Showing papers on "Condition monitoring published in 1995"

Condition-based maintenance: tools and decision making

Abstract: The primary objective of equipment maintenance programmes is to preserve system functions in a cost‐effective manner. Surveys a wide variety of techniques commonly used to monitor the condition of mechanical systems. Reviews some of the decision models designed to address this type of inspection problem; concludes with some suggestions for future research directions in this type of decision problem.

Learning methodology for failure detection and accommodation

Abstract: A major goal of intelligent control systems is to achieve high performance with increased reliability, availability, and automation of maintenance procedures. In order to achieve fault tolerance in dynamical systems many algorithms have been developed during the past two decades. Fault diagnosis and accommodation methods have traditionally been based on linear modeling techniques, which restricts the type of practical failure situations that can be modeled. This article presents a learning methodology for failure detection and accommodation. The main idea behind this approach is to monitor the physical system for any off-nominal behavior in its dynamics using nonlinear modeling techniques. The principal design tool used is a generic function approximator with adjustable parameters, referred to as online approximator. Examples of such structures include traditional approximation models such as polynomials and splines as well as neural networks topologies such as sigmoidal multilayer networks and radial basis function networks. Stable learning methods are developed for monitoring the dynamical system. The nonlinear modeling nature and learning capability of the estimator allow the output of the online approximator to be used not only for detection but also for identification and accommodation of system failures. Simulation studies are used to illustrate the learning methodology and to gain intuition into the effect of modeling uncertainties on the performance of the fault diagnosis scheme. >

Danger avoidance system for a vehicle

Abstract: The present invention relates to a danger avoidance system capable of avoiding the collision of a vehicle against obstacles and of reducing the effect of collision on a vehicle. A traveling condition deciding unit decides whether or not the vehicle is in a dangerous traveling condition on the basis of the respective outputs of a vehicle monitoring system and an ambient condition monitoring system. A danger recognition confirming unit decides whether or not the driver is aware of the dangerous traveling condition on the basis of the respective outputs of a driver monitoring system, the ambient condition monitoring system and the vehicle monitoring system. When the traveling condition deciding unit decides that the vehicle is in the dangerous traveling condition, measures to avoid danger or to reduce the effect of collision on the vehicle are taken only when the driver is not aware of the dangerous traveling condition. Therefore, a danger avoidance effecting unit does not execute the measures to avoid danger or to reduce the effect of collision on the vehicle when the driver is aware of the dangerous traveling condition.

On-line failure diagnosis for compression refrigeration plants

Abstract: A prototype condition monitoring and diagnostic system has been developed for compression refrigeration plants, which can be used under variable operational conditions. Based on a combination of causal analysis, expert knowledge and simulated failure modes, a failure mode symptom matrix has been created. Healthy system behaviour is predicted based on a regression analysis model. Using multi-valued (or ‘fuzzy’) logic, real-time recognition of failure modes, at an early stage, proved to be possible. Future developments for improvement of diagnostic systems in compression refrigeration plants are discussed.

Automated Monitoring of Manufacturing Processes, Part 2: Applications

Abstract: In Part 2 of this paper, three applications of automated monitoring of manufacturing processes are presented to demonstrate the use of monitoring methods discussed in Part 1 of the paper. These applications are: (1) tool condition monitoring in turning, (2) machining condition monitoring in tapping, and (3) metallographic condition monitoring in arc welding. For each application, a background review, monitoring index selection and experimental setup are first presented. Then, monitoring methods discussed in Part 1 of the paper were applied and the test results are investigated. Discussions of the monitoring success rate, sensitivity, robustness, monitoring index selection, and decision under uncertainty are also included.

Analysis of a 2-phase model for optimization of condition-monitoring intervals

Abstract: Condition monitoring is a maintenance strategy where decisions are made depending on either continuously or regularly measured equipment states. It reduces uncertainty with respect to actual states of equipment, and can thus avoid unnecessary repair or replacement. However, it involves capital expenditure and/or operational costs to perform measurements. This paper presents a basic model for the economic evaluation and optimization of the interval between successive condition measurements (also called inspections), where measurements are expensive and cannot be made continuously. It assumes that the technique can detect an intermediate state to failure for a failure mode of interest. The influence of competing risks is analyzed, leading to the conclusion that once the cost-effectiveness of the condition-monitoring has been established, competing risks need not be considered in determining the optimum condition monitoring interval. Inspection is cost-effective if the intermediate state has a: (1) nondecreasing hazard rate, and (2) shorter mean residence time than the good state (good-as-new condition), while costs of failure are high enough compared with inspection and repair costs in the intermediate state. Assuming that the distribution of the residence time in the second state is unimodal, estimation of the mean (or scale parameter) and standard deviation of this state, in many cases, provides enough information to make a good decision on the inspection interval. The most important model parameters are identified by sensitivity analyses; it is shown that the model can be simplified without seriously affecting optimal decision making. >

The application of modern signal processing techniques to rotor fault detection and location within three phase induction motors

Abstract: Previous work at The Robert Gordon University has shown that faults within the rotors of large three phase induction motors, such as broken rotor bars, can be detected by monitoring and analysing the line current taken by the machine during a no-load starting transient. This line current has been shown to contain frequency components which are indicative of these fault conditions. Under transient conditions these components are however nonstationary in both the time and frequency domains. A variety of signal processing strategies applicable to time variant data such as the spectrogram, The Wigner Ville distribution and wavelet decomposition have been implemented and their success in detecting these non-stationary components evaluated. The most suitable of these techniques has been used to determine the occurrence and severity of motor faults. Preliminary work suggests that these techniques may also be used to detect frequency components indicative to the location of the fault

Decomposition of gear motion signals and its application to gearbox diagnostics

Abstract: The decomposition of gear motion and the related dynamic measurements for the condition monitoring and fault diagnosis of gearboxes are described. The motion error signal is separated according to fundamental frequencies into the harmonic error and the residual error, which are used to quantify the gear condition. High-order accelerations, such as jerk, are considered and shown to be more sensitive to some classes of early damage to gear teeth. Analysis of the time domain average of a gearbox casing vibration signal enables early detection of gear damage. Several methods to represent and enhance the fault information in the signal are introduced, based on the representation of different forms of the motion errors

Concurrent error detection using monitoring machines

Abstract: In circuits implementing system level functions, the correctness of the overall operation is critically dependent on the correctness of the control part. Therefore, concurrent error detection techniques for controllers implemented in integrated circuits have previously received wide attention. This paper presents a new technique for concurrent error detection in finite state machine (FSM) controllers. It is based on the use of monitoring machines. In a monitored FSM controller, an auxiliary monitoring machine operates in lock-step with the main FSM, such that any fault in either of the two machines is immediately detected. It is shown how the monitoring machine provides a uniform mechanism for the detection of stuck-at faults as well as delay faults. Besides being less costly than the main machine, it is also not identical to it. These features yield designs which compare very favourably with previous implementations. Not only is the fault coverage higher, also the hardware cost of the monitored sequential circuit is significantly lower.

Architectures and monitoring techniques for active databases: an evaluation

Abstract: The need for active capability for non-traditional applications and its concomitant benefits are well-established. Although the event-based technique for monitoring conditions (leading to the integrated architecture) is the most versatile of all the techniques, from a practical viewpoint there is a need for enhancing pre-existing non-active DBMSs to support active capability. The set of techniques that can be used for providing this add-on active capability (leading to the layered architecture) imposes certain limitations on the extent of active capability that can be supported. Insights into the details of techniques as well as their impact on the architecture entails a design that better meets the active database objectives. This paper identifies a repertoire of techniques for condition monitoring and discusses their suitability to different architectures. This paper argues that from a pragmatic viewpoint, both layerd and Integrated approaches to support active capability need to be pursued. Then it comes polling and event-based or asynchronous monitoring techniques using an implementation on Symbolics using Common Lisp with Flavors. The focus of this comparison is on: techniques, performance, influence of implementation strategies on performance, and identification of opportunities for optimization.

Envelope enhancement system for detecting paper machine press section anomalous vibration measurements

Abstract: A vibration detecting and condition monitoring system for the press section, including felts, rollers and presses, employing a feature extracting circuit having a selectable plurality of Bessel type bandpass filters and an enveloping circuit to square the bandpassed signal output prior to FFT analysis of the signal. The system enables the enhancement and detection of highly impulsive signals in the frequency domain, that are indicative of defects in machinery operation.

Comparing hidden Markov models with artificial neural network architectures for condition monitoring applications

Abstract: The use of hidden Markov models (HMM's), multilayer perceptrons (MLP's) and Kohonen self-organising maps (SOM's) has been proposed previously as efficient analysis and detection tools for condition monitoring of industrial plants and processes (Yin, 1993). Work on such applications with these techniques, has identified a need for a reassessment of these alternative recognition systems with a view to establishing their relative merits. In this paper the three systems are compared for two test data sets, one identifying the response of the systems to varying fault severity, the other showing recognition of faults which are independent of load. It is shown that for the MLP and SOM, implementing multiple networks improved the recognition of faults of varying severity. Possibly of more importance, this technique provided a means of diagnosing combinations of faults. For faults produced under differing load conditions, it is shown that the data cannot be classified by the SOM, and the supervised training regimes of the HMM and MLP provide the only means of classifying the data. Improved recognition is obtained with the HMM, although it is believed that the spectral pre-emphasis which was carried out on the input data could have contributed to this fact, suggesting that implementation of such pre-processing for the artificial neural network architectures, may be beneficial.

Acoustic evaluation of faults in electrical machines

Abstract: Condition monitoring of induction motors traditionally involves current measurements to detect electrical faults, and accelerometer measurements to determine mechanical faults. This work focuses on the use of acoustic measurements for condition monitoring, and proceeds from preliminary work done by Reynders, et al. (1992). The feasibility of identifying loose stator coils by acoustic measurements is shown, as is the identification and verification of a suitable neural network to perform this task automatically.

Fuzzy diagnosis and fuzzy navigation for plant inspection and diagnosis robot

Abstract: Plant inspection and diagnosis robot (IDR) must have the ability to widely monitor the condition of plant machinery with only a few sensors, and to quickly discriminate machine failures. Sound information can be used for monitoring the condition and diagnosing the failure of many machines at the same time, so a sound measuring system is suitable for an IDR. However, when using sound information for condition monitoring and inspection, several problems, such as the effect of noise and a dull sensibility to failure signals, etc., must be solved. In this paper, the method of failure detection and navigation for an IDR is discussed by using sound information and fuzzy logic. >

Application of Kohonen's self-organizing network to the diagnosis system for rotating machinery

Abstract: In this paper, we focus on the diagnostic algorithm of the diagnosis system for plant maintenance. In the current plant, the reduction of maintenance costs and the avoidance of sudden fault of equipment are required to improve the operation rate of the plant. The fault diagnosis is performed for this requirement, and the collection and analysis of actual fault data are indispensable to diagnose machinery conditions. For some reasons, however, there are many kinds of equipment whose fault data collection are difficult. In this paper, we select the motor in plant as such kind of equipment and propose a new diagnostic algorithm.

Condition-based maintenance of railway signalling equipment

Abstract: The scope and motivation of the work reported in this paper relates to the virtues and means of active condition-based maintenance of railway signalling equipment, train-stops and points machines in the first instance. An overview of evolving maintenance principles are introduced and their application to railways considered. This provides the motivation for the development and implementation of modern computer-based condition monitoring techniques. These provide early warning signals which could be used to increase the reliability and hence quality of the service to passengers, as well as to complement and direct the maintenance work of permanent way and signalling engineers in order to make efficient use of resources and save costs associated with too frequent maintenance routines.

Fault Diagnosis in the Linear Lightwave Network

Abstract: This paper proposes an algorithm for fault localization in a Linear Lightwave Network (LLN). A LLN is an all-optical network using Linear Dividers Combiners (LDCs) as network nodes. These LDCs can perform either wavelength or waveband selective linear operations on optical signals including controllable power dividing and combining [l]. In general, the network environment is divided into a number of static disjoint subnetworks, as in Figure l(a), each managed by an individual management process. A subnetwork might be connected to other subnetworks, and/or has inputs from Tx stations and outputs to Rx stations. A subnetwork manager is responsible for monitoring the power level at all the entering and leaving links in its domain, in order to detect failures [2] as shown in Figure l(b). Whenever a node fails the power levels in its output ports deviate from the acceptable power levels. We assume that if a node fails it fails completely (all its outputs are erroneous). The effect of a failure propagates in the nodes that are connected directly or indirectly to the faulty node. A straightforward way to localize a fault would be to constantly measure the input and output powers of all the components in the network. That approach, requires a large number of tests. Since testing is a time consuming process, it is essential to provide a fault localization algorithm which minimizes the number of required tests in order to localize the fault while keeping its computational complexity low. In this paper we propose such an algorithm which is based on measurements only at the peripheral nodes of the network. We examine the single fault case, where exactly one component can become faulty at any time. We also assume that the system is a zero time system '. For fault management purposes a subnetwork can be modeled as a digraph G = (V, E ) , where the vertices represent the network components that might fail and the edges the fault propagation between them. The manager keeps the set R (routing table) for all active routes & at any given moment in its domain. We define S(&) the set of nodes that compose the route R,, Head(%) and Tail(&) the starting and

Plasma process control with optical emission spectroscopy

Abstract: Plasma processes for etching and desmear of electronic components and printed wiring boards (PWB) are difficult to predict and control. Non-uniformity of most plasma processes and sensitivity to environmental changes make it difficult to maintain process stability from day to day. To assure plasma process performance, weight loss coupons or post-plasma destructive testing must be used. The problem with these techniques is that they are not real-time methods and do not allow for immediate diagnosis and process correction. These tests often require scrapping some fraction of a batch to insure the integrity of the rest. Since these tests verify a successful cycle with post-plasma diagnostics, poor test results often determine that a batch is substandard and the resulting parts unusable. These tests are a costly part of the overall fabrication cost. A more efficient method of testing would allow for constant monitoring of plasma conditions and process control. Process anomalies should be detected and corrected before the parts being treated are damaged. Real time monitoring would allow for instantaneous corrections. Multiple site monitoring would allow for process mapping within one system or simultaneous monitoring of multiple systems. Optical emission spectroscopy conducted external to the plasma apparatus would allow for this sort of multifunctional analysis without perturbing the glow discharge. In this paper, optical emission spectroscopy for non-intrusive, in situ process control is explored along with applications of this technique towards process control, failure analysis and endpoint determination.

Swept frequency response test for condition monitoring of power transformer

Abstract: The paper discusses the modelling and testing of a large 500 kVA distribution type power transformer with iron cored disk-type windings. The models developed show a close trend between the measured and predicted frequency response. The results are believed to be very encouraging for further work in this area. Furthermore, the method used in this paper is sufficiently general to be easily applied to other electromagnetic devices.

Early detection and continuous monitoring of dissolved key fault gases in transformers and shunt reactors

Abstract: Insulation breakdown in oil-filled equipment will produce gases. These gases will dissolve in the oil. The early detection and continuous monitoring of the key fault gases dissolved in oil provide a reliable warning of incipient and developing failure conditions. This paper reviews the basics and dynamics of failure evolution, the impact on the insulation system, the characterization of the defects using dissolved gas analysis, the use of a fuel cell based technology to detect and monitor fault gases in oil, as well as the economics of equipment condition monitoring. Reference is made to current industry practices and documented results.

A Data Acquisition System for Machine Tool Condition Monitoring

Abstract: A flexible, generically applicable and inexpensive data acquisition system (DAS), for machine tool condition monitoring, has been designed, constructed and installed as part of a European Union sponsored project. The DAS is more than just a data logger and an array of sensors. It also consists of a methodology for analysing data logging requirements and a relational database that supports this methodology. The database is held on a central ‘maintenance management’ computer. The monitoring to be carried out by the DAS is specified through this database, which contains detailed information about the DAS's facilities. This feature makes it simple to reconfigure the DAS to implement new monitoring requirements and to customize its operation to meet the needs of different machines. The information in the database is transformed into a look-up table that is read by the software that sets up and controls the data logging processes.

The PD pulse behaviour in rotating machine stator windings

Abstract: The acquisition, storage and processing of partial discharge (PD) signals is widely used as a condition monitoring tool to determine the integrity of the solid insulating systems of high power plants Despite the advances in the instrumentation available for this purpose, the ability of partial discharge data to provide the necessary information to enable accurate assessments to be made is compromised in rotating machines by the modification which partial discharge pulses undergo in their passage from their site(s) of origin to the point where detection instrumentation is connected Calibration strategies in general use with such instrumentation do not address this problem This paper presents results of an experimental investigation into pulse propagation in induction motor stator windings In particular, the paper focuses on the mechanisms of pulse propagation, ie series mode propagation through the windings and capacitive coupling and on the factors influencing their relative effects

The use of artificial neural networks for the estimation and classification of vibration behavior in power transformers

Abstract: It is widely recognized that the vibrational behavior of both the windings and the core of power transformers can be a good indicator of the transformer`s state, or health. However, measurements of vibration levels in large power transformers, are seldom, if ever, available in an on-line fashion. Additionally, the fitting of vibration transducers (accelerometers) to transformers already in service is both technically and economically infeasible. This paper reports on work undertaken using Artificial Neural Networks (ANNs) to both estimate and classify the vibrational behavior of a power transformer. The inputs to the ANNs consist of data which is typically available on-line such as voltage, current and temperature measurements. Extensive test data from an actual power transformer, fitted with the necessary instrumentation, has been used to train the ANNs and test their performance. The results of these tests are very encouraging, as will be demonstrated in the paper. Facilities for physically and electrically over-stressing the test transformer were used to generate the train/test data set, representing the transformer under a variety of normal and abnormal operating conditions. This work has generated significant interest among both utilities and manufacturers, and it is hoped that this project may culminate in the incorporation ofmore » this methodology in an integrated transformer condition monitoring system.« less

Reducing unscheduled plant maintenance delays-field test of a new method to predict electric motor failure

Abstract: Most electric motor predictive maintenance methods have drawbacks that limit their effectiveness in the mining environment. The US Bureau of Mines (USBM) is developing an alternative approach to detect winding insulation breakdown in advance of complete motor failure. In order to evaluate the analysis algorithms necessary for this approach, the USBM has designed and installed a system to monitor 120 electric motors in a coal preparation plant. The computer-based experimental system continuously gathers, stores, and analyzes electrical parameters for each motor. The results are then correlated to data from conventional motor maintenance methods and in-service failures to determine if the analysis algorithms can detect signs of insulation deterioration and impending failure. This paper explains the online testing approach used in this research, and describes monitoring system design and implementation. At this writing, data analysis is underway, but conclusive results are not yet available.

A gas turbine condition monitoring system

Abstract: This paper describes a program to develop a condition-based maintenance system for the Navy's LM2500 main propulsion engine. The effort began with research and evaluation of the engine failure modes and effects. Using a well-defined reliability-centered maintenance (RCM) methodology, we identified the most appropriate maintenance approach for each subsystem and failure mode. Expert rules were developed for the equipment and failures that will benefit from a condition-based maintenance approach. The expert system is based on the Navy's Integrated Condition Assessment System (ICAS). The proposed system applies new and innovative expert system logic and instrumentation using an upgraded capable ICAS to identify degradation in the marine gas turbine (MGT) engine and its subsystems.

A LabVIEW based virtual instrument for on-line induction motor parameters identification

Abstract: An automatic procedure to identify online the induction machine parameters of the steady-state equivalent circuit is presented. The procedure starts with the acquisition of input current and voltage instantaneous values and their processing to obtain the following dataset of variables: current and voltage first harmonic values; their displacement; and the slip value. The slip value is obtained through the current spectrum lines due to the rotor slotting. This first step is automatically iterated if the operating condition changes, in order to obtain more sets of variables. Different methods identify the parameters of the equivalent circuit, starting from the available sets of variables. All this process, configured as a virtual instrument, is implemented in the LabVIEW environment and can be included in a diagnostic system to detect machine health conditions, in order to expand system capabilities.