47719 Chapman,C.E.; Fahr, A.: Pelletier,A.; Hay, D.R. Artifldal intelligence in the eddy ourrent inspection of aircraft engine components Materials Evaluation, Vol. 49. No. 9, pp. 1090-1094 (Sep. 1991) pettem recognition and signal processing can be used together to build expert systems which can automatically detect and classify defects in real-time. This article de~ribes an application of such expert systemsto detecting and characterizing holthole cracks in aircraft compre~or disks by using eddy currant signals. Such automated eddy current techniques offer improved ~nsitivity reliability and speed. 47639 Shankar,R.; Williams, R.; Avioli, MJ. Knowledge-based ultrasonic examination assistant Materials Evaluation. Vol. 49, No. 10, pp. 1316-1327 (Oct. 1991) This paper describes an effort at the Electric POwerResearch Institute (EPRI) NDE Center to assist in the interpretation of NDE data acquired by automatic systemsduring ultrasonic weld examination of boiling water reactors (BWRs). A personal computer (PC) hased expert system "shell" was u~d to encode rules and assemble knowledge to addre,~s the di~rimination of intergranular stress-corrosion cracking (IGSCC) from benign reflectors in the inspection of pipe-to-compouent welds. The rules attempt to factor in plant inspection history, ultrasonic examination data. and. if available. radiography testing data; a majority of them deal with specific ultrasonic signal temporal and spatial behavior during automatic ~anning. The expert system is configured in a question-and-answer format. It has been integrated on a PC with a feature-biL~ed imaging system capable of displaying and computing image features pertinent to the consultation; an ultrasonic ray-tracing program simulates ultra.~nic examination scenarios. The system was evaluated on field-removed samples with ~rvice-induced IGSCC. 47589 Shankar, R.; Williams. R.; Smith, C.; Selby, G. A PC-based expert system for nondestructive testing Expert Systems Applications for the Electric Power Industry, Orlando, Florida (United States). 5-8 Jan. 1989, pp. 573-590. Edited by J.A. Nasar, Hemisphere (1990) This paper describes an effort underway at the EPRI NDE Center to assist in the interpretation of NDE data acquired by automatic systems during ultrasonic weld examination of boiling-water reactors (BWRs). A personal computer (PC)--based expert system "shell" was u~d to encode rules and assemble knowledge to eddre.~sthe discrimination of intergranular stress corrosion cracking (IGSCC) from benign reflectors in the inspection of pipe-to-component welds. The system was evaluated at the EPRI NDE Center on field-removed samples with service-induced IGSCC and is currently being evaluated by utiliti~. 47585 Langdon,A.; Abroad, K.; Frieze, P.A. An expert system for offshore structure inspection and maintenance 4th International Conference on Civil and Structural Engineering Computing (CiviI-Comp 89), pp. 221-232, Edited by B.H.V. Topping, Vol. 4 (1989) ISBN 094874913X The full extent of knowledge required to completely rationally execute inspection and maintenance of offshore structures is beyond the scope of one individual without the assistance of a comprehensive system with access to the necessary datebases and algorithms. The PLAIM (Platform Lifetime Assessment through Analysis, Inspection and Maintenance) project has been established to provide such a system. 'Knowledge- Based' Expert Systems methods and techniques, particularly those related to the acquisition of experimental knowledge and the repre~ntation of the knowledge for solving problems, have been u~d in the PLAIM project. The major deliverabies of this project include: the documentation of informal and qualitative knowledge together with a prototype expert system for eliciting details of defects in structures, empirically as~ssing the severity of the defect and recommending a series of remedial actions. 47516 Rodgers,J.M. New systems concepts for on-line digital evaluation of acoustic emission signals Progress in Acoustic Emission V. The ]Oth International Acoustic Emission Symposium, Sendal (Japan), 22-25 Oct. 1990. pp. 369-379. Edited by K. Yamaguchi, H. Takahashi and H. Niitsuma. The Japanese Society for Non-Destructive Inspection (1990) The next generation AE systems will be characterized by greater signal processing flexibility, and may lose their distinction as uniquely designed for AE signal analysis. The growing popularity of adaptive training strategies for signal characterization, such as statistical pattern recognition classifiers and neural networks, is also prompting the need for systems with signal pruce~sing flexibility. The ideal system would allow the selection and training of a signal classification approach for a given application, then use software routine to optimize the system for use in the run-time application. This paper will describe the fundamental design of the new generation system required to meet tbe~ objectives, and its implementation into a practical instrument for both re,arch and field use. 47413 Grabec, I.; Sachse, W. Automatic mndeling of ultrasonic phenomena Progress in Acoustic Emission V. The 10th International Acoustic Emission Symposium, Sendai (Japan), 22-25 OCt. 1990. pp. 196-202. Edited by K. Yamaguchi, H. Takahasi and H. Niitsuma. The Japanese Society of Non-Destructive Inspection (1990). This article deals with modeling of ultrasonic phenomena as information processing which can be automatically performed by a system composed like a neural network out of an array of sen~rs, a memory, an estimator and an array of actuators. The focus is on explanation of those properties of the system which enable prediction of data from incomplete measurements. Performance of the system is demonstrated on examples of AE and ullrMonic scattering phenomena. It is shown that forward and inverse AE problems can be treated, and also the material inhomogoneities can be characterized by the same modeler simulated on a data acquisition system. 47031 Ishizuka, M. The report of Ad-hoc committee for nondestructive evaluation expert systems in 1990 (In Japanese) Journal of JSNDI, Vol. 40, No. 6, pp. 375-377 (Jan. 1991) 46941 Workman, G.L.; Bryson, C.C. Eddy current inspection of graphite fiber components National Aeronautics and Space Administration, Huntsville, Alabama (United States), N91-10294/7/GAR, 52 pp. (Oct. 1990) This project attempts to utilize current capabilities in eddy current instmmontation. artificial intelligence, and robotics in order m provide insight into defining geometrical aspects of being evaluated using eddy current inspection techniques. The unique capabilities of E-probes and horseshoe probes for inspecting graphite fiber materials were evaluated and appear to hold great promise once the technology development matures. The initial results are described of modeling eddy current interactions with certain flaws in graphite fiber samples. 46868 McNab,A.; Dunlop, i. Artificial intelligence techniques for the automated analysis of ul- trasonic NDT data Advances in Transducers. Equipment and Data processing to Improve the Reliability of NDT, Colloquium, 4 Dec. 1990. pp. 6/1-6/8 lEE Digest No. i 9901172. Artificial intelligence (A.I.) and other automated computer-hnseddata processing techniques offer advantagesof improved speed and accuracy in the interpretation of nondestructive testing results over the usual manual interpretation. This paper discusses the automated analysis of ultra.,~nic data, from weld inspections, for example, and includes image processing and A.I. techniques such as pattern recognition and neural networks. Several defect shapes am selected for analysis. Future work on A.I. would extend the number of defect types and use real defects. A three-layer neural network is aim proposed,and an expert system framework would be used to relate a defect region to a number of external parameters. 46850 Win6Lvor, C.G. The classification of defects from ultrasonic measurements Conference on Neural Networks from Models to Applications (NEUR088), Paris (France), 6-9 Jun. 1988, pp. 592-601. Edited by L. Personnaz and G. Dreyfus. (1990) ISBN 2903667039 The Rumelhart back propagation and Hopfield algorithms have been used to classify the nature of the defects found in steel pressure ve,ssels. Ultrasonic data from ~ m e 83 defects of known type had previously been processed into six feature parameters describing the ultrasonic reflections. Both neural network methods performed as well as a weighted minimum distance algorithms in classifying the defect type. 46845 Javed, M.A.; Sanders, S.A.C. Artificial neural networks as Intelligent condition monitoring devices Condition Monitoring and Diagnostic Technology, Vol. 2, No. 1. pp. i 7- 20 (Jul. 1991) This report describes the recent advances made in artificial neural network technology and its potential applications in the rapidly growing field of condition monitoring and diagnostic engineering. The architecture of a multi-layer perceptmn type neural network and the fundamentals of the procedure required to train these devices is described. Some useful recent applications of this technique in the field of condition monitoring and diagnostic engineering systemsare aim discussed. 46841 Bahr,B.; Naheel, T.M. Neural networks for detecting defects in aircraft structures Wichita State Univ., KS. Inst. for Aviation Research, Kansas (United States), PB91-118760/GAR, 16 pp. (Apr. 1990). The problem of the reliability encountered in nondestructive testing(NDT)/nondestructive evaluation (NDE) was examined in the paper, and the necessity to automate the process was confirmed. Since automatic interpretation of the results is critical for the success of such a system, neural networks with their unique abilities in the field of complex pattem recognition were examined with some very simplified mathematical explanation of how they perform, and how they may be applied to NDE problems. The Neum-NDE can be an aid to the inspector to locate problem areas with greater confidence, thereby increasing productivity and reliability. It is worthwhile to mention that two cases mentioned as applications to Neuro-NDE are not the only possible cases to use neural networks but rather they indicate that a lot can, and should be done. 46832 Bahr, B.; Maari, S. Robotic-aided system for Inspection of aging aircraft National Inst. for Aviation Research, Wichita, Kansas (United States), PB91-I 18802/GAR, 15 pp. (May 1990) The objective of the work is to illustrate how a robotic system can be used for inspection of aircraft. A new mobile robot concept for inspection of aircraft will be presented. The approach will be especially useful for areas that are on the aircraft fuselage and not within eaay watch.The robot can be prognmm~ to follow a specified path while inspecting the stntcture. Vision and other inspection systems will be used to record the surface condition of the aircraft. The results of inspection regardless of technique (vision, eddy current, etc.) can he transmitted to the technician or an expert NDT&E International Volume 25 Number 1 1992 41