method of processing information loosely based on the manner m which neurons function in the human brain, has strong applicability to the job of interpreting the type of signals produced by NDE equipment. Neural networks have shown strong capabilities in pattern recognition tasks and are being utilized in a variety of applications ranging from object recognition to sonar signal classification. A system for NDE based on neural networks can make rapid, reliable evaluations of large areas, identifying potential trouble spots for human evaluation.
Chan, R.W.Y.; Hay, D.R.; Matthews, J.R.; MacDonald, H.A. Automated ultrasonic system for submarine pressure hull inspection
techniques, through the formulation of test procedures, to the analysis and interpretation of test results. In this particular case, an expert system shell was used to develop a system which assists in the derivation of procedures for the ultrasonic inspection of welds in ferritic steel. The components of the system are described and examples given of its use and subsequent outputs. The advantages of such an approach are improvements in the speed and accuracy of procedure writing, especially for less experienced personnel, the potential as a training aid and the ability to document the system consultation.
42015
Signal Processing and Pattern Recognition in Nondestructive Evaluation of Materials, Proceedings of the NATO Advanced Research Workshop, Lac Beauport, Quebec, Canada, 19-22 Aug. 1987, Edited by C.M. Chen, pp. 175-188. Springer-Verlag (1988) To meet the increasingly demanding requirements for ultrasonic testing of submarine pressure hulls, a project was undertaken to design an inspection system that incorporated artificial intelligence signal interpretation and data logging capabilities in a workstation configuration that could be used by field inspectors. The system is described and some preliminary results of the use of the artificial intelligence system for signal classification are presented.
Benoist, B.; Morizet, P.; Galliard, P.; David, B., Pigeon. M Increasing reliability of defect characterization of SG tubings using a combination of signal processing and expert system
41295
Proceedings of the 12th World Conference on Non-Destructive Testing, Amsterdam (Netherlands), 23-28 Apr. 1989, Vol. 2, pp. 1702-1707. Edited by J. Boogaard and G.M. van Dijk, Elsevier, 1989. The NDT laboratory Commissarait a L'Energie Atomique (CEA-SACLAY) and Universite de Technologic de Compiegne (UTC) are developing an expert system for automatic analysis of eddy current (EC) signals provided by the multifrequency control of steam generators tubing. This article describes the results of the eliminating pilgrim noise, and the expert system which uses signal analysis and signal processing in unison.
41903
Dornfi'ld, D.A. Monitoring of the machining process with acoustic emission sensors
41248
Journal of Acoustic Emission, Vol. 8, Nos. I-2, pp. 5227-5230 (Jan. - Jun. 1989). Special S u p p l e m e n t - Extended Summaries of papers to be presented at the World Meeting on Acoustic Emission, Charlotte, North Carolina, USA, 20-23 Mar. 1989 This paper discusses some of the motivations and requirements for sensing in automated or untended machining processes as well as reviews the research on acoustic emission (AE) sensing of tool condition (wear and fracture) in machining. The background for AE generation in metal cutting and its relationship to the condition of the cutting tool for single and multiple point tools (turning and milling) is presented. Research results are summarized relating to the sensitivity of AE signals to process changes. AE signal sensitivity to tool condition for wear and fracture, AE signal processing methodologies for feature extraction including time .series modeling to remove influences of machining conditions on wear tracking and AE ~nsor fusion using neural networks for process monitoring with several sensors.
NDT International, Vol. 22, No. 2, pp. 97-105 (Apr. 1989) Neural network algorithms offer a method of classification of multi- parameter data which is both rapid and tolerant of noisy data. Here the Hopfield model is used to classify processed ultrasonic data from various known classes of defect within steel test welds. Some 83 defects, from four known categories, and described by up to six feature parameters, were used in the analysis. A randomly chosen fraction of the set was used to produce images characteristic of each defect class. These were memorized into the network. Defects in the remaining fraction of the dataset were then recalled by the network, together with statistical information on the degree of confidence of the identification. An accuracy of 100% was often achieved when 50% fractions of the data were used for training and for testing. The accuracy achieved is comparable with that given by conventional minimum distance classification algorithms.
Gravec. I.; Sachse, W. Solving AE problems by a neural network 41885
Journal of Acoustic Emission, Vol. 8, Nos. 1-2, pp. 520-523 (Jan. - Jun. 1989). Special Supplement - Extended Summaries of papers to be presented at the World Meeting on Acoustic Emission, Charlotte, North Carolina, USA, 20-23 Mar. 1989 The contribution describes the application of a simulated neural network to characterization of Acoustic Emission (AE) phenomena. It is shown that a network. which works on an experimental series of AE data, is capable of approximately solving the forward and inverse wave problems on a completely empirical basis. As examples, the AE source characterization on ground of experimental signals and determination of Ae signals from source properties are presented in the article.
Harris, R.W.; Wood, B.R.A. Facilitation of AE integrity assessment by an expert system
41884
Journal of Acoustic Emission, Vol. 8, Nos. I-2, pp. 514-515 (Jan. - Jun. 1989). Special Supplement - Extended Summaries of papers to be presented at the World Meeting on Acoustic Emission, Charlotte, North Carolina, USA, 20-23 Mar. 1989 The assessment of the mechanical integrity of a structure using results from an acoustic emission evaluation involves a decision making .process which can he considered as a form of decision tree, and hence it should be possible to construct a knowledge base which can be incorporated into an Expert System shell. This will allow the externalization of the interpretive procedures into a computer based system and so that it will be possible to define the logical steps needed to make a valid assessment suggestions are presented for how this could he done.
Gravec, I.: Suchse, W. Application of an intelligent signal processing system to acoustic emission analysis
41750
Journal of the Acoustical Society of America, Vol. 85, No. 3, pp. 1226- 1235 (Mar. 1989) This article describes a novel approach for analyzing elastic wave signals to obtain a solution to simple inverse source problems. By using a simulated intelligent system that resembles the structure of a neural network and an acoustic source with its associated field of elastic waves in a plate, a ~ t of pattern vectors is generated. The memory of the system is formed through a learning process in which a systematic series of experiments is pre~nted to the system. In the experiments described here. the system was trained using simulated acoustic emission signals generated by a normal force acting on a thick plate to which information about the source related to its location and characteristics was appended, h is demonstrated, that one can recover the characteristics of an unknown .sourer from the acoustic signals emitted by it, and that one can synthesize from the memory the acoustic signals corresponding to an arbitrary source without using any elastodynamic theory.
McNah, A.; Young, H.S. Knowledge-based approach to the formulation of ultrasonic nondestructive testing procedures. 41656
lEE Proceedings, Vol. 136, Pt. A, No. 3, pp. 134-140 (May 1989) In nondestructive testing there is considerable potential for the application of knowledge-based systems. This ranges from the selection of suitable testing
46
Baker, A.R.; Windsor, C.G. The classification of defects from ultrasonic data using neural networks: the Hopfield method
Martinez-Ona, R.; Cabrera, E.; Gonzalez, L. Computerized system for automatic ultrasonic inspection
40975
Proceedings of the 12th World Conference on Non-Destructive Testing, Amsterdam (Netherlands) 23-28 Apr. 1989, Vol. I, pp, 183-188. Edited by J. Boogaard and G.M. van Dijk, Elsevier, 1989. The automation of different inspection tasks is an increasingly important, factor, especially with respect to the reduction of inspection time and the increase of reliability. This paper describes a system that integrates and automates the different inspection tasks. The evaluation module incorporates an expert system providing graphic and alphanumeric outputs of results, emulating the human evaluator.
Benas, J.C.; Lefevre. F.M. Automatic system for eddy current examination of steam generator tubes
40873
Proceedings of the 12th World Conference on Non-Destructive Testing, Amsterdam (Netherlands) 23-28 Apr. 1989, Vol. 1, pp. 854-859. Edited by J. Boogaard and G.M. van Dijk. Elsevier, 1989 The Expert System (ES) EXTRACSION CdF (systeme Expert de Traitement, d'Analyse et de Classification des Signaux d'Origine Nucleaire en Courants de Foucault) is described. The purpo~ of this system is to diagnose faults in Steam Generators (SG) pipes, which are tested by eddy currents. This ES drives algorithmical programs and describes the real world with an object oriented repre~ntation. This method is used for metal structures in SGs. The important problems of a priori knowledge (apK) and coherence between the different areas of knowledges are broached.
40465 Cabrera, E.; Gonzalez, L. Towards an ultrasonic automatic decision system Non-Destructive Testing, Proceedings of the 4th European Conference, London (United Kingdom), 13-17 Sup. 1987. Vol. 4, pp. 2525-2533. Edited by J.M. Farley and R.W. Nichols. Pergamon Press, 1988 This paper considers the general aspects of two projects currently being developed by TECNATOM, S.A. within the framework of pattern recognition technique~ The~ projects, which are aimed at increasing the reliability of Non-Destructive Testing (NDT) by ultrasonic methods, are: Automatic Classification by means of Pattern Recognition and Machine Learning Techniques, and an Expert System for Automatic Evaluation. The initial foundations for the automatic classifier are bar~:l on statistical and parametrical studies of ultra~nic signals. This paper presents certain characteristics and results of a signal analysis project, known as ANSE (ANalisis de SEnal), which has been designed to cover this objective.
Gondard, C.; Papezyk, E.; Wident, P.; Mauny, P.; Miglianico, T., Mougel, J.F.; Juvin, D. 'SERUS' An expert system for the ultrasonic examination of fuel r o d s
40457
Non-Destructive Testing, Proceedings of the 4th European Conference, London (United Kingdom), 13-17 Sup. 1987. Vol. 4, pp. 2454-2460. Edited by J.M. Farley and R.W. Nichols. Pergamon Press, 1988 The use of pattern recognition functions and the models of human expert reasoning, allow the automatic identification of defects in welds or structures. The proposed application uses an ultrasonic examination to detect and classify 3 types of defects in end plug weld.,;of PWR fuel rods.
NDT&E International Volume 25 Number 1 1992