Process Fault Detection Using Hierarchical Artificial Neural Network Diagnostic Strategy

Abstract

Kertas kerja ini menerangkan mengenai kegunaan jaringan neural tiruan (ANN) untuk mengesan dan membaiki kesilapan dalam loji proses. Dalam penyelidikan ini, ANN menggunakan dua lapisan dalam strategi diagnostik hirarki. Lapisan pertama mengenal pasti nod di mana kesilapan bermula sementara lapisan kedua membahagikan kesilapan yang berlaku pada nod tertentu. Arkitek model ANN adalah berasaskan beberapa lapisan rangkaian suapan hadapan dan menggunakan algoritma luncuran belakang dalam skema latihan. Untuk mendapatkan konfigurasi ANN yang terbaik, analisis topologi dilakukan. Keberkesanan kaedah ini ditunjukkan oleh kajian kes melibatkan turus pemecahan asid lemak. Keputusan menunjukkan sistem ini berjaya mengesan kesilapan tunggal dan fana yang terdapat dalam proses tersebut. Kata kunci: Pengenalpastian dan diagnostik kesilapan proses, strategi diagnostik hirarki, jaringan neural tiruan, turus pemecahan asid lemak This paper focuses on the use of artificial neural network (ANN) to detect and diagnose fault in process plant. In this work, the ANN uses two layers of hierarchical diagnostic strategy. The first layer diagnoses the node where the fault originated and the second layer classifies the type of faults or malfunctions occurred on that particular node. The architecture of the ANN model is founded on a multilayer feed forward network and used back propagation algorithm as the training scheme. In order to find the most suitable configuration of ANN, a topology analysis is conducted. The effectiveness of the method is demonstrated by using a fatty acid fractionation column. Results show that the system is successful in detecting original single and transient fault introduced within the process plant model. Key words: Process fault detection and diagnosis, hierarchical diagnostic strategy, artificial neural network, fatty acid fractionation column

Machine Learning and Deep Learning in Chemical Health and Safety: A Systematic Review of Techniques and Applications

“…The hierarchical neural network (HNN) and fuzzy neural network (FNN) are two approaches to improve the performance of a single hidden layer neural network. Othman et al and Rusinov et al used the HNN to operate faults diagnosis. , This strategy was able to shorten the training time and narrow the diagnostic focus of the system from node to type of fault. Even if the cause of the abnormal situation was not recognized by the lower-level neural network, the location information was obtained by the higher-level neural network.…”

Diagnostics of sintering processes on the basis of PCA and two‐level neural network model

“…That is why, it is important to ensure small time for network retraining in the case of possible change in the normal process state (for example, as a result of a drift of the normal state of the process). This can be achieved by introducing a hierarchy in the model structure . The latter entails a necessity for process decomposition and separation of a number of structural units (blocks), which should be as autonomous as possible.…”

“…This can be achieved by introducing a hierarchy in the model structure. 17,21 The latter entails a necessity for process decomposition and separation of a number of structural units (blocks), which should be as autonomous as possible. Thus, the high-level neural network (HL-NN) is responsible for the fault localization.…”

“…The NNs were mainly used, however, for simulating the sintering process, and there are only a few articles devoted to diagnostics of this process faults . This fact was possibly determined by the abovementioned shortcomings of the NN approach.…”

Hierarchically Distributed Fault Detection and Identification through Dempster–Shafer Evidence Fusion

“…Faults as well as identification methods can be organized following the compositional hierarchy of the process. Hierarchical neural network classifiers have been proposed − consisting of a top-level neural network that examines the overall process and identifies the section in which the fault originates. Each of the lower level neural networks is trained to classify the faults in that particular section.…”