IB MET L9: AI in Manufacturing

What is the difference between data analytics, machine learning, and artificial intelligence?

Data analytics

The science of studying data to uncover and interpret hidden patterns and trends.

Includes:

• Descriptive analytics

• Predictive analytics

• Prescriptive analytics

• Statistics

• Visualisation

• Data/graph mining

Machine learning (ML)

A subset of AI where systems learn patterns and relationships from data to make predictions or decisions automatically.

Artificial intelligence (AI)

A broader concept involving systems capable of:

• Intelligent decisions

• Situational/contextual awareness

• Automated reasoning and action

AI may include:

• Machine learning

• Rule-based systems

• Optimisation and automation

👉 Key idea:
Data analytics extracts insight from data, machine learning learns patterns from data, and AI is the broader goal of creating intelligent decision-making systems

Types of Analytics

Descriptive analytics

Analyses historical data to understand:

• What happened

• Patterns and trends in past performance

Predictive analytics

Uses data and models to estimate:

• What is likely to happen in the future

Prescriptive analytics

Uses optimisation and decision models to determine:

• What actions should be taken to achieve desired outcomes

👉 Key idea:
Descriptive explains the past, predictive estimates the future, and prescriptive recommends actions (All relies on business domain knowledge)

Why can data science methods be applied across very different systems?

Data science works by:

• Converting complex systems into abstract datasets and relationships

Very different systems can share similar underlying patterns, such as:

• Networks

• Connections

• Clustering

• Flows

• Dependencies

Examples:

• Global automotive industry networks

• Italian mafia networks

Because these systems share structural patterns, similar data science methods can be applied to both.

Highly connected hubs are especially important:

• If key hubs or routes fail (e.g. major ports, the Strait of Hormuz, Suez Canal), disruptions can cascade through the network

• This can create global shortages and widespread supply chain disruption

👉 Key idea:
Data science reveals common network structures across different systems, helping analyse vulnerabilities, cascading failures, and critical hubs in global industries

How is data science applied in supply chains?

Descriptive analytics

Finds hidden patterns in supply chain networks using:

• Graph mining

• Network analysis

Predictive analytics

Predicts future behaviour such as:

• Supplier links

• Delivery delays

• Supply disruptions

Prescriptive analytics

Uses optimisation and autonomous algorithms to:

• Allocate safety stock

• Schedule production/delivery

• Improve resilience and performance

Applications include:

• Automotive

• Aerospace

• FMCG industries

👉 Key idea:
Supply chain data science progresses from understanding patterns, to predicting disruptions, to optimising and automating decisions

What is Artificial Intelligence (AI)?

AI (Artificial Intelligence), also called:

• Machine Intelligence

• Computational Intelligence

…describes systems that mimic or enhance human cognitive functions such as:

• Learning

• Reasoning

• Problem solving

• Decision making

AI involves:

• Intelligent decision-making capability

• Autonomous actuation of decisions

Unlike procedural programming:

• AI can exhibit emergent behaviour and learn from data

AI often uses data analytics methods, but:

• Data analytics also includes non-AI approaches such as:

  • Mathematical optimisation

  • Simulation

👉 Key idea:
AI goes beyond analysing data by enabling systems to learn, reason, and autonomously make intelligent decisions

What are common types of AI methods?

• Logic and reasoning

  • Knowledge representation, expert systems, problem solving

• Classification

  • Assigning observations to categories

• Clustering

  • Grouping similar observations to discover relationships

• Regression

  • Modelling relationships between variables for prediction

• Prediction

  • Using historical/current data to estimate future outcomes

• Search and optimisation

  • Finding the best solution from feasible alternatives

• Generative AI

  • Creating new content such as text, images, or code

• Intelligent agents

  • Autonomous systems that pursue goals using available tools and AI methods

👉 Key idea:
AI consists of many overlapping methods for reasoning, prediction, optimisation, pattern recognition, content generation, and autonomous decision-making

Why is data important in manufacturing and where does it come from?

Data supports manufacturing decision-making across:

• Suppliers and supply chains

• Production planning

• Quality control

• Maintenance

• Sales and customer behaviour

Typical questions include:

• Which suppliers should we choose?

• What products should we make today?

• Are there quality issues?

• Do machines need maintenance?

• How are products used by customers?

Common data sources Structured data

• ERP/MRP systems

• SPC data

• Maintenance logs

• Sales data

• Warranty claims

Typically:

• Lower volume/velocity/variety

Contemporary unstructured data

• Machine sensors

• Smart products/telematics

• Social media

• CRM systems

• Weather data

Typically:

• High volume

• High velocity

• High variety

👉 Key idea:
Manufacturing increasingly relies on large, diverse data sources to improve operational, supply chain, and business decision-making

What are Industry 4.0 and Industry 5.0?

Industry 4.0

The Fourth Industrial Revolution refers to:

• Automation and data exchange in manufacturing

It combines:

• IoT (internet of Things)

• Cloud computing

• Sensors

• AI and data analytics

A key feature is integration across:

• The business

• The product lifecycle

• The supply chain

This improves:

• Productivity

• Flexibility

• Quality control

• Mass customisation

Industry 5.0

Builds on Industry 4.0 but focuses more on:

• Human-centred systems

• Sustainability

• Resilience

👉 Key idea:
Industry 4.0 integrates data and automation across businesses, product lifecycles, and supply chains, while Industry 5.0 extends this toward sustainable and human-focused manufacturing systems

How is AI used in engineering design?

AI supports engineering design through:

• Generative design for optimising structures under simulated loads

• Automating CAD tasks such as topology optimisation

• Predicting product performance using trained simulation models

Applications occur early in the product lifecycle to improve:

• Structural efficiency

• Design speed

• Performance prediction

👉 Key idea:
AI enhances engineering design by automating optimisation and predicting product performance before physical manufacturing begins

How is AI used inside factories?

AI in factories is used for:

• Automated quality inspection

• Buffer and workflow optimisation

• Closed-loop process control

• In-situ monitoring of processes such as 3D printing and laser deposition

AI also improves:

• Safety and ergonomics

Examples:

• Detecting unsafe worker behaviour

• Monitoring robotic workcells

• Identifying hazardous shifts/processes

• Smart helmets and wearable devices for hazard detection and strain reduction

👉 Key idea:
AI improves factory performance through intelligent monitoring, automation, optimisation, safety management, and adaptive process control

How is AI used in supply chains and logistics?

Descriptive AI

Used to identify patterns and relationships:

• Classification

• Clustering

Applications:

• Detecting fraudulent transactions

• Categorising customers, suppliers, orders, and demand

Predictive AI

Used to forecast future outcomes:

• Demand prediction

• Disruption prediction

• Product quality prediction

Often uses:

• Probabilistic machine learning methods

Prescriptive AI

Used to recommend or automate decisions:

• Route optimisation

• Supply chain configuration

• Inventory optimisation

• Facility location decisions

Uses:

• Optimisation

• Logic/expert systems

• Intelligent agents

Intelligent agents

Autonomous systems that:

• Model and simulate supply chains

• Coordinate decisions and actions

• Automate low-level operations

Example:

• Public transport systems in London can be viewed as logistics networks distributing people to destinations efficiently using prediction, scheduling, routing, and optimisation methods.

👉 Key idea:
AI helps supply chains and logistics systems understand patterns, predict future behaviour, and optimise movement of goods, resources, or even people through complex networks

How is AI used for supply link and disruption prediction?

AI systems combine:

• Supply chain mining

• Natural language processing (NLP)

• Event monitoring

• Link prediction algorithms

• Disruption prediction algorithms

To estimate:

• Which suppliers are connected

• Which parts/programs may be affected

• Disruption pathways through the supply network

• Confidence/probability of disruption impacts

• Delivery delays to factories

Predictive problems are often decomposed into smaller models which are then integrated together, causing uncertainty propagation through the system.

👉 Key idea:
AI predicts supplier relationships and disruption risks by combining multiple predictive models and data sources across the supply network

How can AI support collaborative logistics?

Problem:

• Logistics companies may not want a central system controlling routes (“lock-in”)

• Limited visibility between companies

• Large-scale optimisation becomes computationally difficult

• Companies lack incentives to collaborate

AI solution:

• Multi-agent systems automate negotiation (“chatter”) between companies

• Distributed AI avoids central control

• Reinforcement learning (RL) reduces optimisation complexity

• Agents negotiate and coordinate route/resource sharing automatically

Example:

• Truck carriers sharing excess transport capacity without a central scheduler

👉 Key idea:
AI enables distributed, collaborative logistics where autonomous agents coordinate decisions and optimise resource sharing without centralised control

How is AI used in service and maintenance?

AI combined with sensors enables monitoring of products after sale to:

• Optimise maintenance operations

• Improve servicing efficiency

• Support future engineering design improvements

Applications include:

• Predictive maintenance

• Service scheduling

• Spare parts planning

👉 Key idea:
AI extends product lifecycle integration by using operational product data to improve maintenance and future product design.

What is Condition-Based Maintenance (CBM)?

Traditional maintenance:

• Uses periodic inspections

• Can be expensive and inefficient

CBM uses:

• Sensors

• Data analytics and AI

…to monitor equipment condition continuously and predict failures.

Benefits:

• Predict remaining useful life (RUL)

• Optimise maintenance timing

• Reduce unnecessary inspections

• Order spare parts in advance

• Improve next-generation designs

Typical applications:

• Aircraft engines

• Wind turbines

• Heavy machinery

👉 Key idea:
CBM replaces fixed maintenance schedules with data-driven maintenance based on actual equipment condition.

How is AI used for spare parts and safety stock optimisation?

AI predicts:

• Component failures

• Fleet remaining useful life

This helps optimise:

• Spare parts inventory

• Logistics and overhaul planning

• Safety stock levels

Reinforcement learning (RL) can:

• Learn optimal inventory policies across complex supply networks

• Reduce stock shortages and lead times

Trade-offs:

• AI methods may outperform analytical models

• But can be slower and less explainable

👉 Key idea:
AI helps balance inventory cost, lead time, and maintenance reliability across large engineering systems.

How are intelligent agents used in maintenance logistics?

Multi-agent systems use autonomous software agents to coordinate:

• Spare part requests

• Supplier searches

• Auctions and negotiations

• Batch ordering and contracts

Agents communicate and make decisions without a central controller.

Benefits:

• Reduces bottlenecks

• Improves coordination across organisations

• Enables distributed decision-making

👉 Key idea:
Agent-based AI automates coordination and communication in complex maintenance and logistics networks

What are key risks of AI in manufacturing?

AI systems can fail due to:

• Poor data quality

• Overfitting

• Changing real-world behaviour

• Incorrect assumptions or spurious correlations

Example:

• Google Flu Trends overestimated flu prevalence because search behaviour changed over time.

Manufacturing-specific challenges:

• Low sample sizes

• Discontinuous datasets

• Complex interactions between batches/processes

• Harsh factory environments

• Connectivity and deployment difficulties

• Cost and practicality concerns

Potential solutions:

• Synthetic data generation

• Anomaly detection methods

👉 Key idea:
AI performance depends heavily on high-quality data, robust models, and practical deployment conditions.

What ethical concerns exist with AI in manufacturing?

Key concerns include:

• Worker monitoring and surveillance

• Algorithmic bias

• Workforce displacement

• Environmental responsibility

Regulation is evolving:

• Example: EU AI Act (2024)

  • Restricts manipulative AI

  • Prohibits emotion recognition in some contexts

  • Classifies worker monitoring as high-risk

👉 Key idea:
AI in manufacturing creates ethical and regulatory challenges around privacy, fairness, safety, and human employment