AI Techniques in Advanced Manufacturing and Edge Computing

AI in Advanced Manufacturing

  • AI can be applied in advanced manufacturing.
  • Various AI applications in advanced manufacturing.
  • Edge computing concepts and its significance.
  • The impact of AI on sustainability.

What is AI in Manufacturing?

  • Historical context: AI in manufacturing has evolved through different industrial revolutions.

Industrial Revolution 4.0

  • Definition: Characterized by the integration of various technologies.
  • Importance: Understanding its impact.

Technologies Driving the 4th Industrial Revolution:

  • Machine Learning (ML)
  • Automation
  • Advanced & Predictive Analytics
  • Internet of Things (IoT)
  • These technologies enable real-time monitoring and data-driven decision-making.

AI as a Manufacturing Game Changer

  • AI tools enable manufacturers to:
    • Process large volumes of production data
    • Detect anomalies in real time
    • Forecast demand & customer behavior
    • Gain end-to-end visibility across sites
    • Continuously learn, adapt & optimize operations
  • 4IR technologies could generate 3.73.7 trillion in value by 2025.
  • AI alone may contribute 1.2–21.2–2 trillion to manufacturing and supply chain management.
  • Companies adopting AI report measurable gains:
    • 16%16\% achieved 10–19%10–19\% cost reductions
    • 18%18\% saw a 6–10%6–10\% revenue increase

AI-Driven Benefits in Manufacturing

  • Predictive Maintenance: Reduces unplanned downtime.
  • Near-shore Operations: Leverage 3D printers & robots for cost savings.
  • Generative Design: Creates efficient, low-waste product designs.
  • Advanced Forecasting:
    • 10–20%10–20\% improvement in accuracy
    • 5%5\% reduction in inventory costs
    • 2–3%2–3\% increase in revenue

AI AND THE FUTURE OF MANUFACTURING

  • Cost savings and revenue growth.
  • End-to-end visibility of manufacturing operations.
  • Reduce labor costs and stay resilient despite supply chain disruptions.
  • Tackle operational challenges and disruptions.
  • Ensure efficiency and reduce waste.

USE CASES OF AI IN MANUFACTURING INDUSTRY

  • Logistics
  • AI-Based Robots
  • Supply Chain Management
  • AI Autonomous Vehicles
  • Factory Automation
  • IT Operations
  • Design and Manufacturing
  • Artificial Intelligence and IoT
  • Warehouse Management
  • Process Automation
  • Predictive Maintenance
  • Product Development
  • Connected Factory
  • Visual Inspections and Quality Control
  • Purchasing Price Variance
  • Order Management
  • Cybersecurity

AI in Supply Chain

  • Evaluate “What-If” Scenarios
    • Simulate outcomes based on time, cost, and revenue
    • Optimize decisions for last-mile delivery performance
  • Improve Last-Mile Delivery
    • Predict optimal delivery routes
    • Track driver behavior and performance in real-time
    • Analyze weather, traffic, and historical data to forecast accurate delivery times
  • Gain Full Supply Chain Visibility
    • Monitor and manage everything from capacity planning to inventory tracking
    • Build real-time, predictive models to assess supplier reliability
    • Receive instant alerts on supplier failures and quantify disruption impact

AI in Supply Chain

  • AI-enhanced supply chains will reduce:
    • Forecasting errors by 20−50%20-50\%.
    • Lost sales by 65%65\%.
    • Over-stocking inventories by 20−50%20-50\%.

AI in Warehouse Management

  • Real-Time Monitoring for Smarter Logistics
    • Continuous warehouse monitoring enables proactive logistics planning
    • Real-time data helps identify bottlenecks and streamline operations
  • Demand Forecasting for Stock Optimization
    • Predict future demand to stock inventory in advance
    • Minimize transportation costs while meeting customer expectations
  • Warehouse Automation with Robotics
    • Robots can track, lift, move, and sort items efficiently
    • Frees up human workers for strategic tasks, reducing workplace injuries
  • Automated Quality Control & Inventory Management
    • Lowers warehouse management costs
    • Boosts productivity and reduces labor dependence
  • Business Impact
    • Improved efficiency leads to higher sales and increased profit margins

AI in Warehouse Management

  • JD.com Fully Automated Warehouse in Shanghai.

AI for Factory Automation

  • Current Challenge: Human-Dependent Monitoring
    • Factory operators often rely on intuition and experience to:
      • Monitor multiple signals across screens
      • Manually adjust equipment settings
    • Problems with This Approach:
      • Error-Prone & Inefficient:
        • Manual processes can lead to mistakes and equipment malfunctions
        • Impacts overall factory efficiency
      • Loss of Knowledge:
        • Skilled operators are hard to replace
        • Their departure results in a loss of critical contextual knowledge

AI for Factory Automation

  • How AI Transforms Factory Operations:
    • Automates complex tasks traditionally done by humans
    • Continuously monitors operations to detect anomalies and send real-time alerts
    • Reduces labor costs while boosting overall productivity and efficiency
    • Builds a centralized operational knowledge base for easier employee transition
    • Minimizes workforce requirements to run the factory effectively
    • Enables flexible scaling of production based on demand shifts and strategy

AI for Factory Automation

  • BMW Factory – Integration of A.I. in the Production Line

AI in Predictive Maintenance

  • AI’s Biggest Value Driver in Manufacturing

  • Predictive maintenance delivers 0.5–0.70.5–0.7 trillion in global value.

  • It is the top Industry 4.0 priority.

  • AI Systems Enable Predictive Maintenance By:

    • Capturing and analyzing large-scale sensor data (including audio, video, and GPS) from the shop floor
    • Detecting anomalies or inefficiencies—e.g., unusual engine sounds or irregular assembly line patterns
    • Preventing unexpected equipment failures, reducing downtime and improving operational efficiency
    • Minimizing repair costs by identifying and addressing component-level issues before full machine breakdowns occur

AI in Predictive Maintenance

  • Semantic AI for Predictive Maintenance of Railway Track Systems

AI in Quality Inspection

  • Computer Vision for Defect Detection
    • Utilizes high-resolution cameras to monitor production lines in real time.
    • Detects subtle defects beyond human perception and triggers automatic corrective actions.
    • Reduces product recalls, minimizes waste, and ensures consistent product quality.
  • Real-Time Anomaly Detection
    • Identifies hazards like toxic gas emissions as they occur.
    • Enhances worker safety by enabling immediate response to operational anomalies.
  • Augmented Reality (AR) for Inspection & Training
    • AR overlays compare physical parts to supplier specs to detect deviations.
    • Supports remote troubleshooting and real-time training, enabling expert guidance from any location.

AI for Purchasing Price Variance

  • For manufacturers, fluctuations in raw material costs can significantly impact profit margins. Accurately estimating these costs and selecting reliable suppliers remain critical challenges.
  • With AI-powered dashboards, manufacturers can monitor:
    • Resource attributes such as material type, pitch, diameter, and surface finish
    • Supplier characteristics including location, brand, and historical performance metrics
  • Using AI-driven algorithms, they can:
    • Group and identify the right product parts required for production
    • Predict standard purchase prices based on historical trends and market data
    • Establish reliable benchmarks for comparing prices across suppliers
  • AI also enables centralized management of procurement data, simplifying the tracking of parts across multiple vendors and enhancing overall procurement efficiency.

AI in Generative Design

  • AI design software generates multiple optimized product designs by using input parameters such as raw materials, dimensions and weight, manufacturing methods, cost and resource constraints, etc.
  • Key Benefits:
    • Creates a wide range of design permutations
    • Simulates real-world conditions
    • Enables engineers to select the most efficient, cost-effective solution
  • Real-World Example: Nissan uses AI to generate breakthrough car designs in seconds—far faster than traditional methods.
  • Sustainability Bonus: AI can also suggest optimal production "recipes" to minimize material waste and reduce energy usage.

AI in Generative Design

  • Autodesk researchers are studying how artificial intelligence can be used with existing generative design technology.
  • The swingarm (blue) was considered “impossible” to create using traditional design and manufacturing techniques.
  • But an AI algorithm found a way by combining generative design for optimization with 3D printing methods.

AI Based Robots

  • AI robots use self-learning algorithms to automate repetitive and decision-based tasks. Over time, they continuously improve, leading to:
    • Higher precision and efficiency
    • Reduced errors and downtime
    • 24/724/7 operation without breaks
  • Boosting Productivity and Safety
    • Robots handle heavy or hazardous tasks, allowing humans to focus on precision work
    • Improves workplace safety and performance
    • McKinsey reports a potential 20%20\% productivity boost in labor-intensive environments

AI Based Robots

  • Real-World Adoption
    • Automotive: Robots manage complex car assembly lines
    • E-commerce & Packaging: Offer faster, cost-effective, and low-error alternatives to human labor
  • Common Applications:
    • Welding
    • Painting
    • Drilling
    • Product inspection
    • Die casting
    • Grinding

AI and IoT

  • The Industrial Internet of Things (IIoT) refers to smart, connected devices with embedded sensors that generate real-time operational data. When combined with AI, IIoT drives higher precision, efficiency, and productivity in manufacturing.
  • Key Use Cases of IIoT:
    • Wearables & Smart Glasses: Hands-free access to instructions and real-time situational awareness
    • Equipment Monitoring: Track machine performance, energy consumption, environmental temperature, and toxic gas levels to enhance safety and uptime
    • Smart Energy Management: Automated lighting and HVAC control for reduced energy usage
    • Edge Analytics: Analyze real-time data from production-floor devices for informed, on-the-spot decision-making

Key Takeaways – Why Manufacturers Must Adopt AI Sooner

  • A McKinsey survey found that companies embracing digital transformation in manufacturing are leading the industry. They adopted 4IR technologies, such as big data analytics, AI, AR and VR, IoT, predictive analytics, automation, and robotics, among others.
  • Key Benefits Reported by Early Adopters:
    • 30−50%30-50\% decrease in machine downtime
    • 15−30%15-30\% increase in labor productivity
    • 10−30%10-30\% improvement in throughput
    • 10−20%10-20\% reduction in quality-related costs
  • Early adopters have already started gaining a competitive advantage by substantially lowering operating costs, improving time-to-market, and optimizing performance. These benefits will only grow over time, increasing the gap between the early adopters and the laggards in the industry.

EDGE COMPUTING

What is Edge Computing

  • Edge computing is a distributed computing model in which computing takes place near the physical location where data is being collected and analyzed, rather than on a centralized server or in the cloud.
    • Performance
    • Safety
    • Availability

Edge Computing Vs. Cloud Computing

  • Comparison of Edge and Cloud Computing.

Edge Computing

  • Idea is to push applications, data and computing power to the edge of the Internet, in close proximity to mobile devices, sensors, and end users
  • An early example is Akamai, with servers around the world to distribute web site content from locations close to the user (content delivery networks, or CDNs)

Why Edge Computing ?

  • Bandwidth Efficiency: Instead of sending a lot of data to the server and back, we can use on-device inference to send tiny amounts of data when needed. This is useful for remote places with no proper internet.
  • Reduced Latency: Sending data to a server involves a round-trip delay which gets in the way when working with real-time data. This is no longer an issue when our model is at the edge. When inference is super fast we can solve high-performance actions like real-time object tracking for a robot.
  • Enhanced Privacy & Security: When data stays on device users benefit from increased privacy and security since personal information never leaves their devices. This benefits privacy-sensitive applications like security webcams and health-care data.

Edge Computing: The Telco View

  • Opportunity for providing edge computing devices in existing infrastructure
    • e.g., micro data centers at the base of cellular towers
  • Edge Computing according to the Open Edge Computing Initiative

Edge Computing: The Cloud Provider View

  • Goal is mainly to provide
    • Content Delivery Network (CDN) services
    • IoT data processing and aggregation for data in transit to the cloud
  • Examples
    • Azure IoT Edge — deploy business logic to edge devices and monitor from the cloud
    • Amazon
      • AWS CloudFront — CDN Service, includes Lambda@Edge
      • AWS Greengrass — connected IoT devices can run AWS Lambda functions and other code on locally-collected data
  • Industrial IoT: IoT to Edge to Cloud

Edge Computing: The “Appliance” View

  • Goal is to provide a “data center in a box” to push cloud computing capabilities to the edge
    • Often combined with networking capabilities such as edge gateways and smart routers
  • Many players in this space, such as Amazon, Cisco, Dell EMC, HPE, etc.
  • Disconnected Operations
    • AWS Snowball Edge — large-scale data transfer service with an embedded computing platform (based on AWS Greengrass plus Lambda functions)

Edge Computing

  • Field Operations
    • People that spend time away from their main offices or labs, such as researchers, medics, and sales personnel, can leverage portable surrogates to support their computation and data needs
  • PowerSense: Image Processing for Dengue Detection
    • Leverages microfluidic paper-based analytical devices (μ\muPADs)

Edge Computing

  • Resource-Challenged Environments
    • Less-privileged regions characterized by limited Internet access, limited electricity and network access, and potentially low levels of literacy can leverage surrogates to obtain information to support their communities
  • AgroTempus Features
    • Surrogates in villages download and cache data from mobile hub
    • Surrogates upload field- collected data to the mobile hub which eventually syncs with the cloud

Edge Computing with 5G

  • Edge computing, paired with 5G, revolutionizes industries by bringing computation and data storage closer to data generation.
  • Key Benefits:
    • Better Data Control
    • Reduced Costs
    • Faster Insights & Actions
    • Continuous Operations
  • The Future of Data Processing
  • By 2025, 75%75\% of enterprise data will be processed at the edge, up from just 10%10\% today.

Edge Computing in Advanced Manufacturing

  • Improve supply chain and asset management
  • Orchestrate management from end to end with intelligent video analytics and AI to monitor stock, automate replenishment and more

Edge Computing in Advanced Manufacturing

  • Enable Industry 4.0
  • Gather insight from distributed machinery and manufacturing processes and respond in real time, optimizing production lines and reducing waste

Edge Computing in Advanced Manufacturing

  • Modernize distributed IT
  • Control remote infrastructure with microcloud computing that performs real-time diagnostics and over-the-air updates, reducing costs and disruption

Common Edge Computing Devices

  • Controller in Autonomous Vehicle (BOSCH)
  • Industry Panel PC Machine Controller (OMRON)
  • Nvidia Jetson Nano Development Kit
  • Google Coral Development Board
  • Raspberry Pi 4
  • Intel Movidius Neural Compute Stick
  • Google Coral USB Accelerator Stick
  • SparkFun Edge Development Board

Mobile Device as Edge Computing Device

  • Using TensorFlow Lite to deploy models on edge devices like Android, Linux Embedded devices - Raspberry Pi, ios, and Microcontrollers.

Deployment Example in Advanced Manufacturing

  • Over 77%77\% of manufacturers have already implemented IoT solutions, laying the groundwork for smarter operations.
  • Edge computing enhances existing processes, making them more intelligent and autonomous, driving improved responsiveness and agility.
  • USA 5G Smart Factory: Verizon’s Trial Use Case
  • The first trial involves Verizon’s Sensor Intelligence Solution. An autonomous mobile robot equipped with a camera scans packages to track inventory and location in the warehouse. Using computer vision, the robot transmits barcode and shipping label data over 5G and mobile edge computing to the inventory management system, enabling real-time analytics to optimize logistics.

Deployment Example in Advanced Manufacturing

  • nROK/VTC/MVS GCloT Series Intel Atom E3950 Processor, Intel® Coffee Lake S/Refresh 9th/8th Gen. Core™/Xeon® LGA1151 socket-type CPU, Google Edge TPU inside.

Deployment Example – Google Glass

  • Google Glass at the AI Application Centre is also an edge device. The model is trained with TensorFlow and then converted to TensorFlow Lite to reduce the model size. The final model was then deployed to the Google Glass.

AI FOR ENVIRONMENTAL SUSTAINABILITY

  • Sustainability is the balance between the environment, equity, and economy

Background

  • Singapore’s carbon tax is a key part of our strategy to meet net zero goals. It sends a clear economic signal to reduce carbon-intensive goods and services, hold businesses accountable, and incentivize low-carbon innovations.
  • Timeline of the Carbon Tax:
    • Launched: January 1, 2019 – Southeast Asia’s first carbon pricing scheme.
    • Tax Rate: S55/tCO2e (2019-2023) for a transition period.
    • Future Increases:
      • S2525/tCO2e (2024-2025)
      • S4545/tCO2e (2026-2027)
      • Goal: S50−8050-80/tCO2e by 2030 to strengthen the price signal and drive emissions reductions.

Background

  • Since 2012, the field of AI has reported remarkable progress on a broad range of capabilities including content generation, object recognition, game playing, speech recognition, and machine translation. Much of this progress has been achieved by increasingly large and computationally intensive deep learning models.
  • AI, the most powerful technology of our time, will become a part of every business. McKinsey & Co. estimates AI will add a staggering 1313 trillion to global GDP by 2030 as deployments grow.
  • AI plays an important role in achieving not only environmental but all other sustainable development goals.

How Al Makes Buildings More Sustainable

  • AI's role in making buildings more sustainable.

AI Can Be Used To Reduce Carbon Footprint

  • AI can be a net positive contributor to environmental sustainability in many industries. Here are some examples:
    • In agriculture, AI can transform production by better monitoring and managing environmental conditions and crop yields. AI can help reduce both fertilizer and water, all while improving crop yields. Companies in this sector include Blue River Technology, Harvest CROO Robotics and Trace Genomics.
    • In energy, AI can use deep predictive capabilities and intelligent grid systems to manage the demand and supply of renewable energy. By more accurately predicting weather patterns, AI can optimize efficiency, cutting costs, and unnecessary carbon pollution generation. Companies in this sector include Stem, ClimaCell and Foghorn Systems.

AI Can Be Used To Reduce Carbon Footprint

  • In transportation, AI can help reduce traffic congestion, improve the transport of cargo (supply chain logistics), and enable more and more autonomous driving capability. AI will eventually help with the “last mile” delivery problem and reduce the need for delivery vehicles. AI can help businesses with demand forecasting, helping to reduce the amount of transport needed. Companies in this sector include Nutomony, Nauto and Sea Machines Robotics.
  • In manufacturing, AI can help reduce waste and energy use in production facilities. Robotics can enable better precision. AI can design more efficient systems. Companies in this sector include Drishti, Cognex Corp and Spark Cognition.

AI Can Be Used To Reduce Carbon Footprint

  • In facilities management, AI can help recycle heat within buildings and maximize the efficiency of heating and cooling. AI can help optimize energy use in buildings by tracking the number of people in a room or predicting the availability of renewable energy sources. Companies in this sector include Aegis AI, IC Realtime and IBM’s Tririga.
  • In water resource management, AI can help reduce or eliminate waste while lowering costs and lessening environmental impact. AI-driven localized weather forecasting will help reduce water usage. Companies in this sector include Innovyze, Kurita Water Industries and Plutoshift.

The Environmental Impact of Training AI Models

  • Did you know that training AI models, particularly for tasks like Natural Language Processing (NLP), can have a significant environmental impact?
  • For instance, training an NLP model can generate emissions equivalent to:
    • 5 times the lifetime emissions of an American car.
    • The emissions from 300 round-trip flights between San Francisco and New York.

Green AI

  • To meet the growing demand for AI while minimizing environmental impact, the concept of Green AI has emerged. Green AI focuses on creating environmentally sustainable AI models by reducing computational costs and carbon emissions.
  • Goal of Green AI: Develop AI that is environmentally friendly—focusing on sustainability through energy-efficient models.
  • Efficient Computation: Both in training and prediction, it’s crucial to maximize computational efficiency and minimize energy consumption.
  • The Role of Model Operation: Unlike research, model operation is a key driver for sustainability in businesses. According to AWS, 90%90\% of a model’s energy consumption occurs during predictions and analysis.
  • Energy-Efficient Hardware: Advances in GPUs and Data Processing Units (DPUs) are helping increase energy efficiency for AI, networking tasks, and high-performance computing (HPC), including simulations in supercomputers and enterprise data centers.

Green AI & Green Computing

  • NVIDIA estimates data centers could save a whopping 19 terawatt-hours of electricity a year if all AI, HPC and networking offloads were run on GPU and DPU accelerators (see the charts below). That’s the equivalent of the energy consumption of 2.9 million passenger cars driven for a year.

Green AI – Best Practice

  • Ecologically Efficient Hardware & Infrastructure Design: Sustainable AI starts with the selection of energy-efficient hardware and designing optimized infrastructure. Key considerations include:
    • Data Center Efficiency: Ensuring minimal energy consumption beyond core calculations.
    • Specialized Hardware: Using purpose-built hardware for machine learning tasks to reduce energy consumption.
  • Model Selection: The complexity of the chosen model directly impacts energy use.
    • More complex models require more data and longer training times.
    • Simplicity is key: Choosing simpler models and leveraging research-backed techniques can improve sustainability.
    • Optimizing model selection and choosing the right programming libraries can significantly reduce resource use.

Green AI – Best Practice

  • Model Training & Optimization: Hyperparameter tuning is crucial but can be resource-intensive, often consuming half of the training time.
    • Efficient Training: Streamlining training processes to find optimal models faster reduces the energy footprint.
    • Research-based optimization strategies, including meta-features, can improve model efficiency during training.
  • Model Operation: In production, the efficiency of model operation is vital.
    • Optimized Execution: Fine-tuning models can cut prediction energy consumption by up to 83%83\%.
    • Smart Monitoring: Monitoring input data and model performance helps reduce unnecessary retraining, lowering energy consumption.

Green AI – Best Practice

  • Awareness is Key: Sustainability begins with awareness. Tracking energy consumption and CO₂ emissions during AI development raises visibility and encourages more responsible practices.
  • Structured Responsibility: Establishing organizational frameworks and standards ensures that sustainability is not left solely to individual developers.
    • These frameworks promote consistent implementation of eco-friendly practices across teams and projects.
    • Responsibility becomes shared and systematic, leading to more effective and scalable outcomes.

Conclusion

  • We’ve witnessed tremendous progress in artificial intelligence in recent years.
  • However, when it comes to sustainability, we are still at the beginning of the journey.
  • To move forward, it's essential to adopt the best systems, methods, and practices for building energy-efficient AI models.
  • With the right strategic approach and technology guidance, we can develop AI solutions that are not only powerful but also environmentally sustainable.