Ovidiu Vermesan Interview: Artificial Intelligence for Digitising Industry Applications
River Publishers is pleased to announce the publication of the latest title 'Artificial Intelligence for Digitising Industry'. This book provides in-depth insights into use cases implementing artificial intelligence (AI) applications at the edge.
The following interview with editor Ovidiu Vermesan covers the incorporation of AI by industries, how AI and edge technologies can be used to reach sustainability goals, Industry 5.0, and the security challenges faced when using such technologies.
Can you please start by explaining a little about the book and how it came about?
OVIDIU VERMESAN: This book explores the research, practical results, and exchange of ideas between the representatives of forty-one organisations participating in the AI4DI project to develop Ai-based technologies and applications to support the acceleration of digitising industry. We have included in the book 26 articles presenting the AI technological developments in different industrial sectors. Based on the search on Internet, we can say that this is the first book addressing AI-based technology development and applications in specific industrial use cases and demonstrators.
One hundred twelve (112) international experts from across the industry, research, and academia have contributed to the book with their knowledge and expertise in presenting the latest research and developments in applying AI algorithms, techniques, and methods for digitising various industrial sectors.
The concepts presented herein reflect interaction with other European and international projects addressing the research, development, and deployment of AI, IIoT, edge computing, digital twins, and robotics in industrial environments to strengthen and sustain a dynamic AI technology ecosystem. These concepts and research results shed light on steps in the evolutionary transition to Industry 5.0.
Can you give some examples of the types of industries this book focuses on?
OVIDIU VERMESAN: The book provides in-depth insights into use cases implementing artificial intelligence (AI) applications at the edge. It covers new ideas, concepts, research, and innovation to enable the development and deployment of AI, the industrial internet of things (IIoT), edge computing, and digital twin technologies in industrial environments.
The focus is on five industries: automotive, semiconductor, industrial machinery, food and beverage, and transportation.
Is there a featured application that stands out to you as particularly innovative?
OVIDIU VERMESAN: There are several featured applications presented in the book that are very innovative and briefly highlighted below.
· Development of an autonomous reconfigurable battery system that overcomes the homogeneity requirements of a battery system by combining battery cells with heterogenous performance requirements using AI-based state parameter prediction algorithms and online activation by bypassing the switching modules.
· Reinforcement Learning (RL) techniques used to address the challenge of bringing autonomy in industrial robotic manipulation by implementing an optimal policy in robotics through RL and provide real-world environments where a robot can actively operate and safely explore the different policies for an extended period.
· A multi-purpose robotic platform deployment for indoor use intended for autonomous transportation of the factory's material, goods, or tools. AI algorithms deliver autonomous and cooperative behaviour even in complex environments, and AI trajectory planning manages distributed intelligent traffic control ensuring fast and reliable delivery within the factory.
· Machine learning techniques implemented in semiconductor manufacturing to predict product parameters of inertial sensors, which are determined by the 3-dimensional shape and dimensions of the MEMS device. Data is collected from several process sources, including product measurements in various process steps and processing machine conditions.
· A device-integrated solution for the silicon wafer classification where the AI-based device gets pictures from the test equipment and perform real-time data analysis, giving the wafer default category and binary faulty/non-faulty information. As wafer images are sensitive data, and access to the production line is not possible, the setup is implemented using an offline computer sending data at production rate to an embedded system performing on-device classification.
· An AI-based continuous wafer fab process quality monitoring system using Scanning Electron Microscope (SEM) cross-section based on images acquisition and extraction of etching shapes, deposited thickness and line, and spacer width for analysis. The implementation advances the computer vision methods for automated analysis techniques of semiconductor front-end technologies and provides automation for image measurement and technology identification.
· Innovative robot implementation based on a universal multi-modal cognitive sensing platform that can support different industries by synthetic real-life data generation for AI-training, intuitive human-machine interaction, and usage of robot operating system (ROS) for adaptability of different industrial robots, sensors, and other equipment.
· Development of novel methods of capturing images and data using an autonomous robot to support cameras and sensors. The data analysis from the vineyards allows precise decision-making regarding the yield, vine diseases, and missing vines. Rapid intervention is the critical challenge to optimise the harvest quality, logistics, and production costs. The autonomous environment-aware robot is designed to navigate vineyards to help automate specific tasks such as yield estimation or disease identification.
· AI-based predictive maintenance solution to implement an intelligent monitoring system that separates the equipment's normal and abnormal conditions. IIoT-based sensors are installed to measure different parameters such as vibration, current, sound, temperature, etc. The system design is based on a heterogeneous wireless sensor network consisting of sensor nodes and IIoT devices with different communication interfaces (e.g., BLE, LoRaWAN, Wi-Fi), computing power, sensing range and AI-based processing capabilities.
· Development of an AI fleet management of Mobility-as-a-Service (MaaS) use case, in which two automated last-mile vehicles are controlled according to the transport demands of the users. Scaling up the number of IoT nodes (i.e., vehicles) is simulated. The approach is to optimise overall MaaS vehicle fleet management.
The book covers a lot of the challenges faced by various industries when trying to incorporate AI. What particular challenge do you think is key to overcome to allow industries to fully embrace AI into the future?
OVIDIU VERMESAN: The description of the solutions developed show a common pattern for addressing the AI-based implementations using mainly the same processing phases such as collecting the data from different sources, pre-processing, storage, model development/selection, and AI-based model deployment. The real-time monitoring and data collection across different manufacturing stages and various industries can vary. In most cases, object detection, feature identification, pattern recognition, diagnostics, fault prediction, data reduction, and performance projection AI, machine learning, and data analytics are typical elements in processing information. In most cases, IIoT is customed to the specific requirements based on the physical assets, processes and systems existing in the industry and defined for each industrial use case.
The hardware/software partitioning and sub-system level designs show that the AI-based solutions explore various ways to combine and integrate AI-based technologies in different industrial sectors. The designs described exploit the strengths of different AI-based techniques to produce more robust, flexible hybrids and exhibit more desired qualities than the one using either technology alone. Because per today the AI-based technologies are not mature enough, most demonstrators apply a "trial and error" strategy to find the best solution to fit the implementation requirements. In this context, each industrial solution provides state-of-the-art AI-based techniques and implementation solutions and shows progress beyond it.
The different industrial use cases have various data sources. Some data were extracted from experiments, other from industrial processes. These sources provide the data sets for historical and real-time data (e.g., sensors, actuators, IoT/IIoT devices, ERP systems, databases, data lakes, collaborators and partners, public data, mobile apps, social media, legacy data, etc.). Data quality is an essential element for the design of AI-based systems. Most of the solutions identify the collection of data and the availability of valid data sets a challenge. Identifying and defining the type of data used (e.g., structured, and unstructured) and the formats (e.g., raw data stream, file, block, object, etc.) is time-consuming and varies from use case to use case and among industrial sectors.
How can AI and edge-based technologies help us progress towards a more sustainable future and help us achieve certain sustainability goals?
OVIDIU VERMESAN: Defining the requirements for industry-grade AI is crucial as advanced machines and IIoT devices with enhanced AI capabilities may operate in ways that were not envisaged when the AI-based system was designed and put into operation.
Industrial applications have much higher requirements for reliability, verifiability, safety etc., than the AI-based products designed for the consumer market.
In many cases, phrases like "the data is the new oil" are used to highlight the digital transformation and the use of data in analytics, without considering that as for the oil, when data is used excessively, it is polluting the environment.
Raw data has no value in itself. Instead, the value is created when collected effectively and accurately, connected to other relevant data, done on time, processed, and refined. When well refined, usable data immediately becomes a decision-making tool – information – allowing companies to use it in the manufacturing decision-making and process automation.
AI industrial applications harness AI to enhance efficiency and sustainability while expediting digital transformations. By applying AI, machine learning, and deep learning, manufacturers can advance operational efficiency, dynamically control, and adapt product lines, customise product designs, and plan technological developments.
Implementing AI-based solutions across industrial sectors aims to optimise processes, use of energy, resources, and provide greener solutions than those of current industrial facilities by deploying end-to-end, environmentally friendly manufacturing deployments with a minimal CO2 footprint.
In the book you talk about Industry 5.0. What is your vision of Industry 5.0 and how does this move on from the current Industry 4.0?
OVIDIU VERMESAN: Industry 4.0 has revolutionised the manufacturing sector by integrating several technologies, including cloud computing, big data, and cyber-physical systems. Industry 4.0 goal is to make the manufacturing industry "smart" by integrating machines and equipment that can be monitored and controlled throughout the life cycle.
Industry 5.0 extends these technological advances to facilitate intelligent machine-machine and human-machine collaboration further. The goal is to combine the speed, precision, repeatability, and replicability of the operation of machines with the vision, decision-making, and critical and cognitive thinking of human beings. Industry 5.0 can significantly increase manufacturing efficiency by extending the use of AI technologies to create a versatile connection between humans and machines, enabling constant monitoring and interaction. Industry 5.0 is characterised by the convergence of technologies and integrates the IIoT with AI-based solutions and digital twins to connect physical and virtual manufacturing environments. This convergence makes possible physical and virtual simulations and operating environments in which models based on predictive analytics and managed intelligence enable faster, more accurate and precise, and more reliable decisions.
What are some of the key security challenges faced when applying AI in this way?
OVIDIU VERMESAN: The AI models and the algorithms are costly and valuable intellectual property to protect against cyber-attacks and tampering.
The industrial architectures for designing and deploying AI-based applications consider security as an end-to-end (E2E) crosscutting function that addresses the convergence of information technology (IT) and operational technology (OT) within manufacturing industrial sectors across all layers of the functional industrial domain (e.g., physical, control, operations, information, applications, and business).
The AI-based industrial use cases implement solutions that can adapt to evolving cyberattacks landscape, have high-grade protection requirements with multi-faceted E2E security by design solutions applied from the edge devices incorporating AI-processing units to the edge computing analytics, cloud learning platforms and industrial enterprise back-end systems.
Security is incorporated starting from the use case requirements and specifications, design, integration, deployment, over the entire lifecycle of the industrial application.
The whole security framework for AI-based industrial systems is based on AI security mechanisms, identification, authentication, authorisation, availability, confidentiality, integrity, secure analytics, network prescribed policy, secure communication, security by default, by-design and best practices.
The implementation of the AI-based use cases combined with IIoT, edge and cloud specifically designed for industrial environment requires addressing the enterprise architectural level to gain complete visibility into the industrial networks, to identify possible security threats, ensure process integrity, build secure infrastructures, drive regulatory compliance, and enforce security policies to control risks in order to provide E2E security by design solutions integrated into the industrial systems.