Automated Incident Detection in Refinery Operations Using Machine Learning
DOI:
https://doi.org/10.63282/3050-9262.IJAIDSML-V7I3P107Keywords:
Automated Incident Detection, Refinery Operations, Anomaly Detection, Machine Learning, Process Safety, Sensor Data, Event Classification, Real-Time MonitoringAbstract
Refinery operations run continuously and involve complex, tightly coupled processes in which a single equipment failure or chemical imbalance can escalate rapidly into a major safety incident. Most refineries still rely on human operators to detect and classify incidents manually from alarm panels and sensor dashboards a process that is slow, cognitively demanding, and error-prone. This paper proposes an automated incident-detection system based on machine learning that monitors real-time process-sensor streams, identifies anomalies, and classifies them into specific incident types without waiting for human recognition. The system couples two unsupervised anomaly-detection techniques (an autoencoder neural network and an isolation forest) with two supervised classifiers (a random forest and a long short-term memory network). The paper states the problem, describes the proposed architecture, explains each machine-learning technique and the rationale for its selection, and presents a step-by-step methodology for building and deploying the system in a production refinery environment.
References
[1] V. Venkatasubramanian, R. Rengaswamy, K. Yin, and S. N. Kavuri, “A review of process fault detection and diagnosis: Part I: Quantitative model-based methods,” Comput. Chem. Eng., vol. 27, no. 3, pp. 293–311, 2003.
[2] V. Chandola, A. Banerjee, and V. Kumar, “Anomaly detection: A survey,” ACM Comput. Surv., vol. 41, no. 3, pp. 1–58, Jul. 2009.
[3] I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning. Cambridge, MA, USA: MIT Press, 2016.
[4] M. Sakurada and T. Yairi, “Anomaly detection using autoencoders with nonlinear dimensionality reduction,” in Proc. MLSDA 2nd Workshop Mach. Learn. Sensory Data Anal., Gold Coast, Australia, 2014, pp. 4–11.
[5] F. T. Liu, K. M. Ting, and Z.-H. Zhou, “Isolation forest,” in Proc. 8th IEEE Int. Conf. Data Mining (ICDM), Pisa, Italy, 2008, pp. 413–422.
[6] L. Breiman, “Random forests,” Mach. Learn., vol. 45, no. 1, pp. 5–32, 2001.
[7] S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Comput., vol. 9, no. 8, pp. 1735–1780, 1997.
[8] P. Malhotra, L. Vig, G. Shroff, and P. Agarwal, “Long short term memory networks for anomaly detection in time series,” in Proc. 23rd Eur. Symp. Artif. Neural Netw., Comput. Intell. Mach. Learn. (ESANN), Bruges, Belgium, 2015, pp. 89–94.
[9] N. V. Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, “SMOTE: Synthetic minority over-sampling technique,” J. Artif. Intell. Res., vol. 16, pp. 321–357, 2002.
[10] D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” in Proc. 3rd Int. Conf. Learn. Represent. (ICLR), San Diego, CA, USA, 2015.
[11] N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhutdinov, “Dropout: A simple way to prevent neural networks from overfitting,” J. Mach. Learn. Res., vol. 15, no. 1, pp. 1929–1958, 2014.
[12] Z. Xiao, A. Kordon, and S. Sen, “Fault detection and diagnosis in Tennessee Eastman process with deep autoencoder,” in Proc. Annu. Conf. PHM Soc., vol. 15, no. 1, 2023.
[13] OPC Foundation, OPC Unified Architecture, IEC 62541, Int. Electrotech. Comm., Geneva, Switzerland, 2020.
