1 Introduction
In this module we will explore Industrial Data Science, a term that is not widely used, defined or standardized (yet).
Generally, it refers to the application of data science techniques within industrial and business domains. This includes optimizing processes, predictive maintenance, quality control, supply chain management and automation by leveraging large volumes of sensor and process data.
The term Industrial Artificial Intelligence (Industrial AI) is more common in literature and appears to be used interchangeably with the fuzzy definition of Industrial Data Science. According to Peres et al. (2020), Industrial AI distinguishes itself within the field of AI in five dimensions:
Algorithms: Combines physical, digital, and heuristic models, requiring complex management and deployment.
Data: Often involves large volume diverse datasets from multiple sources.
Decision-making: Demands high reliability, low error tolerance, and efficient handling of uncertainty.
Infrastructure: Large emphasis on real-time processing capabilities.
Objectives: Targets value creation, such as reducing waste, improving quality, and boosting operational efficiency.
Unlike typical courses that use clean, polished datasets, in this lecture we will get our hands dirty and work with (synthetic) raw machine data with the ultimate goal of setting up a data science product on a simulated machine using live data.
In Chapter 1 to Chapter 3, we will start with an introduction to Industrial Data Science, its significance, its history, and its applications.
In Chapter 4 to Chapter 8, we will prepare ourselves by exploring curated datasets of machine sensors.
Chapter 9 to Chapter 16 provide an overview of time series data, including common applications and methods.
In Chapter 17 to Chapter 23, we will work directly with simple simulated machines.
Finally, Chapter 24 to Chapter 28 will give a brief outlook on selected topics in Industrial Data Science.
Please note that Industrial Data Science is a vast and rapidly evolving field, encompassing a wide range of industries, technologies, and specialized methods. Each industrial context presents unique challenges and data architectures. Covering more of the relevant tools, algorithms, and real-world scenarios in depth would require far more time and resources than a single module allows. The focus of this material, therefore, is to introduce key concepts and practical skills, laying a foundation for further exploration.