Industrial processes are large scale, highly complex systems. The flow of mass and energy, as well the compensation effects closed-loop control systems, cause significance cross-correlation autocorrelation between process variables. To operate systems stably efficiently, it is crucial to uncover inherent characteristics both variance structure dynamic relationship. Long-term dependency slow fea...

An RNN-based forecasting approach is used to early detect anomalies in industrial multivariate time series data from a simulated Tennessee Eastman Process (TEP) with many cyberattacks. This work continues a previously proposed LSTM-based approach to the fault detection in simpler data. It is considered necessary to adapt the RNN network to deal with data containing stochastic, stationary, trans...

Orthonormal subspace analysis (OSA) is proposed for handling the decomposition issue and principal component selection in traditional key performance indicator (KPI)-related process monitoring methods such as partial least squares (PLS) canonical correlation (CCA). However, it not appropriate to apply static OSA algorithm a dynamic since pays no attention auto-correlation relationships variable...

In the recent years, deep learning has been widely used in process monitoring due to its strong ability extract features. However, with increasing layers of network, compression features by model will lead loss some valuable information and affect model’s performance. To solve this problem, a fault detection method based on discriminant enhanced stacked auto-encoder is proposed. An network stru...

In this paper, we present a Bayesian framework for fault detection. Principal component analysis (PCA) technique and quadratic test statistics are incorporated under a Conditional Gaussian Network (CGN). The proposed network, given an observation, is able to project it into an orthogonal space and to give a decision about the system state (faulty or not). The paper demonstrate, the probability ...

The fact that modern Networked Industrial Control Systems (NICS) depend on Information and Communications Technologies (ICT) is well known. Although many studies have focused on the security of NICS, today we still lack a proper understanding of the impact that network parameters, e.g. network delays, packet losses, background traffic, and network design decisions, have on cyber attacks targeti...

A feed-forward neural network is proposed for monitoring operating modes of large scale processes. A Gaussian hidden layer associated with a Kohonen output layer map the principal features of measurements of state variables. Subsets of selective neurons are generated into the hidden layer by means of self adapting of centers and dispersions parameters of the Gaussian functions. The output layer...

Abstract With the rapid development of deep learning methods, variational auto-encoder (VAE) has been utilized for nonlinear process monitoring. However, most VAE-based methods hardly consider inner independent and related relationship each variable. To overcome this problem, a novel VAE named variable information concentrated (IRVIC-VAE) is proposed. concentrate information, loading weight mat...

The concept of globally optimal controlled variable selection has recently been proposed to improve self-optimizing control performance of traditional local approaches. However, the associated measurement subset selection problem has not be studied. In this paper, we consider the measurement subset selection problem for globally self-optimizing control (gSOC) of Tennessee Eastman (TE) process. ...