Article

Prediction of the glass transition temperature of (meth)acrylic polymers containing phenyl groups by recursive neural network

Department of Chemistry and Industrial Chemistry, University of Pisa, via Risorgimento 35, 56126 Pisa, Italy
Polymer (Impact Factor: 3.77). 11/2007; 48(24):7121-7129. DOI: 10.1016/j.polymer.2007.09.043

ABSTRACT A recursive neural network QSPR model that can take directly molecular structures as input was applied to the prediction of the glass transition temperature of 277 poly(meth)acrylates. This model satisfactorily predicted the chemical–physical properties of high and low molecular weight acyclic compounds. However, side-chain benzene rings are present in about one half of the selected polymers. In order to render cyclic structures, the molecular representation through hierarchical structures was extended by two methods, named group and cycle breaking, respectively. The latter approach exploits standard unique molecular description systems, i.e. Unique SMILES and InChI. In all cases the prediction was very good, with 15–16 K mean absolute error and 19–21 K standard deviation. This result confirms the robustness of our method with respect to the inclusion of different structures. Moreover, the good performance of the cycle breaking representation paves the way for the investigation of data sets that contain a variety of poorly sampled cyclic structures.

0 Followers
 · 
208 Views
  • [Show abstract] [Hide abstract]
    ABSTRACT: Quantitative structure-property relationships (QSPR) studies were performed for kinetic chain-transfer constants of 90 agents on styrene polymerization at 60°C. By using stepwise multilinear regression analysis (MLRA) and artificial neural network (ANN), linear and nonlinear models containing seven descriptors were obtained with R2 of 0.7866 and 0.8661 for the training set, respectively. The reliability of the proposed models was validated through the test set. The descriptors involved in the models are related to the molecular conformational changes and some functional groups. As these descriptors are directly calculated from the molecular structure, the proposed models can be employed to estimate the chain-transfer constants for styrene. © 2011 Wiley Periodicals, Inc. J Appl Polym Sci, 2011
    Journal of Applied Polymer Science 01/2012; 123(1). DOI:10.1002/app.34255 · 1.64 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper we present results on prediction of elongation at break (target property) for a group of 77 amorphous polymers of high molecular weight. Novel descriptors are proposed in order to better represent structural features related to the target property. These proposed descriptors along with the classic ones, were calculated for the set of polymers. The final descriptors of the predictive model were obtained by using a combination of variable selection method and domain knowledge. The model consisted of three descriptors: Cross-head Speed (CHS), Number Average Molecular Weight/Main Chain Surface Area ratio (Mn/SAMC), and Normalized Main Chain Mass (nMMC). By means of a multi-layer perceptron (MLP) neural network a good prediction model (R2 = 0.88 and MAE = 1.89) was achieved, which was internally and externally validated. The model shows the advantages of using well-known parameters in the field of polymers and of capturing the structural characteristics of the main and side chains. Thus, more intelligent tools are developed for the design of new materials with a specific application profile.
    Chemometrics and Intelligent Laboratory Systems 12/2014; 139:121. DOI:10.1016/j.chemolab.2014.09.009 · 2.38 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The use of quantitative structure–property relationships is proposed for the calculation of temperature of 10 % mass loss (T 10) for a number of 50 optically active polymers. The descriptors involved in these models were calculated from the structures of the repeating units. The important descriptors were selected applying genetic algorithm–partial least squares (GA–PLS) technique. A PLS method was used to select the best descriptors and the selected descriptors were used as inputs for support vector machine (SVM) model. The root mean square errors for the SVM calculated T 10 of training and prediction sets are 9.842 and 10.384, respectively, which are smaller than those obtained by PLS model (27.970 and 34.416, respectively). The results obtained showed the ability of the developed SVM to predict T 10 of various chiral polymers. Also results revealed the superiority of the SVM over the PLS model.
    Journal of Thermal Analysis and Calorimetry 05/2014; 116(2):989-1000. DOI:10.1007/s10973-013-3587-0 · 2.21 Impact Factor

Full-text

Download
3 Downloads
Available from
Apr 7, 2015