System Decomposition for Distributed Multivariate Statistical Process Monitoring by Performance Driven Agglomerative Clustering

Khatib, Shaaz; Daoutidis, Pródromos; Almansoori, Ali

doi:10.1021/acs.iecr.8b01708

Cited by 20 publications

(17 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The key step of statistical process monitoring is to define normal operating regions by applying statistical techniques to data samples obtained from the process system. Typical examples of such techniques include principal component analysis (PCA) [3][4][5][6], partial least squares [7][8][9], independent component analysis [10,11], and support vector machine [12,13]. Any data sample that does not lie in the normal operating region is then classified as a fault, and its root cause needs to be identified through fault diagnosis.…”

Section: Introductionmentioning

confidence: 99%

Statistical Process Monitoring of the Tennessee Eastman Process Using Parallel Autoassociative Neural Networks and a Large Dataset

Heo

Lee

2019

Processes

View full text Add to dashboard Cite

In this article, the statistical process monitoring problem of the Tennessee Eastman process is considered using deep learning techniques. This work is motivated by three limitations of the existing works for such problem. First, although deep learning has been used for process monitoring extensively, in the majority of the existing works, the neural networks were trained in a supervised manner assuming that the normal/fault labels were available. However, this is not always the case in real applications. Thus, in this work, autoassociative neural networks are used, which are trained in an unsupervised fashion. Another limitation is that the typical dataset used for the monitoring of the Tennessee Eastman process is comprised of just a small number of data samples, which can be highly limiting for deep learning. The dataset used in this work is 500-times larger than the typically-used dataset and is large enough for deep learning. Lastly, an alternative neural network architecture, which is called parallel autoassociative neural networks, is proposed to decouple the training of different principal components. The proposed architecture is expected to address the co-adaptation issue of the fully-connected autoassociative neural networks. An extensive case study is designed and performed to evaluate the effects of the following neural network settings: neural network size, type of regularization, training objective function, and training epoch. The results are compared with those obtained using linear principal component analysis, and the advantages and limitations of the parallel autoassociative neural networks are illustrated.

show abstract

Section: Introductionmentioning

confidence: 99%

Statistical Process Monitoring of the Tennessee Eastman Process Using Parallel Autoassociative Neural Networks and a Large Dataset

Heo

Lee

2019

Processes

View full text Add to dashboard Cite

show abstract

“…In a distributed monitoring method, a decomposition of the system must first be selected where a set of subsystems is defined, each of which contains a different set of sensors. The monitoring method is then applied to each of these subsystems, and the monitoring results of the subsystems are combined using a consensus strategy . For example, in a distributed pattern recognition method, each subsystem uses the pattern recognition method to diagnose a faulty sample of the measurements of its local sensors.…”

Section: Introductionmentioning

confidence: 99%

“…Feature selection for a distributed monitoring method, i.e., the selection of the system decomposition, has a significant impact on its performance. , Some distributed monitoring methods select a decomposition based on the layout of the plant so that sensors in the same unit are allocated to the same subsystem. ,,, The distributed monitoring methods in refs − use normal operation data to place sensors with strong correlation between them in the same subsystem. In refs − , normal operation data is used to place sensors with a similar probability distribution within the same subsystem.…”

Section: Introductionmentioning

confidence: 99%

“…In ref , the decomposition is selected by applying Sparse Principal Component Analysis to the process data so that most of the variance of the sensors in each subsystem is explained along a single orthogonal direction, and the sensors in each subsystem satisfy certain operational constraints. Note that the optimal decomposition depends on the distributed monitoring method that is to be implemented and the faults that are expected to affect the system . These factors are not taken into account in the above methods.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Optimal Feature Selection for Distributed Data-Driven Process Monitoring

Khatib

Daoutidis

2019

Ind. Eng. Chem. Res.

Self Cite

View full text Add to dashboard Cite

Two methods are proposed in this article to address distributed data-driven process monitoring. The first method selects sensors and allocates these sensors among subsystems with the objective of optimizing the diagnostic performance of a pattern recognition method when it is implemented in a distributed configuration subject to constraints that are imposed to meet operational requirements. The second method finds the minimum number of monitoring computers required to implement the methods used for fault detection and diagnosis in a distributed configuration subject to certain operational constraints. It also finds the sensors that need to be transmitted to each of the monitors. The proposed methods rely on concepts and algorithms from graph theory and optimization, and their efficiency is shown through a case study on the benchmark Tennessee Eastman Process.

show abstract

“…In Ref. , a distributed multivariate statistical process monitoring is proposed where system partitioning is carried out using agglomerative clustering search strategy to find the optimal system decomposition. Recent results have enabled diagnosis of simultaneous actuator and sensor faults in nonlinear uncertain networked systems including faults in sensors (but not actuators) in shared interconnections, that is, sensors that affect multiple units (see e.g., Ref.…”

Section: Introductionmentioning

confidence: 99%

Distributed fault diagnosis of heating, ventilation, and air conditioning systems

et al. 2018

View full text Add to dashboard Cite

The problem of distributed fault detection and isolation (FDI) for heating, ventilation, and air conditioning (HVAC) systems has been addressed in this work. First, a linear model is identified for subunits of an HVAC system. Next, a local FDI (LFDI) framework is designed for each unit under consideration. A distributed FDI architecture is designed where the LFDI frameworks communicate to exchange information to achieve enhanced FDI in each unit. As a result, each LFDI framework functions as intended even in the presence of faults that affect multiple units. Effectiveness of the proposed distributed FDI framework is shown for various commonly occurring fault scenarios. © 2018 American Institute of Chemical Engineers AIChE J, 65: 640–651, 2019

show abstract

System Decomposition for Distributed Multivariate Statistical Process Monitoring by Performance Driven Agglomerative Clustering

Cited by 20 publications

References 36 publications

Statistical Process Monitoring of the Tennessee Eastman Process Using Parallel Autoassociative Neural Networks and a Large Dataset

Statistical Process Monitoring of the Tennessee Eastman Process Using Parallel Autoassociative Neural Networks and a Large Dataset

Optimal Feature Selection for Distributed Data-Driven Process Monitoring

Distributed fault diagnosis of heating, ventilation, and air conditioning systems

Contact Info

Product

Resources

About