Knowledge Based Assistant for Ultrasonic Testing Methodology of Materials
by Reynold
Franklin,* Udaya B. Halabe† and
B. Gopalakrishnan‡
This month's article presents an interesting concept for testing metals. The proposed expert system will be a useful tool for ultrasonic testing of metals. In this paper, the authors discuss the development of an artificial intelligence system for choosing a suitable ultrasonic testing procedure for a given application.
Ripi Singh
Guest Editor
Introduction
Ultrasonic testing (UT) is one of the most widely used NDT techniques for discontinuity detection and material characterization in various engineering fields. The advantages of UT as compared to other nondestructive techniques for metal tests are higher penetration, higher sensitivity and greater accuracy (Bar-Cohen et al., 1992). A major disadvantage of UT is that extensive technical expertise is required for its application. Extensive knowledge is required to identify the right testing methodology for a given application to obtain reliable results. Also, deciphering the received signals and correlating them to discontinuities or other indications requires the knowledge of an expert. Computer based tools representing the knowledge of an expert are ideal for industrial applications of UT, but they have to be user friendly and provide meaningful results. Several researchers have developed artificial intelligence (AI) systems to interpret the received ultrasonic signals and link them to various parameters of NDT characterization (Wendal and Dual, 1996; Sinha and Reddy, 1996; Rose, 1996; Roy et al., 1995; Shankar et al., 1991). However, very little effort has been spent on developing AI tools to assist an ultrasonic inspector in choosing a suitable testing method (transducer type and frequency, propagation mode and type of ultrasonic couplant).
The strength of an expert system lies in its ability to store, formalize and process qualitative data.
A complex relationship exists between the type of NDT characterization desired and the UT procedure. This relationship is further complicated by site conditions such as the presence of a coating on the metallic component and limited access for placing the transducers while the component is in use. Hence, extensive knowledge is required on the part of the ultrasonic inspector for choosing the suitable testing procedure. Therefore, an AI system that would help the ultrasonic inspector in choosing a suitable testing methodology will greatly enhance the quality of testing and reduce its cost while maintaining quality and reliability. Such an AI system could also greatly reduce the chances of generating ambiguous signals and hence facilitate the use of the other AI systems available for signal processing. The development and application of a rule based decision logic that can incorporate the expertise of humans for UT in metals are discussed in this paper.
Why Use an Expert System?
Ultrasonic testing procedures are developed by extensive research involving various factors. The success of each testing procedure is demonstrated by experimental work in the laboratory and/or field environment. Generally, the test procedure developed for a particular application may not be directly translated to other applications. Moreover, UT procedures for metals have been well established for several engineering applications. Hence, for choosing the appropriate test procedure, there is little scope for human reasoning but more for human knowledge. Expert systems can be used effectively for translating human knowledge into computer programs (Buchanan and Shortliffe, 1984). Expert systems have been successfully used in transforming human knowledge into problem solving tools by various researchers (Forouraghi, 1997; Gopalakrishnan et al., 1997; Gopalakrishnan et al., 1994; Sinha and Reddy, 1996; Anderson, 1992; Komai et al., 1992; Shankar et al., 1991; Srinivasan and Kane, 1990; Iudica, 1987; Fisher, 1985; Lacoe, 1984).Several qualitative factors such as metal type, surface condition and accessibility will influence the testing procedure. The strength of an expert system lies in its ability to store, formalize and process qualitative data. Moreover, the factors associated with the testing procedure can be formalized into sequential data which means decision trees and rules can be used to represent the underlying theory. This makes the use of an expert system a logical choice. Also, expert systems have a good user interface as an inherent part of the system which can be effectively used to explain to the user why and how a testing procedure works. Details related to decision making and recommendations can also be explained to the user. Hence, an expert system is an appropriate tool for developing the AI system on the UT methodology for metals.
Development of the Expert System
An expert system is defined as an intelligent system composed of a knowledge base, an inference engine, a working memory, a user interface and an explanation based subsystem. Expert systems have been developed to function as advisors in manufacturing, financial and other industrial applications. There are several established methods of knowledge representation such as the predicate calculus based approach, production rule based representation, frame based representation and object oriented representation. The expert system for UT presented in this paper makes use of rule based knowledge representation where the knowledge is presented as if/then rules. These rules exhibit the connection between conditions and conclusions using "and" and "or" connectives. The inference engine searches the knowledge base, one rule at a time, in order to arrive at the goal variables (Gopalakrishnan et al., 1997).Ultrasonic testing of metals constitutes four basic stages as shown in Figure 1. The testing specifications that are adopted would result in several wave propagation phenomena which are analyzed and quantified for the desired NDT characterization. Hence, a strong dependent relationship exists between the required NDT characterization and the test specifications to be adopted. This relationship is broken down into several hundred rules and incorporated into the expert system. A major concern involved in developing the expert system is to obtain the knowledge from the experts. Usually, an expert's knowledge is obtained through interviews and verbal communications which are extremely time consuming. Hence, inference rules were formulated based on the information available in several standard publications. Books by Krautkrämer and Krautkrämer (1990), Bray and Stanley (1989), Ensminger (1988), Szilard (1982) and Cartz (1995) provided very useful information. Standard reference handbooks by Bar-Cohen et al. (1992) and ASNT (1991) were also great sources of information. Facts collected from major journals in nondestructive testing such as Ultrasonics, Materials Evaluation, Journal of Nondestructive Evaluation and NDT & E International were also used. Rules were also generated from the information available in conference proceedings. It is important to note that the information available in the above references is based on extensive theoretical and experimental research by several investigators and represents a fairly comprehensive knowledge base.
Figure 1 — Stages of ultrasonic testing.
Typical examples of the rules developed based on the above literature are shown in Figure 2. As illustrated in Figure 2, based on the user inputs for application type, testing need, type of component and surface coating conditions, rule 1 makes interim conclusions about the type of ultrasonic waves to be used for testing. Based on this rule, if surface cracks are to be detected in a tube structure for mechanical applications (for example, in machines), the recommended waves are surface, critically refracted longitudinal and lamb waves. Higher confidence numbers are allotted to the wave type recommended the most in the literature. Rule 2 is a simple rule, which recommends the mode of propagation for the chosen wave type. Based on rule 1, if the user chooses surface waves, the propagation mode recommended as per rule 2 would be pulse echo or direct transmission.
Figure 2 — Example of rules in the expert system.
The expert system currently incorporates only the ultrasonic A-scan technique and contact transducer methods. VP-Expert, which is a rule based expert system development tool, was used to code the rules and develop the knowledge base. The early development of the expert system has been presented in Franklin and Halabe (1997). The rules were incorporated into several knowledge base files, which are chained together. The expert system makes use of the backward chaining process for the search because of their popularity and their method of operations. The inference engine searches the database for data to substantiate the goal conditions (ultrasonic test specifications). The expert system incorporates five common applications (nuclear, civil, mechanical, aerospace and pipeline) where UT is most likely to be conducted on metal components. The question and answer menu of the expert system has been generated in a graphics mode to provide the best user interface. Testing specifications for several NDT characterizations such as welds, surface cracks, subsurface cracks, internal discontinuities, elastic constants, corrosion and residual stresses have been incorporated in the expert system. Factors such as metal type, surface roughness, protective coating and access to opposite sides in the metal are also obtained from the user. These factors are required to decide the suitable type of ultrasonic waves for the given application. The expert system incorporates several complex geometries which are: thin plates, thick plates, welded connections, bolted connections, helical springs, shafts, tubes, turbine blades and boiler plates.
The expert system is intended to provide essential information to help the user in identifying a suitable testing procedure. Hence, the expert system is formatted in a manner to provide maximum flexibility for the user. The flow diagram of the knowledge base in the expert system is shown in Figure 3. The system will recommend four major parameters of the testing specifications based on the user input. The recommended parameters are:
- ultrasonic wave type
- propagation mode
- couplant type
- transducer frequency and characteristics.
The recommended parameters are shown in bold text in Figure 3. The user's input to the questions related to surface conditions, specimen geometry and objective of NDT characterization are used to determine the type of ultrasonic waves to be used in the testing. Depending upon the input, the system may recommend one or more possible ultrasonic waves that can be used for the testing. Whenever more than one type of wave can be used for testing, the system lists the waves in an order of preference. The user is asked to choose one wave type from the list recommended by the expert system. Based on the wave type selected by the user and the access condition (user input), the system recommends the possible propagation modes. Again, the expert system will list all possible propagation modes and the user selects one of them. The user selection from the list of expert system recommendations is shown in italics in Figure 3. Based on the selected propagation mode, metal type (user input) and temperature of testing (user input), the system recommends the transducer frequency and its characteristics such as resolution and bandwidth, which are affected by the damping level of the transducer. The transducer frequency is primarily controlled by the objective of the NDT characterization. The propagation mode primarily determines the characteristics of the transducer. The temperature of the metal and the objective of NDT characterization also helps in identifying the required transducer characteristics. The temperature and metal type are used to determine the properties of the ultrasonic couplant. Some important properties of a suitable couplant are displayed along with an example of a couplant type.
Figure 3 — Flow diagram of the structure of the expert system.
Displaying all possible ultrasonic waves and propagation modes that can be used for the given set of user inputs and requiring the user to select one wave type and one propagation mode from the displayed list provides flexibility in the system. Also, the system suggests the range of frequency and transducer characteristics rather than displaying a transducer model by a manufacturer. Moreover, the couplant characteristics are displayed to help the user. Thus, the expert system is formatted in a manner to provide maximum flexibility to the user. For a novice, selection of the primary wave and mode type will suffice for easy data analysis. In addition, the expert system provides necessary information on the properties of the various ultrasonic waves, propagation modes and acoustical properties of common metals in hypertext format. Relevant parts of several ASTM (1995) standards (shown in Table 1) are also incorporated in the expert system in hypertext format to help the user.
Table 1 ASTM standads incorporated into the expert system
Standard Title
E 114 Standard Practice for Ultrasonic Pulse-echo Straight-beam Examination by the Contact Method E 164 Standard Practice for Ultrasonic Contact Testing of Weldments E 213 Standard Practice for Ultrasonic Examination of Metal Pipe and Tubing E 273 Standard Practice for Ultrasonic Examination of Longitudinal Welded Pipe and Tubing E 494 Standard Practice for Measuring Ultrasonic Velocity in Materials E 587 Standard Practice for Ultrasonic Angle-beam Examination by the Contact Method E 797 Standard Practice for Measuring Thickness by Manual Ultrasonic Pulse-echo Contact Method E 1315 Standard Practice for Ultrasonic Examination of Steel with Convex Cylindrically Curved
Entry Surfaces
Consultation with the Expert System
The application of the expert system is demonstrated with an example of a consultation. The user input window is shown in Figure 4. When the user has clicked on the boxes, a pull down menu appears and the user chooses the appropriate answer to the questions. The input for testing a steel shaft for internal discontinuities under normal testing conditions is shown in Figure 4. Based on the user input, the system begins its recommendation of the testing procedure. The system's first recommendation is regarding the ultrasonic wave types which can be used for the test, as shown in Figure 5. Here, the system recommends the use of refracted shear or longitudinal waves with a preference for refracted shear waves. When the user is asked to choose one of the wave types, the user selects the refracted shear waves. The system then recommends the propagation modes as shown in Figure 6. Since the pitch catch technique is widely recommended for testing internal discon-
tinuities using refracted shear waves, it is recommended first and is followed by pulse echo, which is also used in a few cases. The user has the choice of selecting one of the modes and the choice is made as a pitch catch technique. Next, the system recommends the necessary characteristics of the transducer and ultrasonic couplant as shown in Figure 7. The transducer properties recommended by the system are high resolution and narrow band. The frequency range commonly used for the testing is from 2 to 10 MHz. A specific frequency is not provided to the user because the choice of frequency depends on several other factors such as the shaft diameter, wave attenuation in the metal specimen, pulser energy and noise level of the equipment. Hence, more information and specific details of the various equipment will be required to determine the specific frequency of the transducer. The properties of a desirable couplant such as being a corrosion inhibitor and being slow drying are also displayed, as shown in Figure 7.
Figure 4 — User input to the expert system for consulta
Figure 5 — System recommendation for the ultrasonic wave type.
Figure 6 — System recommendation for the propagation mode.
Figure 7 — System recommendation for the transducer and couplant prop
The user has the option of terminating the consultation or continuing in order to gain additional information. Necessary information on the various wave types, modes, ASTM (1995) standards and acoustical properties of common metals are presented in hypertext format. The main menu for the additional information is shown in Figure 8. The hypertext words shown in the highlighted format will lead the user into submenus or a text display screen. Choosing hypertext on the acoustical properties of metals will lead into the display screen shown in Figure 9. The table was extracted from the information provided in Bar-Cohen et al. (1992).
Figure 8 — Main menu for additional information on ultrasonic testing.
Figure 9 — Display of acoustical properties for various metals.
Conclusion
The development and application of an expert system on UT methodology for contact transducer methods with an A-scan technique are described in this paper. The expert system presented in this paper would be a useful tool for ultrasonic inspectors for choosing an appropriate testing methodology for testing metals. This tool, when used in conjunction with other AI tools for data analysis, could greatly reduce the need for highly skilled labor for UT. In the current expert system, decision rules were used and coded using a rule based development tool to provide flexibility and user interface. The expert system incorporates several common applications of UT, many NDT characterizations and most geometries. Rules for additional testing conditions and procedures could easily be added to the knowledge base on a requirement basis. The system recommends the ultrasonic wave types, propagation modes, transducer characteristics and the properties of a suitable ultrasonic couplant. Additional information such as details on the ultrasonic waves, propagation modes, relevant ASTM standards and necessary acoustical properties of common metals are also incorporated in the expert system in hypertext format for display to the user.
Acknowledgments
The authors wish to acknowledge the American Society for Testing and Materials for granting permission to display part of the relevant ASTM standards in hypertext format in the expert system.
References
Anderson, D.B., "Expert Systems and Material Property Databases," Computerization and Networking of Materials Databases, Vol. 3, ASTM, 1992, pp. 243-252.ASTM, Annual Book of ASTM Standards — Nondestructive Testing, Vol. 03.03, Philadelphia, the American Society for Testing and Materials, 1995.
Bar-Cohen, Y., A.K. Mal and the ASM Committee on Ultrasonic Inspection, "Ultrasonic Inspection," ASM Handbook on Nondestructive Evaluation and Quality Control, Vol. 17, 1992, pp. 231-277.
Birks, Albert S., Robert E. Green, Jr. and Paul McIntire, eds., Nondestructive Testing Handbook, second edition: Volume 7, Ultrasonic Testing, ASNT, Columbus, 1991.
Bray, D.E. and R.K. Stanley, Nondestructive Evaluation — A Tool in Design, Manufacturing and Service, New York, McGraw-Hill Book Company, 1989.
Buchanan, B.G. and E.H. Shortliffe, Rule-Based Expert Systems, Reading, Massachusetts, Addison-Wesley, 1984.
Cartz, L., Nondestructive Testing — Radiography, Ultrasonics, Liquid Penetrant, Magnetic Particle, Eddy Current, ASM International, 1995, pp. 81-125.
Ensminger, D., Ultrasonics — Fundamentals, Technology, Applications, 2nd edition, New York, Marcel Dekker, 1988.
Fisher, E.L., "Expert Systems Can Lay Groundwork for Intelligence CIM Decision Making," Industrial Engineering, 1985, pp. 56-60.
Forouraghi, B., "Inductive Learning of Materials Performance Indices from a Material Properties Database," Materials Evaluation, Vol. 55, No. 9, 1997, pp. 1007-1012.
Franklin, R. and U.B. Halabe, "Knowledge-Based Assistant for Ultrasonic Inspection in Metals," Proceedings of the SPIE Conference, Pittsburgh, 1997.
Gopalakrishnan, B., R.W. Plummer and S. Nagarajan, "ENERGEX: Expert Systems for Industrial Energy Conversion and Management," Energy Engineering Journal, AEE Publication, Vol. 94, No. 2, 1997, pp. 58-79.
Gopalakrishnan, B., R.W. Plummer, S. Nagarajan and R. Kolluri, "Energy Analysis and Diagnostics: A Computer Based Tool for Industrial Self Assessment," Proceedings of the Seventeenth National Energy Technology Conference, Houston, April 1994, pp. 70-75.
Iudica, N.R., "GUMMEX — An Expert System for Generating Process Plans," Nachrichten für Dokumentation, Germany, Vol. 36, No. 1, 1987, pp. 22-27.
Komai, K., K. Minoshima and M. Koyama, "Development of Diagnostic Expert Systems for Environmental Assisted Cracking (EXENAC) and Importance Evaluation of Rules in Inference," Computerization and Networking of Materials Database, Vol. 3, ASTM, 1992, pp. 255-270.
Krautkrämer, J. and H. Krautkrämer, Ultrasonic Testing of Materials, 4th Edition, New York, Springer-Verlag, 1990.
Lacoe, D., "Expert Systems: An Answer to Factory Management," Robotics World, Vol. 2, No. 11, 1984, pp. 30-33.
Rose, J.R., "Neural Nets for Flaw Classification Potential with Guided Waves," Proceedings of the 14th World Conference on NDT, New Delhi, Vol. 1, 1996, pp. 41-46.
Roy, A., P. Barat and S.K. De, "Material Classification through Neural Networks," Ultrasonics, Vol. 33, No. 3, 1995, pp. 175-180.
Shankar, R., R. Williams and M.J. Avioli, Jr., "Knowledge-Based Ultrasonic Examination Assistant," Materials Evaluation, Vol. 49, No. 10, 1991, pp. 1316-1321.
Sinha, V.K. and V.B. Reddy, "An Expert System for Defect Analysis," Proceedings of the 14th World Conference on NDT, New Delhi, Vol. 3, 1996, pp. 1821-1824.
Srinivasan, S. and R.D. Kane, "Expert System for Material Selection in Corrosive Environments," Materials Performance, Vol. 29, No. 10, 1990, pp. 69-73.
Szilard, J., Ultrasonic Testing — Non-conventional Testing Techniques, New York, John Wiley & Sons, 1982.
Wendel, R. and J. Dual, "Application of Neural Networks to Qualitative Nondestructive Evaluation," Ultrasonics, Vol. 34, 1996, pp. 461-465.
* Stueve Construction, 2201 East Oak Street, Algona, IA 50511.
† West Virginia University, Department of Civil and Environmental Engineering, PO Box 6103, Morgantown, WV 26506-6103; (304) 293-3031 X2617; fax (304) 293-7109; e-mail <uhalabe
@alum.mit.edu>.‡ West Virginia University, Department of Industrial and Management Systems Engineering, PO Box 6107, Morgantown, WV 26506-6107.
Copyright © 2001 by the American Society for Nondestructive Testing, Inc. All rights reserved.
Copyright © 2012 by the American Society for Nondestructive Testing, Inc. ASNT is not responsible for the authenticity or accuracy of information herein. Published opinions and statements do not necessarily reflect the opinion of ASNT. Products or services that are advertised or mentioned do not carry the endorsement or recommendation of ASNT.
IRRSP, NDT Handbook, The NDT Technician and www.asnt.org are trademarks of the American Society for Nondestructive Testing, Inc. ACCP, ASNT, Level III Study Guide, Materials Evaluation, Nondestructive Testing Handbook, Research in Nondestructive Evaluation and RNDE are registered trademarks of the American Society for Nondestructive Testing, Inc. ASNT exists to create a safer world by promoting the profession and technologies of nondestructive testing.