Data-based process monitoring has become a key technology in industries for safety, quality, and operation efficiency enhancement. This paper provides timely update review on this topic. First, the natures of different industrial processes are revealed with their data characteristics analyzed. Second, detailed terminologies data-based method illustrated. Third, based each main that exhibits process, corresponding problem is defined illustrated, conducted discussions connection comparison...

Data mining and analytics have played an important role in knowledge discovery decision making/supports the process industry over past several decades. As a computational engine to data analytics, machine learning serves as basic tools for information extraction, pattern recognition predictions. From perspective of learning, this paper provides review on existing applications The state-of-the-art are reviewed through eight unsupervised ten supervised algorithms, well application status...

Soft sensors are widely constructed in process industry to realize monitoring, quality prediction, and many other important applications. With the development of hardware software, industrial processes have embraced new characteristics, which lead poor performance traditional soft sensor modeling methods. Deep learning, as a kind data-driven approach, shows its great potential fields, well sensing scenarios. After period development, especially last five years, issues emerged that need be...

Data-driven soft sensors have been widely utilized in industrial processes to estimate the critical quality variables which are intractable directly measure online through physical devices. Due low sampling rate of variables, most developed on small number labeled samples and large unlabeled process data is discarded. The loss information greatly limits improvement prediction accuracy. One main issues data-driven sensor furthest exploit contained all available data. This paper proposes a...

In order to deal with the modeling and monitoring issue of large-scale industrial processes big data, a distributed parallel designed principal component analysis approach is proposed. To handle high-dimensional process variables, first decomposed into blocks priori knowledge. Afterward, in solve data chunks each block, processing strategy proposed based on framework MapReduce then components are further extracted for block. With all these steps, statistical can be established. Finally,...

For plant-wide process monitoring, most traditional multiblock methods are under the assumption that some knowledge should be incorporated for dividing into several sub-blocks. However, is not always available in practice. In this case, monitoring scheme implemented through an automatic way. This paper intends to develop a new sub-block principal component analysis (PCA) method which named as distributed PCA model. By constructing sub-blocks different directions of components, original...

Many of the current multivariate statistical process monitoring techniques (such as principal component analysis (PCA) or partial least squares (PLS)) do not utilize non-Gaussian information data. This paper proposes a new method based on independent analysis−principal (ICA−PCA). The Gaussian and can be extracted for fault detection diagnosis. Moreover, mixed similarity factor is proposed. used to identify mode. Because non-orthogonal nature components, "main angle" proposed calculate...

Industrial process plants are instrumented with a large number of redundant sensors and the measured variables often contaminated by random noises. Thus, it is significant to discover general trends data latent variable models in probabilistic framework before soft sensor modeling. However, traditional such as principal component analysis mostly static linear approaches. The dynamics nonlinearities have not been well considered. In this paper, novel weighted dynamic system (WLDS) proposed...

Just-in-time learning (JITL) is a commonly used technique for industrial soft sensing of nonlinear processes. However, traditional JITL approaches mainly focus on equal sample sizes between process (input) variables and quality (output) variables, which may not be practical in processes since are usually much harder to obtain than other variables. In order handle unequal length dataset with only few labeled data, novel semisupervised framework proposed sensor modeling processes, based...

In this paper, a new fault detection and identification scheme that is based on the global–local structure analysis (GLSA) model proposed. By exploiting underlying geometrical manifold simultaneously keeping global data information, GLSA constructs dual-objective optimization function for dimension reduction of process dataset. It combines advantages both locality preserving projections (LPP) principal component (PCA), under unified framework. Meanwhile, can successfully avoid singularity...

The principal component regression (PCR) based soft sensor modeling technique has been widely used for process quality prediction in the last decades. While most industrial processes are characterized with nonlinearity and time variance, global linear PCR model is no longer applicable. Thus, its nonlinear adaptive forms should be adopted. In this paper, a just-in-time learning (JITL) locally weighted kernel (LWKPCR) proposed to solve time-variant problems of process. Soft sensing performance...

Abstract This article intends to address two drawbacks of the traditional principal component analysis (PCA)‐based monitoring method: (1) nonprobabilistic; (2) single operation mode assumption. On basis framework probabilistic PCA (PPCA), a Bayesian regularization method is introduced for performance improvement, through which effective dimensionality latent variable can be determined automatically. For processes with multiple modes, extended its mixture form, thus PPCA has been developed....

With the growing complexity of modern industrial process, monitoring large-scale plant-wide processes has become quite popular. Unlike traditional processes, measured data in process pose great challenges to information capture, management, and storage. More importantly, it is difficult efficiently interpret hidden within those data. In this paper, road map a distributed modeling framework for introduced. Based on framework, whole decomposed into different blocks, statistical models are...

Just-in-time learning (JITL) is one of the most widely used strategies for soft sensor modeling in nonlinear processes. However, traditional JITL methods have difficulty dealing with data samples that contain missing values. Meanwhile, noises and uncertainties not been taken into consideration relevant sample selection existing approaches. To overcome these problems, a new probabilistic (P-JITL) framework proposed this brief. In P-JITL, variational Bayesian principal component analysis first...

The problem of fault classification in industry has been studied extensively. Most algorithms are modeled on the premise data balance. However, difficulty collecting industrial different modes is quite different. This inevitably leads to imbalance, which will adversely affect performance. article proposes a novel augmentation classifier (DAC) for imbalanced classification. Data based generative adversarial networks (GANs) an effective way solve unbalanced randomness GAN generation process...