Data processing: measuring – calibrating – or testing – Measurement system in a specific environment – Mechanical measurement system
Reexamination Certificate
2000-06-05
2003-02-11
Hoff, Marc S. (Department: 2857)
Data processing: measuring, calibrating, or testing
Measurement system in a specific environment
Mechanical measurement system
C702S038000, C702S039000, C073S592000, C706S021000
Reexamination Certificate
active
06519535
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to a method for predicting burst pressures in tubing. More particularly this invention relates to a method for predicting burst pressures in thin-walled tubing by deconvolving eddy current data to identify electrical characteristics of tubing defects and analyzing the deconvolved data using artificial neural network modeling to predict burst pressures.
BACKGROUND OF THE INVENTION
Steam generators have a long history of being a troublesome major component in pressurized water reactor nuclear power plants, and steam generator tube failures remain a costly concern for nuclear utilities. The thousands of thin-walled tubes of a steam generator form a containment boundary between the high-pressure primary and low-pressure secondary water systems. Therefore, plant operators must repair or plug tubes if significant cracking is detected. Tube cracking has directly led to the decommissioning and replacement of many U.S. steam generators. Eddy current (EC) techniques are currently the predominant technology used for the in-service inspection of nuclear power plant steam generator tubing in the U.S. and elsewhere. Current EC measurement techniques using differential bobbin-coil probes are extremely sensitive to the presence of axial cracks in the tube wall, but they are equally sensitive to the presence of tubing artifacts, i.e. tube dents, fretting, support structures, and corrosion products. EC signal interpretation is further complicated by cracking geometries more complex than a single axial crack. Since cracks are difficult to distinguish from artifacts and even more difficult to characterize, operators are often forced to repair or plug a tube upon detection of a defect, regardless of its effect on tube integrity.
Current analysis techniques based on EC bobbin-probe measurements have been relatively successful in detecting cracks, but fail almost completely when applied to the characterization of cracks and their effect on tube integrity. For example, EC inspection software is available to linearly mix signals taken at different frequencies. These mixing algorithms can accentuate signals from defects, aiding defect detection, but signal distortions due to mixing make this technique less useful for defect sizing. Since they respond only to an aggregate disruption of electrical current along the circumference of a tube at a given axial position, bobbin coils cannot differentiate multiple defects along the tube circumference. Bobbin probes, therefore, are ineffective for characterizing complex cracking. Some limited success has been demonstrated using rotating pancake coils to estimate crack depth. This invention emphasizes the use of differential bobbin coils, but is not limited to that technique.
Additionally, EC signals from cracks, wastage, and other physical sources have similar, if not identical, frequency responses. Frequency-based signal processing techniques such as Fourier filtering and wavelet transformations are ineffective at winnowing out artifact signals, since their frequency signatures are indistinguishable from those of crack signals.
Likewise, algorithms based on EC expert rules applied to the identification of defects in steam generator and heat exchanger tubing have met with only limited success. Attempts to use expert systems to estimate the depth of steam generator tubing flaws in tube support plate regions based on EC voltages and phase angles have met with even less success because complex defect morphologies have limited the accuracy of those algorithms to only ±20 % of the tube wall thickness. Thus, current EC-based expert rules have resulted in poor decisions about whether to plug suspect steam generator tubes.
Artificial intelligence methods have held the promise of providing improved modeling of tube defects compared to conventional empirical modeling techniques. Beginning in the late 1970s, pattern-recognition algorithms have been used to determine which EC Lissajous-pattern features best correlated with the defect classification of tubes with electrodischarge-machined (EDM) slots, machined elliptical wastage, and uniform thinning. These parametric studies considered many signal features based on the shape of the Lissajous figures and ones based on EC voltage readings. Along with classifying the simulated flaws, the researchers attempted to predict the depth of the uniform thinning based on a least-squares regression of EC features. However, these attempts to predict the size of the axial slots were largely unsuccessful.
More recently, an artificial intelligence technique of case-based reasoning has been applied to the classification and characterization of flaws detected through EC inspection and other nondestructive examination methods. Case-based reasoning relies on a comparison of input features (e.g. as from an eddy current measurement) to values used previously for training the system. Cases with similar input features would be expected to have similar solutions. Although case-based reasoning has potential advantages over other artificial intelligence techniques, it still has limitations, including the requirement of a large data base to cover the range of possible input-feature combinations, especially for problems that depend on several input variables. Additionally, case-based reasoning is particularly vulnerable to the effect of data noise. Input data distorted by the presence of artifacts can baffle the analysis system's attempt to find a matching comparison case in the data base.
Artificial neural networks (ANNs) have held particular promise as an artificial intelligence tool for modeling steam generator tube integrity. ANN techniques have been applied to the eddy current identification of flaws in flat plates, and similar neural network defect-identification studies using tubes with machined flaws have been performed. Researchers have used ANNs to characterize the defect depth and artifact type for tubes with drilled holes and artifacts such as tube supports, copper, and magnetite. Employing this same drilled-hole data, others have applied ANNs to eddy current signal analysis for defect classification and for defect sizing. In order to separate EC crack signals from those due to the artifacts, a reference signal was subtracted from the test signal, where the reference signal was obtained from a tube without holes. Accordingly, the common features of the two EC measurements were removed, and the hole effects in the test signal were enhanced. The earlier research was later extended to estimate the depth of laboratory-generated outer-diameter stress corrosion cracks in simulated steam generator tubes. Despite these limited successes in classifying and sizing defects, this research has not translated into a predictive capability that allows modelers to accurately assess the integrity of a damaged tube.
More recently, researchers have proposed a hybrid system that combines rule-based logic, fuzzy syntactic pattern recognition techniques, and artificial neural networks for the detection and basic classification of flaws from eddy current signals. Similarly, others are developing a hybrid eddy current diagnostic system. However, these systems have yet to address the issue of distinguishing crack signals from other signal sources, nor do they attempt to quantify tube burst pressure, which is a more accurate predictor of tubing integrity.
It is therefore an object of the present invention to provide an improved method for assessing the physical integrity of a tube having a defect.
It is another object of the present invention to provide a novel method for predicting the burst pressure of a tube.
It is yet another object of the present invention to provide an improved method for more accurately assessing the integrity of a damaged tube.
It is a further object of the invention to provide a novel method for predicting the burst pressure of a tube having a critical tubing defect from inspection data of the critical tubing defect.
Other objects and advantages of the invention will become app
Kupperman David S.
Morman James A.
Petri Mark C.
Reifman Jaques
Wei Thomas Y. C.
Cooley Marc
Hoff Marc S.
Rechtin Michael D.
The University of Chicago
LandOfFree
Eddy current technique for predicting burst pressure does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Eddy current technique for predicting burst pressure, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Eddy current technique for predicting burst pressure will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3143175