The naphtha cracking process heavily relies on the composition of naphtha, which is a complex blend of different hydrocarbons. Predicting the naphtha composition accurately is crucial for efficiently controlling the cracking process and achieving maximum performance. Traditional methods, such as gas chromatography and true boiling curve, are not feasible due to the need for pilot-plant-scale experiments or cost constraints. In this paper, we propose a neural network framework that utilizes chemical property information to improve the performance of naphtha composition prediction. Our proposed framework comprises two parts: a Watson K factor estimation network and a naphtha composition prediction network. Both networks share a feature extraction network based on Convolutional Neural Network (CNN) architecture, while the output layers use Multi-Layer Perceptron (MLP) based networks to generate two different outputs - Watson K factor and naphtha composition. The naphtha composition is expressed in percentages, and its sum should be 100%. To enhance the naphtha composition prediction, we utilize a distillation simulator to obtain the distillation curve from the naphtha composition, which is dependent on its chemical properties. By designing a loss function between the estimated and simulated Watson K factors, we improve the performance of both Watson K estimation and naphtha composition prediction. The experimental results show that our proposed framework can predict the naphtha composition accurately while reflecting real naphtha chemical properties.

In multiobjective particle swarm optimization (MOPSO), the global‐best particle is randomly selected for each population particle from a nondominated solution set. However, this Roulette wheel‐based global particle selection is ineffective for convergence and diversity when the problem has numerous decision variables or a large number of global‐best candidates. Thus, this study proposes the cluster‐based MOPSO (CMOPSO). In CMOPSO, the similarities between particles are considered when selecting the global‐best particle. The cluster for each particle is determined based on the Euclidean distance in the decision or objective space. The proposed approach is demonstrated by applying an operating condition optimization problem to the hydrogen production process. The target process is a representative chemical plant with a large search space and strong nonlinearity. Furthermore, the performance of CMOPSO is assessed by comparing it with that of MOPSO. The results indicate that CMOPSO considered in the decision space exhibits superior performance in terms of convergence and diversity.

Cover Caption: The cover image is based on the Research Article Development of physical property prediction models for polypropylene composites with optimizing random forest hyperparameters by Chonghyo Joo et al., https://doi.org/10.1002/int.22700.

The physical properties required in polypropylene composites (PPCs) vary depending on the purpose of use. In the manufacturing of PPCs, it is crucial to determine the types and quantities of numerous reinforcements to meet the required physical properties. Owing to industrial complexity, most PPC manufacturers produce the composites repeatedly until the desired physical properties are obtained. Hence, to reduce trial and error, we developed prediction models for the physical properties of PPCs based on commercial recipe data. The recipe data included information about five physical properties of composites manufactured using 90 materials. In complex industrial environments, because one recipe is usually composed of 2–12 materials, numerous combinations of data sets are created. It causes the lack of the same material combination data sets and thus makes it difficult to develop a good performance model. Therefore, a novel categorization process is suggested as data preprocessing to overcome the data imbalance problem. The models for predicting the five physical properties (flexural strength, melting index, tensile strength, specific gravity, and flexural modulus) were developed using random forest, and the performance of the prediction models was improved via hyperparameter optimization. Furthermore, the effects of the materials on the performance of the models were numerically described through variable importance analysis. Finally, a software was developed to implement the prediction models in the industry. The software was applied to a commercial composite and achieved high accuracy, demonstrating the effectiveness of this study. Thus, the software suggests decision-making solutions to save cost and time by reducing the trial and error in the industrial environment with high complexity.