Big Data Analytics and Machine Learning for Industrial Cyber-Physical Systems

The objective of this chapter is to illustrate, through the concepts of big data and machine learning, how it is possible to exploit the immense mass of data that can be capitalized at the level of the cyber layer of ICPSs.

Indeed, an ICPS is more than the networking and exploitation of information technology. Information and knowledge are embedded in objects within their physical part and are connected to their cyber part. By integrating perception, communication, learning, behavior generation, and reasoning into these systems, a new generation of intelligent and autonomous systems can then be developed. Industry 4.0 technologies used in production are an example of this. Linked to cyber-physical systems, the Internet of Things (IoT) and cloud computing, they can generate benefits from a circular economy perspective, as they make it possible for circularity to be designed from information gathered from customers as well as across the entire production process.

Regardless of the application context of ICPSs, the component concept is a central concept. It serves as a model for representing the properties of an ICPS, for example, real objects in a production environment connected to virtual objects and processes. A component of an ICPS can be a production system, a transportation system, a piece of equipment, an individual machine, or an assembly within a machine. Some of the fundamental concepts of ICPS can be traced back to research and technologies related to sensor nodes and sensor networks. A sensor node integrates sensors, actuators, computing elements (a processor, a memory, etc.), communication modules, and a battery. The sensor network interconnects many small sensor nodes via a wireless or wired connection. Referred to as wireless sensor networks (WSNs), a large number of sensor nodes equipped with a wireless network connection can be deployed in the physical phenomenon environment. These sensor nodes can provide raw data to the nodes responsible for data fusion or they can process the raw data using their computational capabilities and relay the required part to other sensor nodes.

Holistic view of CPS. 

The main objective of the implementation of new technologies is related to the effective and efficient customer-oriented adaptation of products (and thus production) and services in order to increase the added value for industrialists, raising their competitive position, while improving customer satisfaction and loyalty. To achieve this goal, companies producing goods and services need to develop and manage new knowledge that is crucial for the organization’s decision-making process and the achievement of related business objectives. Therefore, the data generated by ICPS needs to be analyzed and contextualized, to make it relevant sources of information. In the context of the factory of the future, this makes ICPSs, and in particular their cyber part in charge of capitalizing these masses of data, a source of information integrating, often implicitly, relationships on the environment and the business domain. This information and relationships constitute a potential source of knowledge that must be extracted, formalized, and reused.

The authors of this chapter are : Yasamin Eslami Mario Lezoche Philippe Thomas

Page available in
  • Français