A deep learning model for process fault prognosis

Arunthavanathan, Rajeevan; Khan, Faisal; Ahmed, Salim; Imtiaz, Syed

doi:10.1016/j.psep.2021.08.022

Cited by 111 publications

(24 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The dependence of processes on engineered devices also gives rise to concerns about their safety and integrity (Khan et al, 2021). Emergent technologies play an important role in increasing process safety (Ahmed, 2021;Sajid et al, 2021), for example, through DT-assisted process fault prognosis (Arunthavanathan et al, 2021;Adumene et al, 2021). From a lifecycle perspective, BPM includes activities, such as process identification, discovery, analysis, implementation, execution, monitoring, controlling and improvement (Recker and Mendling, 2016;Kerpedzhiev et al, 2020).…”

Section: Business Process Management and Improvementmentioning

confidence: 99%

See 1 more Smart Citation

Success factors of process digitalization projects – insights from an exploratory study

Baier

Lockl

Röglinger

et al. 2022

BPMJ

View full text Add to dashboard Cite

PurposeIn an exploratory approach, the authors conducted a structured literature review to extract candidate process digitalization project (PDP) success factors (SFs) from the literature on business process management (BPM), project management (PM) and digitalization. After that, the authors validated, refined and extended these intermediate results through interviews with 21 members of diverse PDP teams. Finally, the authors proposed the PDP success model by linking the candidate SFs with relevant success criteria.Design/methodology/approachDigitalization substantially impacts organizations, which increasingly use digital technologies (DTs) to improve and innovate their business processes. While there are methods and tools for identifying process digitalization ideas and related projects (PDPs), guidance on the successful implementation of PDPs is missing. Hence, the authors set out to explore PDP SFs.FindingsThe PDP success model covers 38 PDP success factor candidates, whereof 28 are already backed by the literature and ten have emerged during the interviews. Furthermore, the SFs are structured according to seven categories from the literature covering a broad range of sociotechnical topics (i.e. strategy, structure, culture, people, process, project and technology) as well as equipped with preliminary success rationales.Originality/valueThe work is the first to systematically explore PDP SFs. The PDP success model shows that PDPs require a unique set of SFs, which combine established and hitherto underrepresented knowledge. It extends the knowledge on BPM and serves as foundation for future (confirmatory) research on business process digitalization and the successful implementation of PDPs.

show abstract

Section: Business Process Management and Improvementmentioning

confidence: 99%

“…Emergent technologies play an important role in increasing process safety (Ahmed, 2021; Sajid et al. , 2021), for example, through DT-assisted process fault prognosis (Arunthavanathan et al. , 2021; Adumene et al.…”

Section: Introductionmentioning

confidence: 99%

Success factors of process digitalization projects – insights from an exploratory study

Baier

Lockl

Röglinger

et al. 2022

BPMJ

View full text Add to dashboard Cite

show abstract

“…Further, to eliminate suspected outliers from the training data, the concept of automatic adjustment in tuning parameters is proposed iteratively, using a bisection algorithm . Arunthavanathan et al proposed the method to reduce the impact of the noise based on the anomaly margin in the training data frame using one-class SVM …”

Section: Introductionmentioning

confidence: 99%

“…18 Arunthavanathan et al proposed the method to reduce the impact of the noise based on the anomaly margin in the training data frame using one-class SVM. 23 An incremental one-class SVM was initially proposed based on the batch algorithm to detect multiple deviation points. 19 An incremental one-class approach was developed using the learning by the model algorithm.…”

Section: Introductionmentioning

confidence: 99%

Autonomous Fault Diagnosis and Root Cause Analysis for the Processing System Using One-Class SVM and NN Permutation Algorithm

Arunthavanathan

Khan

Ahmed

et al. 2022

Ind. Eng. Chem. Res.

Self Cite

View full text Add to dashboard Cite

In this era of Industry 4.0, there are continuing efforts to develop fault detection and diagnosis methods that are fully autonomous; these methods are self-learning, with little or no human intervention. This paper proposes a methodology for the autonomous diagnosis of the root cause of a detected fault in a complex processing system. The methodology comprises steps to detect and classify any newly encountered fault, classify the known faults, and find the root cause of the detected fault condition. The one-class support vector machine (SVM) model is used in the framework to detect the unlabeled fault, and the neural network is used for fault classification and root cause analysis. The developed methodology is capable of self-updating the fault database by detecting and diagnosing any new fault condition. A permutation algorithm is applied in the neural network framework to extract the variable’s contribution to the classified fault condition. Also, Spearman’s rank correlation approach is used to investigate and justify the data correlation and causation. The proposed framework is tested using a continuous stirred tank heater and the benchmark Tennessee Eastman process.

show abstract

“…Zhang et al design a fault detection index based on log conditional probability of the CRF. 33 Arunthavanathan et al 34 utilize convolution neural network and long short-term memory for fault sequence learning.…”

Section: Introductionmentioning

confidence: 99%

Linear Chain Conditional Random Field for Operating Mode Identification and Multimode Process Monitoring

Wang

2022

ACS Omega

View full text Add to dashboard Cite

As a supervised machine learning algorithm, conditional random fields are mainly used for fault classification, which cannot detect new unknown faults. In addition, faulty variable location based on them has not been studied. In this paper, conditional random fields with a linear chain structure are utilized for modeling multimode processes with transitions. A linear chain conditional random field model is trained by normal data with mode label. This model is able to distinguish transitions from stable modes well. After mode identification, the expectation of state feature function is developed for fault detection and faulty variable location. Case studies on the Tennessee Eastman process and continuous stirred tank reactor (CSTR) testify the effectiveness of the proposed approach.

show abstract

A deep learning model for process fault prognosis

Cited by 111 publications

References 29 publications

Success factors of process digitalization projects – insights from an exploratory study

Success factors of process digitalization projects – insights from an exploratory study

Autonomous Fault Diagnosis and Root Cause Analysis for the Processing System Using One-Class SVM and NN Permutation Algorithm

Linear Chain Conditional Random Field for Operating Mode Identification and Multimode Process Monitoring

Contact Info

Product

Resources

About