This white paper originally appeared in Automation Alley’s 2020 Technology in Industry Report The Rise of Artifi cial Intelligence in Manufacturing

Written by:

Richard Olawoyin, Ph.D., CSP Associate Professor of Industrial and Oakland University

“Artifi cial intelligence will change 100% of jobs, 100% of industries, 100% of professionals. This is not a technology issue.”

– Ginni Rometty, Executive Chairman, IBM he fourth industrial Technological advances over analyses such as object and speech revolution (Industry 4.0) the past few decades and the recognition, machine tuning and is based on the use of refinement of AI algorithms calibration, robotic motion control, Tcyber-physical models have offered manufacturers the process automation, cognitive quality and digital integration across leading-edge in delivering timely assessment and factory assistance. the connected value chain. This end-to-end solutions, helping to has led to the next generation manage data, deploy applications There are three main categories of technological capabilities on cloud of choice to transform of companies that can benefit and the expansion of the digital products, reskill and upskill the substantially from the enormous ecosystem across industries, workforce to develop the next potential that AI brings to including manufacturing. Digital generation of relevant skills and manufacturing; transformation in manufacturing enhance operational performance is bringing opportunities to at enterprise scale. Hardware • Digital natives – new areas and capitalizing on and software performance have companies that started exiting values in more efficient also significantly improved the with digital core holdings and continue to innovate by adopting digital technologies for all business sectors. Technological investments continue to advance • Digitized incumbents – the relevance of artificial intelligence (AI), where these are incumbent digital incumbent manufacturers take advantage of businesses with over 20% intelligent automation, transcending conventional digitalized core holdings and performance tradeoffs to achieve unprecedented continue to transform the levels of efficiency and quality. core by launching new digital businesses with over 25% organic revenue growth. • Traditional incumbents – ways by optimizing operations, application of AI techniques incumbent companies that increasing productivity, linking including machine learning, continue to compete mainly suppliers, transforming customers teaching and translation, in traditional ways and experiences, increasing workforce knowledge mining, cognitive over 80% of their business productivity and developing new learning, language processing, holdings is not digitized. ways to realize or monetize value. etc., allowing manufacturing and assembly line machines to The Artificial intelligence (AI) jump- process large volumes of predicts that manufacturing starts the transformational data at high velocities. companies that adopt AI-enabled process and powers Industry cloud technologies at scale 4.0 through integration with AI technologies rely on neural (either through partnerships the Internet of Things (IoT) networks that are synonymous or acquisitions of digital assets) and interconnected machines, to the biological neurons in the will lead in the emerging market, enabling real-time visibility across human brain. These networks providing end-to-end solutions in entire supply chains and allowing use various architectures to cloud and edge meeting end-users for more accurate guidance or process information and are demands, and will see more than decisions far beyond what human comprised of nodes and layers 120% in productivity growth in intelligence could offer. that perform different functional the next few years (Martin and

2 | An excerpt from Automation Alley’s 2020 Technology in Industry Report Leurent, 2017). Manufacturing companies are implementing AI tools to optimize digital operations, using intelligent supply chain networks to reimagine their brand and are empowering their workforce as AI can handle mundane and repetitive tasks across the organization, freeing people up to solve more complex problems and focus on getting more impactful work done.

With the adoption of AI, other Emergence of AI in Manufacturing manufacturing areas will have substantial gains as well, including In the 1950s, English sage Alan Turing began to examine the ability of a additive manufacturing by machine to simulate human cognitive functions of reasoning, perception, creating new efficient products learning and teaching, problem solving and decision making. Over time, it and reducing waste and was realized that systems and machines can learn (machine learning - ML) emissions, robotics and process from data and utilize the obtained information for better visualize detections automation by enhancing human- and high accuracy predictions, which are helpful for making valuable and machine collaborations and informed decisions. With the desire to observe and analyze more complex increasing workplace safety and systems, the concept of deep neural networks (deep learning - DL) was cybersecurity by embedding smart introduced in 2010. Convoluted neural networks (CNN), self-learning safeguards into manufacturing and other problem-definition and problem-solving AI approaches were systems to combat vulnerabilities subsequently developed. and threats. Automotive manufacturing is poised for Today, in manufacturing, AI can, for example, power machine learning making vehicles that will be fully in metal cutting equipment. Through AI, machines cut metal with higher autonomous by 2030 and electric accuracy and speeds by making more timely adjustments compared to battery vehicles are expected operator-run equipment. The application of AI can be applied across the to experience an astronomical value chain. For example, AI is particularly popular in autonomous vehicles growth of about 4,000% with an where the onboard software has been trained exhaustively to be self- average vehicle generating roughly functioning. Parts of the training involve pattern recognition tests where 30 terabytes of data per day large volumes of images are presented to the neural nets. The network (McKinsey, 2016). These potentials parameters are adjusted until there is enhanced accuracy in the visual will largely depend on AI and recognition of pedestrians, road features, expected and unexpected data analytics. These disruptive objects, etc. Future improvements in the application and adaptation of AI trends require manufacturers to technologies will include reducing the volume of data required to train and realign their operating model to test them. Within the systems, AI-enabled edge computing (which is at the remain competitive. These product device level), is emerging as the preference for applications with secure changes will introduce new cyber-physical components. competitors, replace component parts with new ones and require The following pages will explore how AI is impacting people, processes and more adaptable supply chains. technology in short, medium and long terms.

3 | An excerpt from Automation Alley’s 2020 Technology in Industry Report People

AI is set to disrupt the world we world (IFR, 2019). ABI Research enablement, accessibility, change live in. According to the McKinsey projects that “the total installed management and the work Global Institute, accelerated base of AI-enabled devices in environment. Digital natives advances in AI and intelligent industrial manufacturing will or incumbent companies are automation will radically impact reach 15.4 million in 2024, with exploring various opportunities our productivity and the way we a compound annual growth rate and people-centered strategies in work. This will include how we of 64.8% from 2019 to 2024” order to secure the value chain in design and manufacture robots to (Robotics Business Review, 2019). the expanding market. perform simple repetitive tasks thereby reducing production As AI technology becomes more Companies competing to become errors. Consequently, fewer entrenched in our daily lives, top economy performers are workers would be required to future products will require expected to invest substantially perform simple tasks, and more more manufacturing innovation in AI technologies. There are nine workers would be needed for that will be demanded by the layers of technology stack for cognitive tasks such as analysis end-users. AI transformation which forward-thinking compa- and programming. In 2020, there of the workforce will take play nies may compete by providing will be over 3 million AI-enabled in six major areas as shown in end-to-end solutions across ser- industrial robots installed and Figure 1: talent management, vices, training, platform, interface functional in factories around the skills enhancement, productivity and hardware requirements.

Fig. 1: AI Transformation in the workforce through connection, prediction and cognition.

End-End Use Cases

Empowering Talent Recruiting Making smart Building shared Build winning talents to management talents hiring decisions experiences teams succeed

AI skills & Skills Technical AI standards & competency enhancement education in AI ethics training training

Guided Productivity Smart factories Machine Remote AI knowledge process enablement & assistance assistance assistance management management

Daily life Accessibility Jobs & tasks Communication Mobility Connectivity activities

Change Lean process in Cross-team Stakeholders Review boards management manufacturing management

Design, Work Smart spaces & Knowledge process/ Safety & health Quality of life environment collaboration workers facilities simulation

4 | An excerpt from Automation Alley’s 2020 Technology in Industry Report Most of the opportunities are not result to profit in the nearest eventually increase in expected to come from layers future. There are three layers for profitability when data can across hardware, accelerator which companies can compete at be generated from end users and head node layers, which the platform level: the framework and third-party providers and will account for up to 50% of layer which includes software appropriately shared. total value to AI vendors. These suites for invoking algorithms on layers coordinate computations the hardware and defining the Manufacturing companies must and perform highly parallel architectures, the algorithm layer have strategic plans to capture the operations based on AI where we have rules for optimal value in this AI enabled emerging requirements. Examples include inference used to adjust artificial market. The consideration must field programmable gate arrays neural network (ANN) weights address core business strategic (FPGAs), graphic processing units such as contrasted divergence questions of where, how and (GPUs), central processing units and back propagation, etc., and when to compete depending on (CPUs) and application-specific the architecture layer which is target value and level of maturity. integrated circuits (ASICs). At the a structured approach for data In the automotive industry, for interface layer, the framework feature extraction such as CNN example, companies that focus facilitates communication between and recurrent neural networks on microverticals (particular hardware and the software such (RNN). The training stack contains use cases) where AI solution are as computation for unified the methods layer with techniques deployed could see big gains, device architecture and Open that optimize the neural net however, they must be on the Computing Language. weights based on data types same page as their suppliers and such as supervised, unsupervised original equipment manufacturers Similarly, about 50% of total value and reinforcement learning (RL), (OEM) as far as embracing AI from AI solutions are expected and the data types layer which solutions. Customers will continue from services at the solutions emphasizes application-specific to be interested in AI solutions and use case layer. Trained DL data presentation to AI systems that can quickly solve specific models will provide improved such as labeled vs. unlabeled. problems and generate strong efficient solutions and efficacy With respect to manufacturing return on investment, this will across all other layers of the stack investments, these layers would reduce costs, reduce downtime in requiring investments. System integrators with direct interface with end users will capture majority of the gains by developing more accurate and insightful AI solutions for customers such as visual recognition for autonomous vehicles. Customers are more likely to pay for AI solutions based on total economic benefits and sustained values.

The other areas of the AI stack will generate indirect value that is expected to spur growth of the digital ecosystem but may

5 | An excerpt from Automation Alley’s 2020 Technology in Industry Report manufacturing, improve predictive maintenance and increase profi tability.

When exploring where to compete, manufacturing digital native and incumbent companies must select the microverticals within their capabilities that will give them a competitive edge in the market, and they must be prepared to address emerging challenges such as cyber vulnerabilities, risks and security. How should the companies compete? Partnerships with other forward-thinking companies or through acquisitions of digital assets to broaden the ecosystem will provide value to the bottom line. By establishing digital partnerships across the value chain (with suppliers, logistics partners, etc.), there could be direct benefi t to customers and digital leaders.

The time to compete is now, the incumbent companies will continue to improve and gain scale that will defi ne the trajectory of progress in the value chain. Traditional incumbent manufacturing companies should begin to focus on appropriate microverticals solutions for their business transformation that will help them establish presence in the market now, then expand to more areas of opportunities in the future. In this rapidly evolving digital market, businesses or units within the manufacturing industry that procrastinate to establish AI solutions in the next two to three years will fi nd it challenging to catch up.

6 | An excerpt from Automation Alley’s 2020 Technology in Industry Report cities’ infrastructure, running systems effectively and efficiently. Some cities are transforming completely into interconnected smart cities. The Trend Force publication “Machine Learning in Manufacturing” highlighted that the global smart manufacturing market will increase to over $320 billion USD at the end of 2020.

Investments by enterprises, local and national governments are already creating prolific prospects for manufacturing companies to compete and provide innovative solutions that will enhance customer experiences, save lives, save time and resources. For example, in the U.S., there are enormous opportunities that government can provide to upgrade traditional city systems to smart city technologies, especially Now (1-2 years) and have lineage. Purposeful with the advent of autonomous partnerships with educational and connected vehicles. Currently, Manufacturers are many years systems in preparing talents for fewer than 1% of traffic lights can away from fully utilizing AI to smart manufacturing will promote be considered as smart traffic replace humans from the shop end-to-end value chain, when the lights. Thus, there are growth floor to the top floor, as ML is right people with keen knowledge potentials in the transportation not yet at the level of human of intelligent business processes industry that manufacturing intelligence and adaptability are involved in problem solving, can support. (Büchel Bettina Floreano, 2018). human centric solutions are The future requires that we achieved. Competencies in how to Near (2-5 years) begin now to in new ways effectively use AI methodologies of developing and operating can help enterprises close the skills Modern manufacturing business and ownership models gap, enhance skills personalization generates vast volumes of data that will strengthen ecosystems. and promote transparency in the from ERP systems and smart Intelligent manufacturing expanding manufacturing market. equipment, and AI systems will using end-to-end engineering be needed to detect and solve techniques will involve vertical Cities around the world are extremely complex problems in and horizontal integration of AI realizing the ease at which they heterogeneous datasets with modeling and IoT technologies. can collect real-time data and highly effective results (El Naqa, Additionally, AI is required with AI capabilities, decisions can 2018). Product and process to be fair, explainable, robust be made quickly for optimizing design in manufacturing use

7 | An excerpt from Automation Alley’s 2020 Technology in Industry Report AI-enabled algorithms based on operational technology (OT), supply chains, thus people will no worker behavior and experience, where it can monitor/control longer be responsible for manual climate change, political status systems and transfer data to intervention in supply chain and economics to predict informational technology (IT) disruptions. Advances in human the inventory levels needed, in a cybersecure convergence, machine interface (HMI) will allow staffing of employees, workforce resulting in optimized intelligent people to control machines and development and how much manufacturing process. analyze data faster, therefore demand there is in the market for a is a solution to a increasing efficiency. In the next certain product (Kozyrkov, 2019). security concern that could drive 5 years, the technical value of Economy of scale suggests that further adoption of AI. It removes HMI is projected to stimulate advancement in AI technologies a hurdle to broader adoption. manufacturing growth by nearly will produce relevant and practical 11% (Transparency Market tools, available to small and Industrial automation using AI Research, 2016). medium-sized enterprises in the will provide better results than next 5 years which will lead to human capabilities, therefore low Broader adoption of AI efficiency gains across the board. and moderate-skilled workers technologies across the However, human’s ability to keep will be displaced (Cheatham 2019, manufacturing industry will up with the data interpretation Moore 2019). This is because enable more to be produced at will lag, while AI will be able to AI powered robots or smart lower costs. Overall, this benefits rapidly interpret the data. In machines are expected to replace the consumers in the market, optimized product manufacturing most human functions on the since this will result in affordable lifecycle management for example, factory floors allowing for capital prices for goods and services. AI can examine product data reallocation and innovation; human On the contrary, many jobs will quickly, to determine averages capacity will be freed-up to create be lost due to lack of required and performance. The trained new products and tasks (Sikorksi, skills (Forbes Insights, 2018). networks will be used to improve 2017). This could allow resources Significant training will be required the output of manufacturing to be traded across global markets for workers in order to adopt capabilities and AI will be able harnessing the benefits of the this technology. As people in the to integrate data from all areas duality of digital capabilities. In factory have more tools at their of the company core business addition, the transformation will disposal, they will need more skills assets to produce quality reduce the safety risks exposures and on-the-job training in order enterprise solutions. for workers. to use the software and controls in a beneficial manner. Though It is predicted that in 3-5 years, Sourcing of tools and AI will help people in many other blockchain technology will be manufacturing parts will surge, ways, it will inevitably create a more embedded with AI in smart expanding the global supply chain. barrier for entry into the technical manufacturing to improve supply Deployment of AI will provide labor pool. As low skill labor jobs chain networks, product quality, significant improvements to disappear, businesses will need supplier order accuracy and track manufacturing efficiency, speed to find new ways to ensure that a and traceability. This will improve and quality of production process, highly skilled labor pool is always the efficiency of how products especially in managing digital available. It is therefore imperative are made, how fast they are supply chain. Expertise in ML for enterprises to invest in training sold, and the quality of services will be needed for certain tasks unskilled workers and potential provided. Blending advanced AI like pattern recognition and to workers so they can be valuable with blockchain will also improve interpret relationships between assets in the industry of the future.

8 | An excerpt from Automation Alley’s 2020 Technology in Industry Report Far (+ 5 years) of Botnets, a network of private in real-time. This will potentially computers infected with malicious help to reduce the numbers Enhanced AI is expected software and controlled as a group of manufacturing workplace to dominate the future of without the owners' knowledge, injuries and fatalities. manufacturing. This will e.g., to send spam messages. As AI substantially reduce the need for technology becomes ubiquitous The AI of the future will be raptly people in various manufacturing in manufacturing, there should be incorporated into people’s lives value chains. The performance equal investments in protecting and will be able to diagnose level and quality of manufacturing assets through strengthened physical and mental health robots will surpass human security measures and the issues that may affect job capabilities. Human error is development of cyber-resilience performance. Through real- inherent in tedious repetitive plan for recovery after an attack. time analytics and monitoring work, while AI completes the of worker’s health and work task the same way, with the One of the benefits of full conditions, AI will be able to same speed, agility and overall integration of AI in manufacturing prescribe corrective actions, consistent quality. is risk reduction. The U.S. reported leading to a happier, healthier over 5,000 job fatalities in 2017 workforce. Advanced AI Experts speculate that robots and alone (U.S Bureau of Labor applications are expected to collaborative robots will replace up Statistics, 2018). Advancements reduce risk in all categories to 20 million factory jobs by 2030, in AI will provide the ability for the of the enterprise. Integrating smart factories will generate $3.7 remaining workers on the factory AI into product lifecycle trillion USD in value, and more floors or other manufacturing management will improve than 50 billion machines will be settings to utilize mixed and human factors and provide connected to the internet (Walker, augmented reality to quickly people with enormous benefits 2019). The cost barrier to entry will identify hazards and control risks from an agile, waste-free system. likely fall to a point where smaller manufacturers and industries will become adopters. As AI tools gain in popularity and application, the jobs that require social skills, empathy and compassion will continue to be held by humans for many more years (Doyle, 2018). Conventional manufacturing jobs will in due course be replaced by software development and cyber- engineering jobs.

In 5 years and beyond, more companies will continue to automate their manufacturing systems. The risk of cyberattacks will also increase since AI can be manipulated by system intruders with destructive motives in forms

9 | An excerpt from Automation Alley’s 2020 Technology in Industry Report Process

Now (1-2 years) The immediate value of AI adoption to enterprises Industry 3.0 led to mass adoption of can help with real-time revision of manufacturing electronics and IT assets to further strategy and production business plans, providing automate production, improving opportunities for businesses to become active visualization for operators. From players in today’s evolving industrial transformation. that period onward, companies have implemented AI-based decision analytics support systems to enable supervisory control systems, and AI provides the benefits of visual on-screen information distributed and advanced process dynamic adaptability to market mode to autopilot mode. controls for digitizing their plants. and operational improvements. The use of machine intelligence However, many manufacturers New or existing software will improve performance and are yet to take advantage of these resources powered by AI are human dependencies such as benefits and still expect operators used to analyze massive volumes variations in skill levels, to manually monitor an array of of data routinely generated for experience and aspirations signals and make decisions based on customizable results. not applicable to machines. instinct, experience and personal judgement. This may have worked for The immediate value of AI Predictive maintenance some, but the fundamental question adoption to enterprises can that remains is that what happens help with real-time revision of Predictive maintenance is an to the operational performance manufacturing strategy and essential business solution for and business profits when the production business plans, manufacturers, where advanced AI experienced operator retires? providing opportunities for algorithms (e.g., ML and ANN) are businesses to become active used to predict process errors and AI can enhance production and players in today’s evolving asset malfunction. In situations continuous maintenance, leveraging industrial transformation. where maintenance cannot be other modern tools like IoT and Big avoided, AI solutions provide Data for providing autonomous and The unique ability of AI to advanced notice on specific smart solutions. The integration of reliably deliver consistent components requiring attention AI in manufacturing processes is complex operational decisions, and advice on optimal methods the driving force behind the Fourth sustain, enhance, and normalize and tools for focused repairs that Industrial Revolution (Harari, knowledge, makes it an are scheduled in advance. 2018). The processing power of indispensable tool especially high-performance AI-enabled where it is becoming more In 2018, McKinsey & Company computing tools together with low challenging to attract and retain reported that the use of a root cost memory requirements (due to skilled operators. Companies cause analysis software tool innovations with ) that are digitizing their core with ML capabilities lead to a offer cost effective alternatives to process operations, are capturing manufacturing yield increase manufacturing companies. Complex the maximum value by offering of 30% while reducing scrap tasks can be automated at reduced operators the ability to readily in manufacturing operations costs, requiring less manpower; make a switch from scheduled (Columbus, 2018). The report

10 | An excerpt from Automation Alley’s 2020 Technology in Industry Report also showed that AI has the For example, Rolls Royce, create substantial challenges capability to reduce testing costs manufacturer of luxury for manufacturers. while increasing the predictive automobiles, provides specialized maintenance accuracy for the maintenance services to airline To attain operational excellence manufacturer’s tooling. Other companies. Their operations and remain competitive, Quality studies have shown that unplanned generate enormous amounts of 4.0 is being implemented where downtime costs manufacturers data and, hence, the company AI tools are integrated with an estimated $50 billion USD deployed Microsoft Cortana data analytics and conventional annually, and that asset failure is Intelligence Suite to perform large quality models. For example, the cause of 42% of this unplanned scale data modelling for anomaly many companies are installing downtime (Kushmaro, 2018). identification in aircrafts, offering 3D cameras or laser inspection clients the opportunity to deliver equipment to replace human AI-assisted predictive uninterrupted schedules. visual inspection. The pieces maintenance is effective for of equipment generate data both planned and unplanned Production Quality and Efficiency that AI uses to improve quality downtimes by taking advantage results. Companies are making of the technology’s essential Digital integration allows different strategic and socio-engineering elements, ANN and ML, to develop units in a manufacturing facility decisions using Quality 4.0 by prediction models for resource to access and use real-time embedding AI into their existing malfunction. The model algorithms information. AI tools can improve quality models for early detection perform regular evaluations process conditions by detecting of quality anomalies, such as and updates of the production quality errors that human subtle abnormalities in machine processes with enhanced operators may not find. The design performance, variations in quality sensitivity, to identify system and and manufacture of products must of raw materials, microscopic program anomalies. This is helpful consider consumer variables such deviations from design intent, for reducing downtimes, providing as changes in customers choices, etc. This is prevalent in semi- just-in-time and focused repairs, durability, high quality, perfection conductor manufacturing where pre-scheduling of maintenance and usability. Meeting these highly magnified images of tasks and for extending the needs with short time-to-market finished wafers are scanned and Remaining Useful Life (RUL) of deadlines while complying with AI compare to clean wafers to production machineries. all required quality regulations detect quality issues.

11 | An excerpt from Automation Alley’s 2020 Technology in Industry Report Real-world Success with AI

A Siemens factory in Amberg, Germany uses digital twins to optimize throughput. Before the products hit the physical assembly line, they are tested in the digital twin to highlight possible bottlenecks or problems in real time. This has increased productivity by over 1,400% (Forbes, 2019).

The use of AI-enabled generative design software can improve design quality of products, where through exhaustive iterations, the optimum product solutions are proposed relative to parameters such as type of materials, budgetary requirements, time to production, manufacturing methods, etc. provided by the design engineers. An advantage of this is that AI algorithms do not make human-like assumptions, optimal solutions are selected based on thorough analysis of actual process performance across a wide range of possible outcomes.

In 2016, Bosch conducted a case study that utilized a network of ML programs to analyze their manufacturing processes. The study lead to isolating bottlenecks in manufacturing processes resulting in a leaner process with 35% reduction in hydraulic calibration test times (Srinivasan, 2016).

Siemens has installed many sensors in its gas turbines, and data is transmitted to an AI-enabled data processing solution which intelligently adjusts fuel valves to maintain the lowest emission levels possible. Cloud-based smart factory service solutions such as Microsoft Azure, DocumentDB and HDInsight provide manufacturers the ability to automate their processes, collect on production history, ensure production efficiency and quality, and project emerging defects.

12 | An excerpt from Automation Alley’s 2020 Technology in Industry Report Manufacturing companies are It was making the right decision patterns (Forbes, 2019) and collecting data from customer for the wrong reasons (Kenyon secure the cyber infrastructure of experience, demographics, political 2018). Based on this possibility manufacturing processes, overall immediacy, climate change, for error, digital developers need strengthen the security of their geographic trends, macro and to monitor and rigorously test systems. socioeconomic variables. With the how AI is coming to conclusions, help of AI algorithms, they can use to ensure it is congruent with the Near (2-5 years) the data to predict how to optimize enterprise’s best interests. AI energy resource consumption, can be prone to “bias” based on The market for AI in manufacturing how to manage the manufacturing the data presented to the model. is projected to exceed $16 billion supply chain and staffing needs in Developers need to ensure they USD by 2025 with compound order to better respond to market are not introducing bias into the annual growth rate of 43% demands. Studies have shown model based on validation data in North America (Figure 2) that AI integration into automated sets or classifications. (Global Market Insight, 2019). manufacturing processes The ML segment for AI in the yielded 10% reduction in annual Cyber-attack is another risk which manufacturing market is predicted maintenance costs, 20% reduction can be from internal and external to be over 47% of the industry in downtime and a 25% reduction corruption from both current . Businesses are exploring AI in costs from quality inspection. and ex-employees as well as technology to improve efficiencies (Kushmaro, 2018) intruders. Information technology in biotech manufacturing and experts have begun to use ML process development to improve Risk and Opportunities for algorithms to identify suspicions their gross margins. Solutions

Presently, most manufacturing- Fig. 2: AI in Manufacturing Market in the next five years. based AI tools are either reactive (Courtesy Global Market Insights) machines or limited memory AI, that assist humans in decision $16 Billion making. A major risk with AI- assisted manufacturing is that it is difficult to fully understand how AI reaches its decision. For example, an object recognition AI was developed to tell the difference between huskies and wolves from a picture. The program reported accurately for some time, but unbeknown to the developers, was analyzing the background rather than the foreground of the images. The 47%: pictures with the wolves had snow Machine Learning in the background and the images of huskies did not. The algorithm taught itself “if snow, then wolf.”

13 | An excerpt from Automation Alley’s 2020 Technology in Industry Report The drivers of this economic There are still several challenges their system can be controlled, indicator include increases in obstructing the wide-spread monitored and analyzed efficiently. venture capital investments in AI adoption of AI technologies, and rapid adoption of industry 4.0, including the complexities of This convergence of IT/OT assets both are projected at 7.7% growth industrial data sets and the is expected to open the door for medium term. Expansion of digital shortage of skilled workface cyber vulnerabilities, risks and data is predicted at about 4.8% needed for full adoption attacks even further. Therefore, within the next five years. All of in manufacturing. investments in AI-based digital which will enhance operational transformation must include efficiency and reduce time to The rapid, across-the-board security and mitigation strategies manufacture and market products. adoption of AI tools, principles and such as recommended by the methods in manufacturing will be SANS Institute (2017): The Asia Pacific (APAC) region is driven by the empowered end- expected to continue to dominate users in the new economy. • documenting network maps the market growth of AI in • procuring tools to analyze manufacturing, largely credited Understandable or explainable network traffic to rapid adoption of Industry 4.0 AI (U/XAI) technologies are • establishing cybersecurity for smart manufacturing plants emerging and will come with assessment and audit plans in China, Japan and South Korea. most AI-based systems by 2025. • establishing credential Increased investments in AI This will help companies better security program technologies in India and China understanding how AI reaches • implementing tools to contribute to this growth. In optimal solutions for intelligent identify and notify security Europe, proliferation of Industry decision making. The UAI or XAI analysts of incidents 4.0 in manufacturing and the rise will be able to generate detailed • regularly reviewing and in labor costs are expected to reports for human understanding patching vulnerabilities drive the growth of AI-powered about the AI’s ability to function • creating a security advanced manufacturing adoption correctly and reach the best protocol for informing in the region. decisions appropriately. Though employees of phishing not currently widely used in and social engineering Of all respondents surveyed in manufacturing (Turek, 2018), the a recent poll by Forbes Insights, tool will further improve how Presently, only about 6% of results showed that 44% from the accurate results are obtained. It manufacturers have implemented automotive manufacturing sector will enhance users’ experience blockchain technologies at-scale consider AI as ‘highly important’ where users can reteach the AI (Columbus, 2018). Growth trends to the manufacturing function in if determined that it is returning are forecasted in coming years the upcoming five years, while 49% erroneous solutions. which will allow for effective deemed it as ‘absolutely critical communication between suppliers, to success.’ In 2020, estimates Based on economic indicators and at the same time reducing the risk of about 1.7 megabytes of new other factors, more manufacturing of fraudulent disruptions. data will be created per second companies are expected to take mainly from IoT devices. It is steps in digital development Far (+ 5 years) critical for manufacturers to adopt and transformation by investing AI technologies for extracting in IT infrastructure that will be Machine learning tools will be used valuable insights from the data. scalable and agile. Enterprises will to optimize the digital supply chain seek IT/OT convergence so that by analyzing big datasets in order

14 | An excerpt from Automation Alley’s 2020 Technology in Industry Report to determine predictive metrics use machine learning algorithms. on Big Data and ML algorithms. that may cause economic and When it is market ready, AGI Manufacturing companies that environmental disruptions. could ideally design, build and realize the value of AI early, and integrate manufacturing processes use the tool to enhance decision Literature shows that researchers with little to no input from skilled making will generate higher market have been working on the concept humans. (Forbes, 2019) This future margins. The stakes are very of developing artificial general will allow automated system to high for traditional incumbents intelligence (AGI). This will be a achieve reality and robots will and other risk averse companies variation of AI where the machine become adaptable, cheaper, faster especially those under capital intelligence will be equal to and and able to redesign and manage market pressures and with volatile exceeds human-like intelligence. production lines. margins. Reluctance to invest in The AGI will not be developed AI technologies will restrict their with any specific application in Opportunity & Cost of capabilities to obtain information mind which would make it to be AI Adoption from their processes for improving super powerful because there end-to-end performance and would be little or no restrictions AI represents a paradigm shift— adapting to the constant changes on how it can be utilized. The AGI from traditionally difficult, in needs of end-users, partners will incorporate fundamental expensive, strict rule-based and suppliers. As such, they will psychological elements including solutions to cost-effective adaptive fall behind and miss out in the object permanence and how to self-learning solutions based competitive new market.

15 | An excerpt from Automation Alley’s 2020 Technology in Industry Report Technology Evolution

Research and innovation into advancements in AI have grown globally about 12.9% since 2015 (Aguis, 2019). Manufacturing problems such as scheduling, diagnosis, modeling and design are very difficult to solve. However, innovations into AI technology have expanded the opportunities for businesses to monitor production quality control and performance, shorten design and production time, significantly reduce material waste and revolutionize how we use predictive maintenance.

Now (1-2 years) division in Munich, Germany, 2018). It has been observed that AI developed a two-armed robot that is evolving from complex integrat- The AI of things (AIoT) is the digital can manufacture products without ed systems to a modular digital nervous system of Industry 4.0. having to be programmed. Both ecosystem. (Satell, 2019). Artificial intelligence is considered arms of the robot work together the functional brain of the system to complete tasks that are inter- Near (2-5 years) while IoT carries out AI’s instruc- preted from CAD models, without tions. The AIoT integration creates the need to program the robot’s Digital twin and digital thread a fully independent functional movements (Thilmany, 2018). The technologies will gain in popularity system (Janakiram MSV, 2019). robot uses digital integration to at the middle of the next decade. The digital core (a central system save time and energy. Digital twin innovations will help for system data and business data) manufacturers save money while allows businesses to combine data Knowledge-based systems: in control of their technology by analytics, AIoT and ML to improve integrated with ANN, CAD and building digital prototypes, while the value chain of businesses. CAE software are used in design the digital thread and blockchain and manufacturing processes will assist to secure digital Intelligent automation and the to increase work efficiency and transactions in the ecosystem. digital core solutions help solve reduce the design cycle time. These tools will be applied to complicated problems in almost support generative design which all areas of engineering, including Researchers used an expert sys- would be used to vastly reduce the manufacturing using rule and tem and ANN for automatic design time it takes for manufacturers to knowledge-based systems. of compound die and predicted its test new ideas. fatigue life. (Salunkhe, 2017). Rule-based system: logic is A few manufacturers, such as currently used to automate The mini applications (mini-IAs) Airbus, Under Armour and Stanley assembly lines. This is not true from Integrated Analysis Systems Black & Decker, presently use AI because humans program provide small and medium size generative design, but based on the machines and the machines companies the capacity for com- performance, generative design is follow the instructions, though the puter, storage, networking, and expected to be widely used in the machines are not self-learning or analytic software they need for coming years. (Akella, 2018). making decisions independently. their operations. The mini-IA is Moreover, the focus is shifting often purpose-built, rack-mount- Far (+ 5 years) quickly to true AI technologies. able, cloud-ready, and built to scale for data warehousing and analytics The goal of manufacturing In 2018, a team of researchers at —a powerful combination of companies will be to have a Siemens Corporate Technology hardware and software (Hemant, 360-degree improvement in the

16 | An excerpt from Automation Alley’s 2020 Technology in Industry Report design of products, digital supply chain networks and enhance the digital core for maximum outcomes. Consequently, it would increase output by improving the speed, quality and flexibility of business operations.

Siemens is already exploring the possibilities of developing future advanced technology such as the Click2Make. This smart application would allow companies to submit an idea for an item they are inter- ested to manufacture. The system will automatically gather bids from numerous manufacturers and allow the customer to select a suitable supplier. Click2Make would then quickly provide the supplier with the plans immediate- ly to ensure a quick turnaround. The technology would focus only on small batches of products.

This technology would increase manufacturers’ creative advantage and optimize their facilities, while enhancing the communication between supplier and manufacturer (Walker, 2019).

The dawn of this era will usher in the Fifth Industrial Revolution (Industry 5.0), where products and services will be readily custom- izable and highly personalized. Automating repetitive and com- plex future tasks using AI-based algorithms will optimize the exist- ing creative process. Inevitably, in the manufacturing industry, robots will become super intelligent and collaborate with humans on inno- vation-based projects.

17 | An excerpt from Automation Alley’s 2020 Technology in Industry Report In the future, coding for robots development. This goal can be solutions that will replace will become more cost effective, achieved either by appropriately humans where needed, but offering more options for upskilling current employees, ultimately offer solutions that assembly parts. The requirement working with expert consultants will promote human value. will be to specify the task and or hiring new multiskilled the system will automatically employees (translators), who To make coding part of the transform these specifications can use technical, business and elementary school curriculum into actionable programs (Zistl, change management human could be a major component in 2017). A people-oriented, robust, knowledge for developing the development of the platform proactive approach can make a new ideas collaboratively with that will support transgenerational world of difference to enhance AI-machines. The translators’ solutions with great potentials the customer experience when function will be to integrate for transformation and disruption integrated with advanced analytics management, business and in manufacturing. Having and emerging digital technologies. market intelligence, Big Data and children learn about how to Traditional and digital incumbent quality engineering, using expert use AI solutions to build and manufacturers can compete in the perspectives for AI solution program machines efficiency with ecosystem and come out on top if delivery that will improve the minimal human error will help they begin now. To design, develop, customer experience. prepare future skilled translators, enhance and sustain reliable AI engineers, scientists, business and integration and utilization, it will The manufacturing industry management analysts and decision require focusing on big ideas that and academic institutions makers that will innovate and will include significant investments must collaborate even more, develop future AI-based end-to- in human capital for skills on developing human-machine end solutions.

18 | An excerpt from Automation Alley’s 2020 Technology in Industry Report AI Solution Use Cases

In advanced manufacturing, where AGI also called strong AI is Some examples of where AI corporate performance is driven the human-level AI that will can help improve end-to-end by operational performance, be required for tomorrow’s performance include: the greatest potential for AI- manufacturing. After this based solutions is in digital period, we will enter the era Predictive maintenance: This manufacturing supply chain and of superintelligence where technology uses the power logistics management. Advanced intellects that vastly exceeds the of machine learning to detect DL, RNN and CNN collectively cognitive capabilities of humans anomalies. Data from IoT-based account for 40% (up to $5.8 trillion will develop, in almost all areas sensors have been analyzed as USD) of the annual potential of interest. Available solutions time series data for predicting value of all analytics techniques like the Microsoft Azure Suites, the RUL of components. High in the market (Chui et al., 2018). Amazon Web Services, Google dimensional data in different Ongoing advancement is helping to Cloud, Baidu AutoBrain, IBM formats and configurations can address the technical constraints or in-house AI algorithms now be analyzed using DL to hindering the full adoption of AI are improving the customer predict future failures and warn techniques including availability experience and changing the way of planned interventions to help of large volume training data sets. businesses operate, but more will drastically reduce costs and Business readiness and societal be needed to keep up with our downtimes and boost productivity. constraints such as AI ethics, changing world. Investments in AI safety, regulation, privacy and must address research challenges Real-time predictive modeling security may slow the adoption of with innovation, AI scalability, powered by AI: Real-time optimi- AI solutions for manufacturing of security, trust and privacy issues. It zation using logistic modeling add products and delivering services in is crucial to accelerate AI research significant value to the bottom public, automotive, food and drugs, collaborations with institutions line and saves cost. For example, medicine and pharmaceutical, to push the boundaries of AI a European trucker in the manu- health care management, banking possibilities farther and the full facturing supply chain optimized and insurance sectors. benefits to humanity nearer. its routing delivery traffic through continuous logistic estimation and behavioral coaching for drivers powered by AI. Fuel consumption and costs were reduced by about 15%, which also reduced the maintenance costs.

Personalized marketing and customer service: Customer experience can be enhanced with natural language processing for advanced speech recognition and emotional intelligence management. DL analysis can help identify problems in

19 | An excerpt from Automation Alley’s 2020 Technology in Industry Report message tone from audio, video or transcripts. For instance, AI can analyze multiple variables in split seconds from the customers initial contacts and recommend the best solution that will make the experience seamless for the customer, including connecting to a human operator, therefore shortening the service time. In sales and marketing, companies like Netflix, Facebook and Amazon have already implemented tools that quickly recommend the “Next Product or Service to Purchase” with individualized pricing to targeted customers. Using customer data to personalize offers can convert a lot of sales items can be a great way to ensure manufacturing enterprise creates and will drive growth in the product quality. vulnerabilities; however, AI is manufacturing sector. an effective predictive tool for At one of the largest adhesive vulnerabilities, risks, and attacks. Smart AI Robots in manufacturers with over 130 The use of AI predictive tools can Manufacturing: Manufacturing plants worldwide supports improve forecast accuracy by companies are developing the aviation, automotive and over 20%, according to McKinsey and acquiring intelligent and electronics industry supply chain, & Company. User and Entity autonomous AI robots for smart Cerebra’s IoT signal intelligence Behavioral Analytics (UEBA), and manufacturing. Industrial analytics platform was implemented to other AI tools are used to analyze systems such as Intelligent process improve quality and increase the Big Data and for detecting odd automation (IPA) and robotic yield by controlling softening patterns such as micromovements process automation (RPA) are used points and viscosity. Within three of data packets. to improve process and product years, sensor data (i.e., torque, quality, reduce downtime, predict RPM, temperature and pressure) These tools can effectively detect performance and optimize labor. from plant operations analyzed exposures to bugs and botnets Advanced AI algorithms are used helped the company save over and use appropriate techniques to improve the supply chains such $140 million USD from to secure the automation with as logistics, warehouse, fleet defective parts. cybersecurity and cyber-resilient and freight management (real- solutions. According to ABI Re- time tracking). New AI-enabled Asset Protection and search, it is projected that by 2021, solutions include the manufacture Cybersecurity: Advancements AI cybersecurity expenditure on of autonomous and connected in technology will also drive Big Data, smart automation, and vehicles, and intelligent drone attackers to innovate in new data analytics will reach $96 billion delivery. Complex AI-powered ways to bypass security and USD. There are opportunities exist vision and Quality 4.0 techniques cause intentional damages. The and for all businesses to excel in to explore defects in produced cyber-physical nature of most this new market.

20 | An excerpt from Automation Alley’s 2020 Technology in Industry Report Action Items

• Manufacturers without concrete plans for AI adoption into their operations must start now by implementing a stepwise adoption approach with the best benefits for the value chain. This will also reduce the dependency on the human labor force. Therefore, all end-to-end solutions must consider the improvement of both human and machine intelligence within an enterprise to ensure the adoption is sustainable. • Upskilling of existing manufacturing workers in automation and AI tools will play to the advantage of the business. Efforts must be made to hire new employees that can use automation and intelligent strategies to solve problems, this will help prepare future operators, engineers, digital experts, translators and innovators for the factory of the future. The business must also engage AI experts that will continuously reflect on market evolution and the value chain. • Based on the cyber-physical nature of digital manufacturing, there will be increased emerging risks to automated and connected systems/processes which may expose manufacturers to cyber-attacks. Appropriate investments in robust physical and cybersecurity systems must be prioritized. • Collaboration within the ecosystem will help develop interactions between AI platforms across similar industries and partners in the value chain, allowing for efficient learning from one another and quick implementation of useful knowledge.

21 | An excerpt from Automation Alley’s 2020 Technology in Industry Report About Automation Alley

utomation Alley is the World Economic Forum’s Advanced Manufacturing Hub (AMHUB) for North America and a nonprofit Industry 4.0 knowledge center with a global outlook and a Aregional focus. We facilitate public-private partnerships by connecting industry, education and government to fuel Michigan’s economy and accelerate innovation. Our programs give businesses a competitive advantage by helping them along every step of their digital transformation journey. We obsess over disruptive technologies like AI, the Internet of Things and automation, and work hard to make these complex concepts easier for companies to understand and implement.

Download the Full Report

Automation Alley members are able to download the 2020 Technology in Industry Report free of charge! Log into your member portal and find your copy under the resources tab.

Not a member? Join Automation Alley today and let us help you increase revenue, reduce costs and think strategically. Contact 800-427-5100 or [email protected] to learn more.

Copyright

All rights reserved. No part of this publication may be reproduced, distributed, or transmitted in any form without the prior written permission of Automation Alley. “Fair use” excerpts may be included in news or research reports provided that a complete citation is given to Automation Alley.

Our Contact Info

2675 Bellingham Troy, MI 48083-2044

Phone: 248-457-3200 Toll Free: 800-427-5100 Fax: 248-457-3210

Email: [email protected]

Website: automationalley.com