A Data-Driven Framework for Monitoring the Operation of Gas-to-Liquid (GTL) Plants

Gas-to-Liquid (GTL) processes widely are used to produce chemicals and fuels from natural gas [1]. Proper operation of GTL plants requires tight monitoring of their operating conditions. Large amounts of GTL data are continuously collected, which contain important information about the wellness of these plants. In this work, a data-driven framework is developed to monitor the operation of GTL plants. The developed framework consists of three main tasks: modeling, fault detection, and fault identification. Various data-driven techniques are developed to enhance the accuracy of each of these tasks. For example, linear and nonlinear unsupervised modeling techniques that enhance the accuracy of existing methods are developed, which include multiscale Bayesian principal component analysis (MSBPCA) [2], Bayesian Optimized Kernel principal component analysis (BOKPCA) [3, 4], and Bayesian Optimized neural network (BONN) techniques [5]. These techniques provide enhanced prediction accuracy, especially for nonlinear and noisy data. Furthermore, new fault detection techniques are developed, which include PCA-based generalized likelihood ratio (PCA-GLR) test, multiscale PCA-GLR, and interval PCA-GLR [6]. These techniques allow detecting various types of faults, such as shifts in the mean or changes in variance. Finally, extensions of support vector machines (SVM) [7] and random forest (RF) [8] are also explored for enhanced fault identification. The developed modeling, fault detection, and fault identification techniques will be integrated in the proposed framework for monitoring the operation of GTL plants. The effectiveness of such a framework will be validated using simulated examples as well as real data collected from a Fischer-Tropsch setup at the Gas and Fuels Research Center (GFRC) at Texas A&M University at Qatar.

Keywords: process monitoring; data-driven modeling; fault detection; fault identification; gas-to-liquid (GTL) plants

References:

[1] Bao, B., El-Halwagi, M., Elbashir, N., “Simulation, Integration, and Economic Analysis of Gas-to-Liquid Processes”, Fuel Processing Technology 91, 703–713, 2010.

[2] Nounou, M., Bakshi, B., Goel, P. and Shen, X., “Bayesian Principal Component Analysis,” Journal of Chemometrics, 16, 11, 576–595, 2002.

[3] Kutateladze, V., “The Kernel Trick for Nonlinear Factor Modeling”, International Journal of Forecasting, 38, 1, 165–177, 2022.

[4] Victoria, A. and Maragatham, G., “Automatic Tuning of hyperparameters using Bayesian optimization. Evolving Systems, 12(1):217–223, 2021.

[5] K.-L. Du, M. N. S. Swamy, Neural Networks and Statistical Learning, 2nd Ed., Springer-Verlag London Ltd., 2019.

[6] Harrou, F., Nounou, M., Nounou, H., and Madakyaru, M., “Statistical Fault Detection using PCA-based GLR Hypothesis Testing”, Journal of Loss Prevention in the Process Industries, 26, 1, 129-139, 2013.

[7] Cortes, C., Vapnik, V., “Support Vector Networks”, Machine Learning, 20, 273-297, 1995.

[8] Mansouri, M., Fezai, R., Trabelsi, M., Nounou, H., Nounou, M., and Bouzrara, K., “Reduced Gaussian Process Regression Based Random Forest Approach for Fault Diagnosis of Wind Energy Conversion Systems”, IET Renewable Power Generation, 15, 5, 3612-3621, 2021.