Latent variable (LV) models have been widely used in multivariate statistical process monitoring. However, whatever deviation from nominal operating condition is detected, an alarm triggered based on classical monitoring methods. Therefore, they fail to distinguish real faults incurring dynamics anomalies normal deviations conditions. A new strategy slow feature analysis (SFA) proposed for the concurrent of point and anomalies. Slow features as LVs are developed describe slowly varying...

Alarm systems play critically important roles for the safe and efficient operation of modern industrial plants. However, most existing alarm suffer from poor performance, noticeably having too many alarms to be handled by operators in control rooms. Such overloading is extremely detrimental role played systems. This paper provides an overview Four main causes are identified as culprits overloading, namely, chattering due noise disturbance, variables incorrectly configured, design isolated...

The detection of direct causality, as opposed to indirect is an important and challenging problem in root cause hazard propagation analysis. Several methods provide effective solutions this when linear relationships between variables are involved. For nonlinear relationships, currently only overall causality analysis can be conducted, but cannot identified for such processes. In paper, we describe a approach suitable both connections. Based on extension the transfer entropy approach, (DTE)...

Latent variable (LV) models provide explicit representations of underlying driving forces process variations and retain the dominant information data. In this study, slow features (SFs) as temporally correlated LVs are derived using probabilistic SF analysis. SFs evolving in a state‐space form effectively represent nominal processes, some which potentially to quality variables hence help improving prediction performance soft sensors. An efficient expectation maximum algorithm is proposed...

Recently, a new process monitoring and fault diagnosis method based on slow feature analysis has been developed, which enables concurrent of both operating point dynamics. In this paper, recursive algorithm for adaptive is put forward to accommodate time-varying processes by updating model parameters statistics once sample arrives. An important algebraic property first established. We then show that such can be violated online with forgetting factor used, remedy suggested. A novel the...

In chemical industrial processes, some quality variables are difficult to measure, and thus soft sensors have been proposed as an effective solution. Deep learning has introduced in deal with the complex nonlinearity of process, yet lacking ability for dynamics. This paper introduces Long Short-Term Memory (LSTM) develops a deep neural network structure based on LSTM sensor method strong dynamics process. The effectiveness improved modeling is validated by sulfur recovery unit benchmark....

In industrial processes, soft sensor models are commonly developed to estimate values of quality-relevant variables in real time. order take advantage the correlations between process variables, two convolutional neural network (CNN)-based this work. By making use unique architecture CNN, first model is capable utilizing abundant data, and complexity remains low. The second integrates finite impulse response dynamics can be reasonably embraced model. effectiveness validated by a simulation...

This paper presents an algorithm for classifying single-trial electroencephalogram (EEG) during the preparation of self-paced tapping. It combines common spatial subspace decomposition with Fisher discriminant analysis to extract features from multichannel EEG. Three are obtained based on Bereitschaftspotential and event-related desynchronization. Finally, a perceptron neural network is trained as classifier. was applied data set <self-paced 1s> "BCI Competition 2003" classification accuracy...

Plant‐wide oscillations are common in many industrial processes. They may impact the overall process performance and reduce profitability. It is important to detect diagnose such oscillations. This paper reviews advances diagnosis of plant‐wide The main focus this study on identifying possible root causes using two techniques, one based data analysis temporal spectral domains other connectivity analysis. data‐based provides an effective way capture difference between cause variable secondary...

Signed directed graph based modeling and its validation from process knowledge data This paper is concerned with the fusion of information connectivity subsequent use in fault diagnosis hazard assessment. The Directed Graph (SDG), as a graphical model for capturing topology to show causal relationships between variables by material paths, has been widely used root cause propagation analysis. An SDG usually built on described piping instrumentation diagrams. complex experience-dependent task,...

Detection of causality is an important and challenging problem in root cause hazard propagation analysis. It has been shown that the transfer entropy approach a very useful tool quantifying directional causal influence for both linear nonlinear relationships. A key assumption this method sampled data should follow well-defined probability distribution; yet may not hold some industrial process data. In paper, new information theory-based measure, 0-entropy (T0E), proposed analysis on basis...

Deep learning technology has been widely used in fault diagnosis for chemical processes. However, most deep technologies currently adopted only use a single network stack or certain with multilayer perceptron (MLP) behind it. Compared traditional technologies, this method made progress both the accuracy and speed, but due to limited performance of network, speed cannot meet requirements greatest extent. In order overcome such problems, article proposes using multimodel fusion. Different from...

Informally, a set of abstractions state space S is additive if the distance between any two states in always greater than or equal to sum corresponding distances abstract spaces. The first known abstractions, called disjoint pattern databases, were experimentally demonstrated produce art performance on certain However, previous applications restricted spaces with special properties, which precludes databases from being defined for several commonly used testbeds, such as Rubik's Cube, TopSpin...

In large-scale industrial processes, a fault can easily propagate between process units due to the interconnections of material and information flows. Thus problem detection isolation for these processes is more concerned about root cause propagation before applying quantitative methods in local models. Process topology causality, as key features description, need be captured from knowledge data. The modelling two aspects are overviewed this paper. From knowledge, structural equation...

Conventional data-driven soft sensors commonly rely on the assumption that processes are operating at steady states. As chemical involve evident dynamics, conventional may suffer from transient inaccuracy and poor robustness. In addition, control performance is unsatisfactory when outputs of serve as feedback signals for quality control. This brief develops a dynamic soft-sensing model combining finite impulse response support vector machine to describe nonlinear static relationships. The...