficulties in quantification of dyes in industrial wastewaters is the fact that dyes are usually in complex mixtures rather than being pure. Here we report the development of two rapid and powerful methods, partial least squares (PLS-1) and artificial neural network (ANN), for spectral resolution of a highly overlapping ternary dye system in the presence of interferences. To this end, Crystal Violet (CV), Malachite Green (MG) and Methylene Blue (MB) were selected as three model dyes whose UV-Vis absorption spectra highly overlap each other. After calibration, both prediction models were validated through testing with an independent spectra-concentration dataset, in which high correlation coefficients (R2) of 0.998, 0.999 and 0.999 were obtained by PLS-1 and 0.997, 0.999 and 0.999 were obtained by ANN for CV, MG and MB, respectively. Having shown a relative error of prediction of less than 3% for all the dyes tested, both PLS-1 and ANN models were found to be highly accurate in simultaneous determination of dyes in pure aqueous samples. Using net-analyte signal concept, the quantitative determination of dyes spiked in seawater samples was carried out successfully by PLS-1 with satisfactory recoveries (90–101%).

Author(s):  
Jatinder Kumar ◽  
Ajay Bansal

The experimental determination of various properties of diesel-biodiesel mixtures is very time consuming as well as tedious process. Any tool helpful in estimation of these properties without experimentation can be of immense utility. In present work, other tools of determination of properties of diesel-biodiesel blends were tried. A traditional statistical technique of linear regression (principle of least squares) was used to estimate the flash point, fire point, density and viscosity of diesel and biodiesel mixtures. A set of seven neural network architectures, three training algorithms along with ten different sets of weight and biases were examined to choose best Artificial Neural Network (ANN) to predict the above-mentioned properties of dieselbiodiesel mixtures. The performance of both of the traditional linear regression and ANN techniques were then compared to check their validity to predict the properties of various mixtures of diesel and biodiesel. Key words: Biodiesel; Artificial Neural Network; Principle of least squares; Diesel; Linear Regression. DOI: 10.3126/kuset.v6i2.4017Kathmandu University Journal of Science, Engineering and Technology Vol.6. No II, November, 2010, pp.98-103


2003 ◽  
Vol 11 (3) ◽  
pp. 211-218 ◽  
Author(s):  
Juliana Paschoal ◽  
Fernando D. Barboza ◽  
Ronei J. Poppi

The feasibility of using near infrared (NIR) transmission spectroscopy for rapid and conclusive determination of contaminants in lubricant oil was investigated. The NIR spectrum in the region from 1300 to 1700 nm was used to predict gasoline and ethylene glycol concentrations present in lubricant oil. A graphically-oriented local multivariate calibration modelling procedure called interval partial least-squares (iPLS) was applied to find variable intervals that featured the lowest prediction error. When compared with the full spectrum PLS model, better results were obtained through the iPLS program. High correlation coefficients and low root mean square errors of cross-validation ( RMSECV) were obtained for gasoline ( R = 0.98, RMSECV = 0.38%, range = 0.2–8.0% w/w) and ethylene glycol determinations ( R = 0.97, RMSECV = 0.04%, range = 0.06 to 0.7% w/w), indicating that the proposed methodology can be used for contaminant determinations in lubricant oil.


In this study, three Artificial Neural Network (ANN) models (Feedforward network, Elman, and Nonlinear Autoregressive Exogenous (NARX)) were used to predict hourly solar radiation in Amman, Jordan. The three models were constructed and tested by using MATLAB software. Meteorological data for the years from 2000 to 2010 were used to train the ANN while the yearly data of 2011 was used to test it. It was found that ANN technique may be used to estimate the hourly solar radiation with an excellent accuracy, and the coefficient of determination of Elman, feedforward and NARX models were found to be 0.97353, 0.97376, and 0.99017, respectively. The obtained results showed that NARX model has the best ability to predict the required solar data, while Elman and feedforward models have the lowest ability to predict it.


Sign in / Sign up

Export Citation Format

Share Document