工业物联网人工智能框架白皮书（EN）.pdf

Industrial IoT Artificial Intelligence Framework 2022-02-22 Authors Wael William Diab,Alex Ferraro,Brad Klenz,Shi-Wan Lin,Edy Liongosari,Wadih Elie Tannous,Bassam Zarkout.Industrial IoT Artificial Intelligence Framework IIC:PUB:IIAIF:V0.10:ID:2022 ii CONTENTS 1 Industrial Artificial Intelligence.6 1.1 Industrial Internet of Things.6 1.2 Industrial Artificial Intelligence.6 1.3 Architecture Viewpoints.7 2 Business Viewpoint.8 2.1 Uncover Valuable Insights From Data Intensive Environments.9 2.2 Enable Digital Transformation.9 2.3 Agent for Future-Proofing the Organization.11 2.4 AI Adoption Readiness.11 3 Usage Viewpoint.12 3.1 Industrial AI Market.12 3.2 Usage Considerations.13 3.3 Trustworthiness.15 3.3.1 Security.16 3.3.2 Privacy.19 3.3.3 Confidentiality.20 3.3.4 Explainability.21 3.3.5 Controllability.21 3.4 Ethical and Societal Concerns.22 3.4.1 Ethics.22 3.4.2 Bias.23 3.4.3 Safety.24 3.5 Impact on Labor Force.25 3.6 Regional and Industry-Specific Considerations.27 3.7 AI as a Force for Good.27 4 Functional Viewpoint.28 4.1 Architecture Objectives and Constraints.28 4.2 Data Concerns.29 4.3 Learning Techniques.30 4.4 General Industrial AI Functional Architecture.32 4.5 System of Systems Issues.34 4.6 Application Horizon of Industrial AI.35 5 Implementation Viewpoint.36 5.1 Implementation Guidance.36 5.2 Implementation Considerations.37 5.2.1 Scope.37 5.2.2 Response Time.37 5.2.3 Reliability.38 5.2.4 Bandwidth and Latency.38 5.2.5 Capacity.39 5.2.6 Security.39 5.2.7 Data Properties.39 Industrial IoT Artificial Intelligence Framework IIC:PUB:IIAIF:V0.10:ID:2022 iii 5.2.8 Temporal Data Correlation.40 5.2.9 Interoperability.40 5.2.10 Running Systems In Parallel.40 5.2.11 Dealing With Technical Debt.41 5.2.12 Portability and Reusability of AI Systems.41 6 The Future of the Industrial AI.42 6.1 Far-Reaching Benefits of AI Despite the Risks.42 6.2 Convergence with Other Transformative Technologies.42 6.3 Standards Ecosystem.44 6.3.1 Enabling intelligent insights.45 6.3.2 Ecosystem approach.46 6.3.3 Program of work and role in enabling DX across industries.46 6.3.4 Summary.48 6.4 Final Thoughts and Takeaways.48 Annex A Artificial Intelligence Background.50 A.1 Brief History of AI.50 A.2 Why Now?.51 A.2.1 Cheap and Powerful Compute Infrastructure.51 A.2.2 Availability of Large Amounts of Data.52 A.2.3 Improvements in Algorithms.52 Annex B Exemplary Use Cases of AI in Industry.52 B.1 Manufacturing.53 B.2 Healthcare.55 B.3 Buildings.56 B.4 Transportation and Logistics.57 B.5 Detecting IIoT System Threats Using AI.57 Annex C IIC References.58 Annex D Authors&Legal Notice.59 FIGURES Figure 1-1.Industrial Internet Viewpoints.Source:IIC IIRA.7 Figure 2-1.Industrial AI Framework Business Viewpoint.Source:IIC.8 Figure 2-2.Digital Transformation Journey.Source:IIC.10 Figure 3-1.Industrial AI Framework Usage Viewpoint and Its Stakeholders.Source:IIC.12 Figure 3-2.Trustworthiness of IIoT Systems.Source:IIC.15 Figure 3-3.Security Framework Functional Building Blocks.Source:IIC IISF.17 Figure 3-4.Skill shift:Automation and the Future of the Workforce.Source:McKinsey.26 Figure 4-1.Industrial AI Framework Functional Viewpoint and Its Stakeholders.Source:IIC.28 Industrial IoT Artificial Intelligence Framework IIC:PUB:IIAIF:V0.10:ID:2022 iv Figure 4-2.AI/Machine Learning Model Process.29 Figure 4-3.Data Processing for AI Modeling.30 Figure 4-4.Industrial AI System.31 Figure 4-5.Industrial AI in Industrial Operation Environment.32 Figure 4-6.Industrial AI High-Level Functional Components.33 Figure 4-7.Example of a System of Systems in the EV Charging Space.Source:Artemis.34 Figure 5-1.Industrial AI Framework Functional Viewpoint and Its Stakeholders.Source:IIC.36 Figure 6-1.Trustworthiness of IIoT Systems.Source:ISO/IEC JTC 1/SC42.46 TABLES Table 3-1.Key Roles of AI in Industrial Applications.14 Table 3-2.Securing Industrial AI Across the IIoT Security Function Building Blocks.18 Table 3-3.Core Principles of the AI Ethics Framework.Source:Department of Industry,Australia.23 IIC:PUB:IIAIF:V0.10:ID:2022 5 Industrial Artificial Intelligence(AI)is the use of AI in applications in industry1 and a major contributor to value creation in the fourth industrial revolution.AI is being embedded in a wide range of applications,helping organizations achieve significant benefits and empowering them to transform how they deliver value to the market.This document provides guidance and assistance in the development,training,documentation,communication,integration,deployment and operation of AI-enabled industrial IoT systems.It is aimed at decision makers from IT and operational technology(OT),business and technical from multiple disciplines,including business decision-makers,product managers,system engineers,use case designers,system architects,component architects,developers,integrators and system operators.The document is structured around the architecture viewpoints as framed in IICs Industrial Internet Reference Architecture,namely business,usage,functional and implementation viewpoints.The document discusses the business,commercial and value creation considerations that drive the adoption of AI.It also elaborates on the concerns that arise from the usage of AI,the use cases in industry,and the ethical,privacy,bias,safety,labor impact and societal concerns related to them.On the technical side,the document describes the architectural,functional and data considerations related to AI,and discusses various implementation considerations,such as performance,reliability,data properties and security.The adoption of AI is expected to accelerate in the industry.AI technology will continue to evolve,given the fast-increasing compute power,wider availability of data that can be used for training and the ever-growing sophistication of algorithms.Current IT standards and best practices must evolve to address the unique characteristics of AI itself and specific considerations related to safety,reliability and resilience of IIoT systems.In addition,the growing maturity of organizations about AI will help them appreciate that its benefits far outweigh its risks.The AI standards ecosystem will also continue to evolve,for example the ongoing standards work by ISO/IEC JTC 1/SC42 that provides guidance to JTC 1,IEC and ISO committees developing AI standards.Based on these trends,there should be little doubt that AI will continue to push the state-of-the-art of what is technologically and functionally possible,and thus what is expected to be the reasonable thing to do will equally evolve.Attitudes towards the technology and business expectations about its use will also continue to evolve.In the future,we can expect the use of AI technologies to become the norm rather than the exception,and given the societal benefits of this technology,“not using AI”may eventually become the irresponsible thing to do.1 Smart manufacturing,robotics,predictive maintenance,health diagnostics,and autonomous vehicles.Industrial IoT Artificial Intelligence Framework IIC:PUB:IIAIF:V0.10:ID:2022 6 1 INDUSTRIAL ARTIFICIAL INTELLIGENCE This section introduces the Industrial Internet of Things(IIoT)and the application of Artificial Intelligence(AI)in IIoT ecosystems(Industrial AI).It focuses on the considerations that must be addressed during the AIs full lifecycle within an IIoT system,from design to implementation and operation.1.1 INDUSTRIAL INTERNET OF THINGS The Industrial Internet of Things integrates the industrial assets and machinesthe thingswith enterprise information systems,business processes and people who operate or use them.With these connections to the industrial assets and machines,modern technologies enable the application of AI to machine and operational process data to gain insights into the operations,optimize them intelligently to boost productivity,increase quality,reduce energy and material consumption,increase flexibility and ultimately create new business value.All this must be done while maintaining commitments to safety,reliability,resilience,security and data privacy as the trustworthiness of the systems and conservation of the environment as social values.IIoT is a natural extension of the industrial and internet revolutions.IIoT is a major force driving economic growth,now and for the coming decades,at a greater pace than prior revolutions.As outlined by the World Economic Forum,2“The first industrial revolution used water and steam power to mechanize production.The second used electric power to create mass production.The third used electronics and information technology to automate production.Now,a fourth industrial revolution is building on what has preceded it,blurring the lines between the physical,digital and biological spheres.”To accelerate this digital revolution the Industry IoT Consortium(IIC)is advancing the technology of IIoT across a diverse set of application domains.1.2 INDUSTRIAL ARTIFICIAL INTELLIGENCE Industrial artificial intelligence is the application of AI to IoT applications in industry,in areas like smart manufacturing,robotics,predictive maintenance,diagnosis of infectious disease with machine learning and autonomous vehicles.The use of AI is pervasive in the enterprise,helping organizations achieve significant benefits in terms of better insight,faster decisions and more effective operations.In particular,AI plays a key role in driving the IT/OT convergence with a growing range of practical applications in industry,for example automating routine labor tasks,driving autonomous vehicles,2 World Economic Forum(2016):The Fourth Industrial Revolution:What it Means,How to Respond.https:/bit.ly/3CRmjzz.Industrial IoT Artificial Intelligence Framework IIC:PUB:IIAIF:V0.10:ID:2022 7 understanding speech and performing medical diagnostics.Industrial AI is a major contributor to value creation in the fourth industrial revolution.1.3 ARCHITECTURE VIEWPOINTS This Industrial AI Framework uses the architecture viewpoints of the IIoT,viewpoints for short,as defined in the IIC Industrial Internet Reference Architecture3(IIRA):Business Viewpoint,Usage Viewpoint,Functional Viewpoint and Implementation Viewpoint.This reference architecture document leverages the ISO/IEC/IEEE 42010:20114 methodology to identify these viewpoints.We highlight four important terms from this standard as they are used throughout this document.Stakeholders are individuals,teams,or organizations that have an interest in a system.System concerns are interests in a system relevant to one or more of its stakeholders.Architecture views are work products expressing the architecture of a system from the perspective of specific system concerns.Architecture viewpoints are work products that establish the conventions for the construction,interpretation and use of architecture views to frame specific concerns The architecture viewpoints identify relevant stakeholders and their concerns and articulates how these concerns are addressed.A stakeholder may have more than one type of concern,for example an executive may have business concerns as well as concerns about implementation;a system architect may have usage concerns as well as functional and implementation concerns.Figure 1-1.Industrial Internet Viewpoints.Source:IIC IIRA.The viewpoints provide different perspectives of the complex IIoT system and taken together(see Figure 1-1)express the systems architecture.3 IIC:Industrial Internet Reference Architecture.Refer to Annex C for details.4 https:/www.iso.org/standard/50508.html Industrial IoT Artificial Intelligence Framework IIC:PUB:IIAIF:V0.10:ID:2022 8 The business viewpoint attends to the concerns of the identification of stakeholders and their business vision,values and objectives in establishing an IIoT system in its business and regulatory context.It further identifies how the IIoT system achieves the stated objectives through its mapping to fundamental system capabilities.The usage viewpoint addresses the concerns of expected system usage,typically represented as sequences of activities involving human or logical(e.g.system or system components)users that deliver its intended functionality,ultimately achieving its fundamental system capabilities.The functional viewpoint focuses on the functional components in an IIoT system,their structure and interrelation,the interfaces and interactions between them,and the relation and interactions of the system with external elements in the environment,to support the usages and activities of the overall system.The implementation viewpoint deals with the technologies needed to implement functional components(functional viewpoint),their communication schemes and their lifecycle procedures.These elements are coordinated by activities(usage viewpoint)and supportive of the system capabilities(business viewpoint).For AI technology,the architecture viewpoints can help current and aspiring providers and operators of AI-enabled IIoT systems to identify and gauge the value that AI technology can bring to the systems design and operation.The viewpoints facilitate a systematic way to identify industrial AI system concerns and their stakeholders and bring similar or related concerns together so they can be analyzed and addressed effectively.The deliberation of the concerns is often performed within each of the viewpoints to which they belong,but they should not be resolved in isolation to those in other viewpoints.2 BUSINESS VIEWPOINT The business viewpoint attends to the concerns of stakeholders including business decision makers,for example executive officers,board of directors,general managers,as well as technical managers and plant managers.The viewpoint encompasses their business vision,values and objectives in establishing an AI-enabled IIoT system in its business and regulatory contexts.Figure 2-1.Industrial AI Framework Business Viewpoint.Source:IIC.Business Viewpoint of Industrial AIMaximize value and improve the ROI Uncover valuable insights from data intensive environments Enable digital transformaon Act as agent for the future-proofing of the organizaonIndustrial IoT Artificial Intelligence Framework IIC:PUB:IIAIF:V0.10:ID:2022 9 As Figure 2-1 shows,the primary consideration of this viewpoint is to maximize value to the organization and its ecosystem through a direct improvement of the ROI.For example,AI can more effectively provide insights such as increasing production throughput,avoiding or reducing operating costs,delivering higher margins,enabling new capabilities,minimizing mistakes,reducing inventory,improving safety,making better and faster decisions and improving the quality of products.The application of AI can also maximize value indirectly as a result of improving societal aspects.For example,AI can increase the accuracy of disease diagnosis,help experts predict natural disasters better,improve education through one-on-one tutoring of students,promote a gradual evolution in the job field and reduce on-the-job hazards by enabling automate