Multivariate sensor data collected from manufacturing and process industries represents actual operational behavior and can be used for predictive maintenance of the plants. Anomaly detection and diagnosis, that forms an integral part of predictive maintenance, in industrial systems is however challenging due to their complex behavior, interactions among sensors, corrective actions of control systems and variability in anomalous behavior. While several statistical techniques for anomaly detection have been in use for a long time, these are not particularly suited for temporal (or contextual) anomalies that are characteristic of multivariate time series sensor data. On the other hand, several machine learning and deep learning techniques for anomaly detection gained significant interest in the recent years. Further, anomaly diagnosis that involves localization of the faults did not receive much attention. In this work, we compare the anomaly detection and diagnosis capabilities, in semi-supervised mode, of several statistical, machine learning and deep learning techniques on two systems viz. the interacting quadruple tank system and the continuous stirred tank reactor (CSTR) system both of which are representative of the complexity of large industrial systems. The techniques studied include principal component analysis (PCA), Mahalanobis distance (MD), one-class support vector machine (OCSVM), isolation forest, elliptic envelope, dense auto-encoder and long short term memory auto-encoder (LSTM AE). The study revealed that MD and LSTM-AE have the highest anomaly detection capability, followed closely by PCA and OCSVM. The above techniques also exhibited good diagnosis capability. The study indicates that statistical techniques in spite of their simplicity could be as powerful as machine learning and deep learning techniques, and may be considered for anomaly detection and diagnosis in manufacturing systems.

No abstract

The 2018 PHM Data Challenge posed the problem of estimating Remaining Useful Life (RUL) for multiple faults in ion etch mills. As with any industrial system, run-to-failure data for the mills is not directly available and the mills experience more than one fault at the same time. We propose a novel data-driven methodology to address these challenges and develop a workflow that can be used for concurrent estimation of RUL for multiple faults in ion etch mills in real time. In the proposed approach, operational data of the ion etch mill is used to build a machine learning model for predicting a health score of the mill and to create a library of truncated degradation curves for each fault. These are then utilized for RUL predictions using Dynamic Time Warping (DTW) curve matching. Application of the proposed approach to test and validation datasets provided during the data challenge showed reasonable agreement between RUL predictions and the ground truth. The approach proposed here can be extended to other industrial systems and equipment for which historical operational data and failure information is available. This framework will help optimize health management and pave the way for predictive maintenance of industrial equipment.

Real-time fault detection, classification and diagnosis in manufacturing and process industries is essential to prevent unplanned downtime and improve the reliability of industrial operations. While several well-accepted machine learning techniques exist for fault detection and classification, there is a need for a reliable and generalized fault diagnosis technique that identifies sensors responsible for industrial faults in real time. In this work, we propose a variable perturbation matrix-based method for fault diagnosis in industrial processes. We utilize the Long ShortTerm Memory for prediction due to its ability to memorize temporal information in time-series data. First, the fault is detected, then one or more independent variables are perturbed across the fault detection point to check the sensitivity of the diagnosis model for the corresponding variables. Thus, a perturbation matrix is calculated and variables with high sensitivity are selected as the variables responsible for fault. The proposed method is applied to an Industry 4.0 quality control test bed set up for electronic components, the dataset for which is provided in the Prognostics and Health Management Euorpe-21 (PHME-21) data challenge. The proposed method accurately detected and classified 6 faults in the test bed and correctly diagnosed the most significant variables. Due to high fault detection accuracy coupled with sensitivity-based fault diagnosis, the method is suitable for multivariate industrial systems.

Real-time root cause identification (RCI) of faults or abnormal events in industries gives plant personnel the opportunity to address the faults before they progress and lead to failure. RCI in industrial systems must deal with their complex behavior, variable interactions, corrective actions of control systems and variability in faulty behavior. Bayesian networks (BNs) is a data-driven graph-based method that utilizes multivariate sensor data generated during industrial operations for RCI. Bayesian networks, however, require data discretization if data contains both discrete and continuous variables. Traditional discretization techniques such as equal width (EW) or equal frequency (EF) discretization result in loss of dynamic information and often lead to erroneous RCI. To deal with this limitation, we propose the use of a dynamic discretization technique called Bayesian Blocks (BB) which adapts the bin sizes based on the properties of data itself. In this work, we compare the effectiveness of three discretization techniques, namely EW, EF and BB coupled with Bayesian Networks on generation of fault propagation (causal) maps and root cause identification in complex industrial systems. We demonstrate the performance of the three methods on the industrial benchmark Tennessee-Eastman (TE) process.  For two complex faults in the TE process, the BB with BN method successfully diagnosed correct root causes of the faults, and reduced redundancy (up to 50%) and improved the propagation paths in causal maps compared to other two techniques.