Fault detection and diagnosis in process system using artificial intelligence-based cognitive technique

Anomaly Detection for Asynchronous Multivariate Time Series of Nuclear Power Plants Using a Temporal-Spatial Transformer

Industrial process monitoring is a critical application of multivariate time-series (MTS) anomaly detection, especially crucial for safety-critical systems such as nuclear power plants (NPPs). However, some current data-driven process monitoring approaches may not fully capitalize on the temporal-spatial correlations inherent in operational MTS data. Particularly, asynchronous time-lagged correlations may exist among variables in actual NPPs, which further complicates this challenge. In this work, a reconstruction-based MTS anomaly detection approach based on a temporal-spatial transformer is proposed. It employs a two-stage temporal-spatial attention mechanism combined with a multi-scale strategy to learn the dependencies within normal operational data at various scales, thereby facilitating the extraction of temporal-spatial correlations from asynchronous MTS. Experiments on simulated datasets and real NPP datasets demonstrate that the proposed model possesses stronger feature learning capabilities, as evidenced by its improved performance in signal reconstruction and anomaly detection for asynchronous MTS data. Moreover, the proposed TS-Trans model enables earlier detection of anomalous events, which holds significant importance for enhancing operational safety and reducing potential losses in NPPs.

Association Measure and Compact Prediction for Chemical Process Data from an Information-Theoretic Perspective

Mutual information (MI) has been widely used for association mining in complex chemical processes, but how to precisely estimate MI between variables of different numerical types, discriminate their association relationships with targets and finally achieve compact and interpretable prediction has not been discussed in detail, which may limit MI in more complicated industrial applications. Therefore, this paper first reviews the existing information-based association measures and proposes a general framework, GIEF, to consistently detect associations and independence between different types of variables. Then, the study defines four mutually exclusive association relations of variables from an information-theoretic perspective to guide feature selection and compact prediction in high-dimensional processes. Based on GIEF and conditional mutual information maximization (CMIM), a new algorithm, CMIM-GIEF, is proposed and tested on a fluidized catalytic cracking (FCC) process with 217 variables, one which achieves significantly improved accuracies with fewer variables in predicting the yields of four crucial products. The compact variables identified are also consistent with the results of Shapley Additive exPlanations (SHAP) and industrial experience, proving good adaptivity of the method for chemical process data.

Fault Diagnosis of the Dynamic Chemical Process Based on the Optimized CNN-LSTM Network

Deep learning provides new ideas for chemical process fault diagnosis, reducing potential risks and ensuring safe process operation in recent years. To address the problem that existing methods have difficulty extracting the dynamic fault features of a chemical process, a fusion model (CS-IMLSTM) based on a convolutional neural network (CNN), squeeze-and-excitation (SE) attention mechanism, and improved long short-term memory network (IMLSTM) is developed for chemical process fault diagnosis in this paper. First, an extended sliding window is utilized to transform data into augmented dynamic data to enhance the dynamic features. Second, the SE is utilized to optimize the key fault features of augmented dynamic data extracted by CNN. Then, IMLSTM is used to balance fault information and further mine the dynamic features of time series data. Finally, the feasibility of the proposed method is verified in the Tennessee-Eastman process (TEP). The average accuracies of this method in two subdata sets of TEP are 98.29% and 97.74%, respectively. Compared with the traditional CNN-LSTM model, the proposed method improves the average accuracies by 5.18% and 2.10%, respectively. Experimental results confirm that the method developed in this paper is suitable for chemical process fault diagnosis.

A Novel Distributed Fault Detection Approach Based on the Variational Autoencoder Model

In large-scale industrial fault detection, a distributed model is typically established on the basis of blocked units. However, blocked distributed methods consider units as independent of one another and disregard the relationship between units, thus leading to incomplete information on local units. In fact, the operation status of a unit is affected by a local unit and its surrounding neighboring units. In addition, the fault detection performance of a system is seriously reduced once data are missing from the data source. Variational autoencoder (VAE) is not only a popular deep generative model but also has a powerful nonlinear feature extraction capability. In this study, VAE is extended to the distributed case. In this study, a distributed fault detection method DVAE based on VAE is proposed. This method can not only describe local and neighboring information, but it can also reconstruct missing data. First, system variables are divided into local and neighboring units in accordance with the system mechanism. Second, for each local unit, a DVAE model is established to map the multivariable data onto the latent variable space. The obtained latent variable contains the information on a local unit and can reflect the complex relationship with its neighboring units. Lastly, Euclidean distance is used to detect system faults. When applied on the Tennessee Eastman process for verification, the proposed method shows good performance in fault detection and missing data reconstruction.

An Open-Source Wireless Sensor Node Platform with Active Node-Level Reliability for Monitoring Applications

In wireless sensor networks, the quality of the provided data is influenced by the properties of the sensor nodes. Often deployed in large numbers, they usually consist of low-cost components where failures are the norm, even more so in harsh outdoor environments. Current fault detection techniques, however, consider the sensor data alone and neglect vital information from the nodes’ hard- and software. As a consequence, they can not distinguish between rare data anomalies caused by proper events in the sensed data on one side and fault-induced data distortion on the other side. In this paper, we contribute with a novel, open-source sensor node platform for monitoring applications such as environmental monitoring. For long battery life, it comprises mainly low-power components. In contrast to other sensor nodes, our platform provides self-diagnostic measures to enable active node-level reliability. The entire sensor node platform including the hardware and software components has been implemented and is publicly available and free to use for everyone. Based on an extensive and long-running practical experiment setup, we show that the detectability of node faults is improved and the distinction between rare but proper events and fault-induced data distortion is indeed possible. We also show that these measures have a negligible overhead on the node’s energy efficiency and hardware costs. This improves the overall reliability of wireless sensor networks with both, long battery life and high-quality data.