(646b) Big Data Approach to Fault Detection, Diagnosis and Maintenance Optimization in Batch Processes

Onel, M., Texas A&M Energy Institute, Texas A&M University
Kieslich, C. A., Texas A&M University
Guzman, Y. A., Princeton University
Pistikopoulos, E. N., Texas A&M Energy Institute, Texas A&M University
Batch process monitoring is an important task in order to attain a safe operability and minimize loss of productivity and profit, yet a challenging one. Characterized with a considerable number of interconnected variables, inherent non-stationarity, nonlinear response, batch-to-batch variability as well as finite duration, batch processes have become the focus of numerous researchers in the past few decades [1]. Today, statistical process monitoring (SPM) and data-driven process monitoring algorithms have been widely used in industry due to the advances in sensing technology which has facilitated enormous amounts of process data collection [2,3]. However, most novel techniques for fault detection and identification have focused on continuous processes, and the need of monitoring algorithm development for batch processes is evident.

In this work, we present a new data-driven framework for process monitoring and intervention in batch processes. Central to the framework are novel theoretical and algorithmic developments in nonlinear support vector machine-based feature selection [4, 5] which encapsulates highly nonlinear relationships between features, thus improving fault detection model accuracy and guiding fault diagnosis. We present results of a recent extensive benchmark, simulation dataset on Pensim model [6, 7] consisting 90,400 batches with numerous and diverse fault types, where we train time-specific models on the pre-aligned batch data trajectories in a moving horizon basis. The developed analysis framework aims for early detection of faulty batches and enables intervention to reduce loss of profit. Furthermore, simultaneous achievement of high process efficiency, safety and profitability is of utmost importance in modern process industries [8]. Therefore by using the insights attained on fault detection latency and preventative action from the built models, we will extend the discussion to the prediction of batch process duration as well as maintenance optimization of batch process operations [9, 10].


[1] Van Sprang, E. N., Ramaker, H. J., Westerhuis, J. A., Gurden, S. P., & Smilde, A. K. (2002). Critical evaluation of approaches for on-line batch process monitoring. Chemical Engineering Science, 57(18), 3979-3991.

[2] Joe Qin, S. (2003). Statistical process monitoring: basics and beyond.Journal of chemometrics, 17(8‐9), 480-502.

[3] Chiang, L. H., Braatz, R. D., & Russell, E. L. (2001). Fault Detection and Diagnosis in Industrial Systems. Springer Science & Business Media.

[4] Guzman, Y.A., Kieslich, C.A., Floudas, C.A. (Submitted). A global optimization framework for feature selection with Support Vector Machines.

[5] Kieslich, C. A., Tamamis, P., Guzman, Y. A., Onel, M., Floudas, C. A. (2016). Highly Accurate Structure-Based Prediction of HIV-1 Coreceptor Usage Suggests Intermolecular Interactions Driving Tropism. PloS one,11(2), e0148974.

[6] Van Impe, J., Gins, G. (2015). An extensive reference dataset for fault detection and identification in batch processes. Chemometrics and Intelligent Laboratory Systems, 148, 20-31.

[7] Birol, G., Ündey, C., Cinar, A. (2002). A modular simulation package for fed-batch fermentation: penicillin production. Computers & Chemical Engineering, 26(11), 1553-1565.

[8] Cinar A., Palazoglu A., Kayihan F., (2006). Chemical Process Performance Evaluation. CRC Press.

[9] Pistikopoulos, E. N.; Vassiliadis, C. G.; Papageorgiou, L. G. (2000). Process design for maintainability: An optimization approach. Computers and Chemical Engineering. 24 (2-7), 203-208.

[10] Thomaidis, T. V.; Pistikopoulos, E. N. (2004). Criticality analysis of process systems. Annual Symposium on Reliability and Maintainability. pp 451-458.