Prediction of the surface tension of binary systems based on the partial least squares method

Abstract

To predict the surface tension of binary liquid systems, an empirical model is proposed using the partial least squares (PLS) based on the multivariate statistical analysis method. Required parameters for the PLS method to predict the surface tension of binary systems are composed of the thermophysical properties of only pure substances such as critical temperature, critical pressure, critical volume, molar volume, viscosity and vapor pressure for input data block (X) and the reported experimental surface tension data for output data block (Y). The data set for the experimental surface tension of binary liquid systems is divided into the training set for regression and the test set for predicting. An average relative error (%) results of regression and prediction indicate that the PLS method can be a useful tool for predicting the surface tension of liquid binary systems.

Adaptive prediction model for fluidized catalytic cracking processes based on the PLS method

“…The PLS method uses LVs that are produced among the input variables ( X ) to be orthogonal with each other and highly correlated with the output variables ( Y ), which are linearly independent of each other . This method is able to produce a prediction model by a regression between the input and output LVs, which are highly correlated each other.…”

“…This method is able to produce a prediction model by a regression between the input and output LVs, which are highly correlated each other. The PLS method is made up of outer relations and an inner relation . Outer relations are produced from the input ( X ) and output ( Y ) variable data blocks.…”

“…The outer relations of the X and Y data blocks are as follows where t h and p h are input score and loading vectors, and E is the residual of X data block.…”

“…Geladi and Kowalski introduced a simple PLS method theoretically . This method has been applied to predict properties of chemical substances, for example, the liquid viscosities, the surface tension, and the flash point . The simple PLS method is accomplished by finding linear relationships between the input and output variables utilized by reducing set of latent variables (LVs) .…”

“…This method has been applied to predict properties of chemical substances, for example, the liquid viscosities, the surface tension, and the flash point . The simple PLS method is accomplished by finding linear relationships between the input and output variables utilized by reducing set of latent variables (LVs) . Most chemical processes require an adaptive model because of time‐varying changes and abnormal behaviors of the processes rather than a fixed model using a simple PLS method.…”