Elsevier

Applied Soft Computing

Volume 8, Issue 1, January 2008, Pages 740-748
Applied Soft Computing

Artificial neural network approach for fault detection in rotary system

https://doi.org/10.1016/j.asoc.2007.06.002Get rights and content

Abstract

The detection and diagnosis of faults in technical systems are of great practical significance and paramount importance for the safe operation of the plant. An early detection of faults may help to avoid product deterioration, performance degradation, major damage to the machinery itself and damage to human health or even loss of lives. The centrifugal pumping rotary system is considered for this research. This paper presents the development of artificial neural network-based model for the fault detection of centrifugal pumping system. The fault detection model is developed by using two different artificial neural network approaches, namely feed forward network with back propagation algorithm and binary adaptive resonance network (ART1). The training and testing data required are developed for the neural network model that were generated at different operating conditions, including fault condition of the system by real-time simulation through experimental model. The performance of the developed back propagation and ART1 model were tested for a total of seven categories of faults in the centrifugal pumping system. The results are compared and the conclusions are presented.

Introduction

The problem of detecting faults in complex real plants is strategically important for its various implications, e.g., avoiding major plant breakdowns and catastrophes, safety problems, fast and appropriate response to emergency situations and plant maintenance. For instance, the following systems represent only a small part of systems where fault detection is in general a very difficult, yet important task: chemical plants, refineries, power plants, airplanes, ships, submarines, space vehicles and space stations, automobiles and household appliances. Generally, in process industries, there is a crucial need for checking and monitoring the equipment condition precisely since they are mostly subject to hazardous environments, such as severe shocks, vibration, heat, friction, dust, etc. So fault detection, fault identification and diagnosis of equipments, machineries and systems have become a vigorous area of work. Due to the broad scope of the process fault diagnosis problem and the difficulties in its real-time solution, many analytical-based techniques [1], [2] have been proposed during the past several years for the fault detection of technical plants. The important aspect of these approaches is the development of a model that describes the ‘cause and effect’ relationships between the system variables using state estimation or parameter estimation techniques. The problem with these mathematical model-based techniques is that under real conditions, no accurate models of the system of interest can be obtained. In that case, the better strategy is of using knowledge-based techniques where the knowledge is derived in terms of facts and rules from the description of system structure and behaviour. Classical expert systems were used for this purpose. The major weakness of this approach is that binary logical decisions with Boolean operators do not reflect the gradual nature of many real world problems. Recently, with the development of artificial intelligence, Computational Intelligence (CI) methods, Neural Networks (NN), Fuzzy Logic (FL), Evolutionary Algorithms (EA), etc., more and more fault diagnostic approaches have emerged as new techniques for fault diagnostic systems [3], [4].

Any complex system is liable to faults or failures. A ‘fault’ is an unexpected change of the system functionality. It manifests as a deviation of at least one characteristic property or variable of a technical process. It may not, however, represent the failure of physical components. Such malfunctions may occur either in the sensors (instruments) or actuators, or in the components of the process itself. In all but the most trivial cases, the existence of a fault may lead to situations related to safety, health, environmental, financial or legal implications. Although good design practice tries to minimize the occurrence of faults and failures, recognition that such events do occur, enables system designers to develop strategies by which the effect they exert is minimized. A system that includes the capability of detecting and diagnosing faults is called the ‘fault diagnosis system’ [5]. Such a system has to perform two tasks, namely fault detection and fault isolation. The purpose of the former is to recognize that a fault has occurred in the system. The latter has the purpose of locating the fault. The following are the set of desirable characteristics one would like the diagnostics system to possess: (a) Quick detection and diagnosis, (b) isolability, (c) robustness, (d) novelty identifiability, (e) classification error estimate, (f) adaptability, (g) explanation facility, (h) modeling requirements, (i) storage and computational requirements and (j) multiple fault identifiability.

Artificial neural network-based methods for fault diagnosis have received considerable attention over the last few years. The advantage of the neural network approach is their generalization capability, which lets them deal with partial or noisy inputs. The neural networks are able to handle continuous input data and the learning must be supervised in order to solve the fault detection and diagnosis problem. The multilayer perception network is the most common network today. Due to their powerful non-linear function approximation and adaptive learning capabilities, neural networks have drawn great attention in the field of fault diagnosis. But the neural network approach needs lot of data to develop the network before being put to use for real-time applications.

Adaptive resonance theory (ART) [6] refers to a class of self-organizing neural architectures that cluster the pattern space and produce appropriate weight vector templates. Conventional artificial neural networks have failed to solve the stability–plasticity dilemma. A network remains open to new learning (remain plastic) without washing away previously learned codes. Too often, learning a new pattern erase or modifies previous training. If there is only a fixed set of training vectors, the network can be cycled through these repeatedly and may eventually learn all. In a real network, it will be exposed to a constantly changing environment; it may never see the same training vector twice. Under such conditions, back propagation will learn nothing. It will continuously modify its weights to no avail, never arriving at satisfactory settings. ART networks and algorithms maintain the plasticity required to learn new patterns, while preventing the modification of patterns that have been learned previously. The objective of the present work is as follows:

  • To detect the fault based on the observed data.

  • To alert the operating personnel about deviations of normal behaviour of the processes that can turn into failures.

  • To monitor and predict the condition of the equipment/system and avoidance of failures and damage.

The paper is organized as follows: in the next section, the description of the experimental system for this study is outlined. Section 3 describes the review of artificial neural network. Sections 4 Model development, 5 Simulation results demonstrate the development of neural network model and detailed discussions on simulation results and finally, in Section 6, conclusions are drawn from the work.

Section snippets

System description

Centrifugal pumps are used in a variety of industrial applications, such as for lifting thin liquids as well as highly dense liquids, such as muddy and sewage water, paper pulp, sugar molasses, chemicals, etc. Centrifugal pumps are classified as rotodynamic type of pumps in which a dynamic pressure is developed which enables the lifting of liquids from a lower to a higher level. The basic principle on which a centrifugal pump works is that when a certain mass of liquid is made to rotate by an

Review of artificial neural network

Neural networks have recently attracted much attention based on their ability to learn complex, non-linear functions. Artificial neural networks [7], [8] can be viewed as parallel and distributed processing systems that consist of a huge number of simple and massively connected processors. These networks can be trained offline for complicated mapping, such as of determining the various faults and can then be used in an efficient way in the online environment.

Model development

The proposed methodology for fault detection in centrifugal pumping system is based on using back propagation and adaptive resonance network in artificial neural network (ANN) for detecting the normal and abnormal conditions of the given parameters, which leads to various faults. The neural network approach for this purpose has two phases; training and testing. During the training phase, neural network is trained to capture the underlying relationship between the chosen inputs and outputs.

Simulation results

This section presents the details of the development and testing of ANN models for fault detection on centrifugal pumping system. Two different ANN models were developed for fault detection, one with back propagation algorithm and the other with adaptive resonance theory (ART1). The neural network model is developed using MATLAB 6.5 Neural Network Toolbox in Pentium 4 with 2.40 GHz processor with 512 MB of RAM. Using real-time simulation on laboratory experimental setup of centrifugal pumping

Conclusion

This paper has presented a neural network-based approach for fault diagnosis in centrifugal pumping system. The data required for the development of neural network model have been obtained through the real-time operational data of the system considered. Totally 7 categories of faults including 20 faults from the centrifugal pumping system were considered in the developed model. For the ANN model the testing data are fed to the designed model to check the accuracy. The testing samples are

References (12)

There are more references available in the full text version of this article.

Cited by (155)

  • A multi-task learning for cavitation detection and cavitation intensity recognition of valve acoustic signals

    2022, Engineering Applications of Artificial Intelligence
    Citation Excerpt :

    Zouari et al. (2004) proposed a vibration signal fault detection method for centrifugal pumps using neural network and neuro-fuzzy technology. Rajakarunakaran et al. (2008) proposed a centrifugal pump fault detection using a feedforward neural network and a binary adaptive resonance network (ART1). Siano and Panza (2018) proposed a method combining ANN and nonlinear regression to diagnose cavitation of time domain vibration signals.

View all citing articles on Scopus
View full text