SCANNER |STANDARDSUPDATE STANDARDIZATION IN ARTIFICIAL INTELLIGENCE: NEEDS, POSSIBILITIES, AND CHALLENGES Artificial intelligence (AI) and machine learning (ML) techniques and their appli- cations are flourishing in a variety of areas including health and medicine, engineering, manufacturing, and nonde- structive testing (NDT). Many industries, companies, and government agencies invest enormously in research on AI for example, €20 billion (US$21.5 billion) per year to the end of 2020 was invested by the European Union alone on AI research and development1. The US National Science Foundation announced US$140 million in funding in May 2023 to launch seven new National AI Research Institutes. This investment will bring the total number of institutes to 25 across the country and extend the network of orga- nizations involved into nearly every state2. Standardization of a technique or a method is the key step toward gener- alization of its application. In addition, standardization has a significant posi- tive impact on technology transfer and emerging technologies by forming common vocabularies and agreed defini- tions of terms. However, in the case of AI, and specif- ically in the field of NDT, the question is that if AI research has already produced mature technologies, and if AI-NDT is ready for standardization. There have been numerous articles published in AI for the NDT domain in the last few years, but practical assessment of the proposed AI methods is limited due to the lack of standardized practices that can be used to validate the performance of the developed tools. From a scientific point of view, there are many open research questions that make AI standardization appear to be premature. As an example, many existing standards in the field of inspection and safety, such as ISO 26262 on functional safety of road vehicles, are not compatible with typical AI methods despite the increasing efforts and interest in advancing technology in passenger cars and autonomous vehicles3. Currently, many standards develop- ment organizations worldwide work on norms for AI technologies and AI-related processes. The International Organization for Standardization (ISO) has run a stan- dardization project on AI since 2018. ISO, in collaboration with International Electrotechnical Commission (IEC), founded the subcommittee ISO/IEC JTC 1/SC 42 to work on an AI standard- ization project4. The scope of work of subcommittee 42 is standardization in the area of AI and consists of five working groups (WGs) and a joint working group with subcommittee 40 (IT Service Management and IT Governance). The WGs include foundational standards (WG 1) big data (WG 2), which used to be covered by a separate working group under JTC 1 trustworthiness (WG 3) use cases and applications (WG 4) and computational approaches and compu- tational characteristics of AI systems (WG 5). Societal concerns have become a subtopic of WG 3. A brief description of each WG follows: Ñ WG 1 attempts to find a workable definition by consensus. Although the concrete wording of the AI defi- nition may not be highly crucial for the quality of the future SC 42 stan- dards, there is a definite need for an AI definition in industry. Ñ WG 2 is assigned to work on big data. Ñ WG 3 works on trustworthiness, including the main tasks of (a) inves- tigating approaches to establish trust in AI systems through transpar- ency, verifiability, explainability, and controllability (b) investigating engi- neering pitfalls and assess typical associated threats and risks to AI systems with their mitigation tech- niques and methods and (c) inves- tigating approaches to achieve AI systems’ robustness, resiliency, reli- ability, accuracy, safety, security, and privacy. Ñ WG 4 works on use cases and applications with the main tasks of (a) identifying different AI application domains and the different context of their use (b) describing applications and use cases using the terminology and concepts defined in ISO/IEC 22989 and ISO/IEC 23053 and extending the terms as necessary and (c) collecting and identifying societal concerns related to the collected use cases. Ñ WG 5 works on computational approaches and computational characteristics of AI systems including (a) main computational characteristics of AI systems and (b) main algorithms and approaches used in AI systems. More updates on the progress of AI standardization will be discussed in future articles as they become available. STANDARDS EDITOR Hossein Taheri, PhD: Georgia Southern University, Statesboro, GA htaheri@georgiasouthern.edu REFERENCES 1 Duthon, P., F. Bernardin, F. Chausse, and M. Colomb. 2018. “Benchmark for the robustness of image features in rainy conditions.” Machine Vision and Applications 29 (5): 915–27. https://doi. org/10.1007/s00138-018-0945-8. 2 The White House. 2023. “Fact sheet: Biden-Harris administration announces new actions to promote responsible AI innovation that protects Amer- icans’ rights and safety.” https:// www.whitehouse.gov/briefing-room/ statements-releases/2023/05/04/ fact-sheet-biden-harris-administration -announces-new-actions-to-promote -responsible-ai-innovation-that-protects -americans-rights-and-safety/. 3 Rao, V. R., 2018. “How data becomes knowledge, part 1: from data to knowledge.” IBM Corp. 4 Zielke, T., 2020. “Is artificial intelligence ready for standardization?” EuroSPI 2020: Systems, Software and Services Process Improvement: 259–274. https://doi. org/10.1007/978-3-030-56441-4_19. 22 M A T E R I A L S E V A L U A T I O N • J U L Y 2 0 2 3 2307 ME July dup.indd 22 6/19/23 3:41 PM

IN DEVELOPMENT The following section provides a summary of new standards, drafts, and revisions that may be of interest to the nondestructive testing and evaluation (NDT/E) community. This summary is provided in Materials Evaluation on a quarterly basis in January, April, July, and October. For the latest informa- tion, please visit the website of the standards provider. PROJECT INITIATION ANSI procedures require notification by ANSI- accredited standards developers of the initiation and scope of activities expected to result in new or revised American National Standards. The following is a list of proposed actions and new standards that have been received recently from accredited standards developers. To view information about additional standards for which a project initiation notification has been submitted, and to search approved American National Standards, please visit ansi.org, which is a database of standards information. Note that this database is not exhaustive. Ñ BSR/AWS C6.3M/C6.3-202x, Recommended Practices for Friction Stir Welding. This is a revision of ANSI/AWS C6.3M/C6.3-2023. This standard provides recommended practices intended to be applicable to all industries for friction stir welding and processing of aluminum and other material and material alloys and addresses design considerations, fabrication, and quality assurance. Ñ BSR/AWS C3.6M/C3.6-202x, Specification for Furnace Brazing. This is a revision of ANSI/AWS C3.6M/C3.6-2022 AMD2. This specification provides the minimum fabrication, equipment, material, process, and procedure requirements, as well as inspection requirements for the furnace brazing of steels, copper, copper alloys, and heat- and corrosion-resistant alloys and other materials that can be adequately furnace brazed (the furnace brazing of aluminum alloys is addressed in AWS C3.7M/C3.7, Specification for Aluminum Brazing). This specification provides criteria for classifying furnace brazed joints based on loading and the consequences of failure and quality assurance criteria defining the limits of acceptability in each class. This specification defines acceptable furnace brazing equipment, materials, and procedures, as well as the required inspection for each class of joint. Ñ BSR/AWS C3.7M/C3.7-202x, Specification for Aluminum Brazing. This is a revision of ANSI/AWS C3.7M/C3.7-2011 (R2021). This specification presents the minimum fabrication, equipment, material, process procedure, and inspection requirements for the brazing of aluminum by all of the processes commonly used—atmosphere furnace, vacuum furnace, and flux processes. Its purpose is to standardize aluminum brazing requirements for all appli- cations in which brazed aluminum joints of assured quality are required. It provides criteria for classifying aluminum brazed joints based on loading and the consequences of failure and quality assurance criteria defining the limits of acceptability of each class. Ñ BSR/AWS D14.9/D14.9M-202x, Specification for the Welding of Hydraulic Cylinders. This is a revision of ANSI/AWS D14.9/D14.9M-2022. This specification provides standards for the design and manufacture of pressure containing welded joints and structural welded joints used in the manufacture of hydraulic cylinders. Manufacturer’s responsibilities are presented as they relate to the welding practices that have been proven successful within the industry in the production of hydraulic cylinders. Included are clauses defining procedure qualification, performance qualification, workmanship, and quality requirements, as well as inspection requirements and repair requirements. Ñ BSR/API 579-1/ASME FFS-1-202x, Fitness-For- Service. This is a revision of ANSI/API 579-1/ ASME FFS-1-2021. This standard provides guidance for conducting FFS assessments using methodologies specifically prepared for pressurized equipment. The fitness-for-service guidelines provided in this standard can be used to make run-repair-replace decisions to help determine if components in pressurized equipment containing flaws that have been identified by inspection can continue to operate safely for some period of time. Ñ BSR/ASME B31P-202x, Standard Heat Treatments for Fabrication Processes. This is a revision of ANSI/ASME B31P-2017. This standard provides requirements for heat treatment of piping assemblies that meet the requirements of ASME B31 code sections. These require- ments apply to (a) preheating, (b) postweld heat treatment (PWHT), (c) postforming heat treatment (PFHT) required by the ASME B31 code sections for other fabricated assemblies including forming operations such as bending, and (d) heat treatments required by contract documents. Ñ BSR/AWS D10.4M/D10.4-202x, Guide for Welding Austenitic Stainless Steel Piping and Tubing. This is a revision of ANSI/AWS D10.4M/ D10.4-2023. This document presents a detailed discussion of the metallurgical characteristics and weldability of many grades of austenitic stainless steel used in piping and tubing. The delta ferrite content as expressed by ferrite number is explained, and its importance in minimizing hot cracking is discussed. Several figures and tables illustrate recommended joint designs and procedures. Annex A presents infor- mation on the welding of high-carbon stainless steel cast pipe and fittings. CALL FOR COMMENT ON PROPOSALS LISTED The public comment period has passed for the following draft American National Standards, which are currently in review. Ñ BSR/AWS D14.0/D14.0M-202x, Machinery and Equipment Welding Specification. This is a revision, redesignation, and consolidation of ANSI/AWS D14.3/D14.3M-2018, ANSI/ AWS D14.4/D14.4M-2019, ANSI/AWS D14.5/ D14.5M-2009, and AWS D14.1/D14.1M. This specification establishes design, manufacture, quality, inspection, and repair requirements for carbon, low-alloy, and alloy steel welded connections in machinery and equipment. It addresses topics including weld joint design, workmanship, quality acceptance criteria, nondestructive examination methods (visual, radiographic, ultrasonic, magnetic particle, and liquid penetrant), repair of weld defects, and post-weld heat treatment. J U L Y 2 0 2 3 • M A T E R I A L S E V A L U A T I O N 23 2307 ME July dup.indd 23 6/19/23 3:41 PM