Search

The study of the styrene–Ground Tire Rubber (GTR) graft radical polymerization is particularly challenging due to the complexity of the underlying kinetic mechanisms and nature of GTR. In this work, an experimental study on two scales (∼10 mL and ∼100 mL) and a machine learning (ML) modeling approach are combined to establish a quantitative relationship between operating conditions and styrene conversion. The two-scale experimental approach enables to verify the impact of upscaling on thermal and mixing effects that are particularly important in this heterogeneous system, as also evidenced in previous works. The adopted experimental setups are designed in view of multiple data production, while paying specific attention in data reliability by eliminating the uncertainty related to sampling for analyses. At the same time, all the potential sources of uncertainty, such as the mass loss along the different steps of the process and the precision of the experimental equipment, are also carefully identified and monitored. The experimental results on both scales validate previously observed effects of GTR, benzoyl peroxide initiator and temperature on styrene conversion but, at the same time, reveal the need of an efficient design of the experimental procedure in terms of mixing and of monitoring uncertainties. Subsequently, the most reliable experimental data (i.e., 69 data from the 10 mL system) are used for the screening of a series of diverse supervised-learning regression ML models and the optimization of the hyperparameters of the best-performing ones. These are gradient boosting, multilayer perceptrons and random forest with, respectively, a test R2 of 0.91 ± 0.04, 0.90 ± 0.04 and 0.89 ± 0.05. Finally, the effect of additional parameters, such as the scaling method, the number of folds and the random partitioning of data in the train/test splits, as well as the integration of the experimental uncertainties in the learning procedure, are exploited as means to improve the performance of the developed models.

This article investigates the applicability domain (AD) of machine learning (ML) models trained on high-dimensional data, for the prediction of the ideal gas enthalpy of formation and entropy of molecules via descriptors. The AD is crucial as it describes the space of chemical characteristics in which the model can make predictions with a given reliability. This work studies the AD definition of a ML model throughout its development procedure: during data preprocessing, model construction and model deployment. Three AD definition methods, commonly used for outlier detection in high-dimensional problems, are compared: isolation forest (iForest), random forest prediction confidence (RF confidence) and k-nearest neighbors in the 2D projection of descriptor space obtained via t-distributed stochastic neighbor embedding (tSNE2D/kNN). These methods compute an anomaly score that can be used instead of the distance metrics of classical low-dimension AD definition methods, the latter being generally unsuitable for high-dimensional problems. Typically, in low- (high-) dimensional problems, a molecule is considered to lie within the AD if its distance from the training domain (anomaly score) is below a given threshold. During data preprocessing, the three AD definition methods are used to identify outlier molecules and the effect of their removal is investigated. A more significant improvement of model performance is observed when outliers identified with RF confidence are removed (e.g., for a removal of 30% of outliers, the MAE (Mean Absolute Error) of the test dataset is divided by 2.5, 1.6 and 1.1 for RF confidence, iForest and tSNE2D/kNN, respectively). While these three methods identify X-outliers, the effect of other types of outliers, namely Model-outliers and y-outliers, is also investigated. In particular, the elimination of X-outliers followed by that of Model-outliers enables us to divide MAE and RMSE (Root Mean Square Error) by 2 and 3, respectively, while reducing overfitting. The elimination of y-outliers does not display a significant effect on the model performance. During model construction and deployment, the AD serves to verify the position of the test data and of different categories of molecules with respect to the training data and associate this position with their prediction accuracy. For the data that are found to be close to the training data, according to RF confidence, and display high prediction errors, tSNE 2D representations are deployed to identify the possible sources of these errors (e.g., representation of the chemical information in the training data).