Industrial Cyber-Physical Systems

Level 7 MSc Applied Artificial Intelligence

Semester 2 - 33 Hours (Lectures + Tutorials)

Principal Lecturer: Dr Bugra Alkan

Aim

This module introduces Industrial Cyber-Physical Systems (ICPS) and Industrial Internet-of-Things (IIoT) analytics, with a focus on time-series data from industrial systems. It covers architectures for ICPS and smart factories, industrial networking and communication (including OPC-UA), IIoT data pipelines, time-series pre-processing, classical forecasting, feature-based machine learning, deep sequential models and anomaly detection. The emphasis is on practical data analytics in Python for industrial time-series, including case studies drawn from robotics and smart manufacturing.

Lecture Topics

L1 Introduction to ICPS and IIoT Analytics ›

Motivation and module overview; from traditional automation to Industry 4.0; Industrial Cyber-Physical Systems (ICPS) at a glance; data, information and value in smart factories; informal ICPS and IIoT architectures; module structure and learning outcomes; Colab environment check and a first industrial time series; summary and exercises.

L2 Cyber-Physical Systems and Smart Factories ›

Definitions and characteristics of cyber-physical systems; smart factories and cyber-physical production systems; architectural views of ICPS; data flows in CPS and typical time-series characteristics (sampling, delays, noise, multivariate tags); safety, reliability and security considerations in industrial settings; summary and exercises.

L3 Internet-of-Things Connectivity and Networking ›

Networking basics relevant to ICPS (latency, bandwidth, reliability); Quality of Service (QoS) in industrial IoT; wired and wireless connectivity technologies (fieldbuses, Ethernet-based solutions, Wi-Fi, LPWAN, 5G); edge, fog and cloud in industrial networking; positioning computation along the edge–cloud continuum; M2M protocols and their suitability; summary and exercises.

L4 Industrial Communication and OPC-UA ›

Overview of the industrial communication landscape; from OPC Classic to OPC-UA; concepts and design goals; OPC-UA address space & information modelling; services, subscriptions & data access; OPC-UA security model; OPC-UA as a data backbone for IIoT analytics; summary and exercises.

L5 Introduction to IIoT Processing and Analytics ›

Motivation for analytics in IIoT and ICPS; types of data (tags, events, logs); monitoring, forecasting, predictive maintenance, optimisation; IIoT data pipelines and storage (historians, TSDBs, data lakes); time-series viewpoint on industrial tags; summary and exercises.

L6 Time Series Pre-processing for IIoT ›

Irregular sampling, bursts, drifts; aligning & resampling signals; missing data handling; outlier detection; smoothing & denoising; scaling; constructing windows; time-based splits; summary and exercises.

L7 Classical Time Series Forecasting Methods ›

Trend/seasonality decomposition; exponential smoothing; AR/MA/ARMA/ARIMA; SARIMA; comparison with ML/deep learning forecasting; summary and exercises.

L8 Feature-based Machine Learning for IIoT Forecasting ›

Turning time series into supervised learning; feature engineering; linear models & regularisation; tree-based forecasting (RF, GBT); evaluation; Colab workflow; summary and exercises.

L9 Deep Sequential Models for IIoT Forecasting ›

RNNs, LSTMs, GRUs; TCNs and 1D CNNs; sequence model training & regularisation; multi-step forecasting; summary and exercises.

L10 Anomaly Detection in ICPS and IIoT Time Series ›

Point, contextual & collective anomalies; forecasting-residual-based anomaly detection; Isolation Forest; autoencoders; thresholding; evaluation; Colab workflow; summary and exercises.

L11 Project Presentations: ICPS and IIoT Analytics in Practice ›

Mini-project presentations using Python/Colab; full pipelines from raw data to forecasting/anomaly detection; visualisation & interpretation; peer feedback & reflection.

Learning Outcome

At the end of the course students should:
  • Understand the role of ICPS, smart factories and IIoT in modern industrial systems
  • Be able to describe and compare ICPS/IIoT architectures, networking options and M2M protocols
  • Understand the purpose of OPC-UA and its role in industrial communication and analytics
  • Be able to design and implement IIoT analytics pipelines for time-series data in Python
  • Understand and apply key pre-processing techniques for industrial time series (alignment, cleaning, resampling, windowing, scaling)
  • Be familiar with classical time-series forecasting methods (exponential smoothing, ARIMA, SARIMA) and know when to use them
  • Be able to engineer features and train machine-learning models (linear and tree-based) for IIoT forecasting and evaluate them properly
  • Understand the principles behind deep sequential models (RNNs, LSTMs, GRUs, TCNs) and apply them to industrial forecasting problems
  • Be able to implement and compare different approaches to anomaly detection in ICPS/IIoT data
  • Appreciate practical safety, reliability and security considerations in deploying analytics in industrial and robotic systems

Recommended reading

At the end of the course students should:
  • Lee, E. A. Introduction to Embedded Systems: A Cyber-Physical Systems Approach (for CPS foundations).
  • Hermann, M., Pentek, T. & Otto, B. "Design principles for Industrie 4.0 scenarios" (Industry 4.0 / smart factories).
  • Hyndman, R. J. & Athanasopoulos, G. Forecasting: Principles and Practice (free online; classical forecasting).
  • Bagnall, A. et al. selected papers/tutorials on time-series classification and forecasting.
  • Chollet, F. Deep Learning with Python (selected chapters on sequence models).