Search for: All records

In this work, we introduce MOLA, a multi-block orthogonal long short-term memory autoencoder paradigm, to conduct accurate, reliable fault detection of industrial processes. To achieve this, MOLA effectively extracts dynamic orthogonal features by introducing an orthogonality-based loss function to constrain the latent space output. This helps eliminate the redundancy in the features identified, thereby improving the overall monitoring performance. On top of this, a multi-block monitoring structure is proposed, which categorizes the process variables into multiple blocks by leveraging expert process knowledge about their associations with the overall process. Each block is associated with its specific orthogonal long short-term memory autoencoder model, whose extracted dynamic orthogonal features are monitored by distance-based Hotelling's T^2 statistics and quantile-based cumulative sum (CUSUM) designed for multivariate data streams that are nonparametric and heterogeneous. Compared to having a single model accounting for all process variables, such a multi-block structure significantly improves overall process monitoring performance, especially for large-scale industrial processes. Finally, we propose an adaptive weight-based Bayesian fusion (W-BF) framework to aggregate all block-wise monitoring statistics into a global statistic that we monitor for faults. Fault detection speed and accuracy are improved by assigning and adjusting weights to blocks based on the sequential order in which alarms are raised. We demonstrate the efficiency and effectiveness of our MOLA framework by applying it to the Tennessee Eastman process and comparing the performance with various benchmark methods. 

Modern industrial processes are continuously monitored by a large number of sensors. Despite having access to large volumes of historical and online sensor data, industrial practitioners still face challenges in the era of Industry 4.0 in effectively utilizing them to perform online process monitoring and fast fault detection and diagnosis. To target these challenges, in this work, we present a novel framework named "FARM" for Fast, Accurate, and Robust online process Monitoring. FARM is a holistic monitoring framework that integrates (a) advanced multivariate statistical process control (SPC) for fast anomaly detection of nonparametric, heterogeneous data streams, and (b) modified support vector machine (SVM) for accurate and robust fault classification. Unlike existing general-purpose process monitoring frameworks, FARM's unique hierarchical architecture decomposes process monitoring into two fault detection and diagnosis, each of which is conducted by targeted algorithms. Here, we test and validate the performance of our FARM monitoring framework on Tennessee Eastman Process (TEP) benchmark dataset. We show that SPC achieves faster fault detection speed at a lower false alarm rate compared to state-of-the-art benchmark fault detection methods. In terms of fault classification diagnosis, we show that our modified SVM algorithm successfully classifies 17 out of 20 of the fault scenarios present in the TEP dataset. Compared with the results of standard SVM trained directly on the original dataset, our modified SVM improves the fault classification accuracy significantly.