IDARTS – Towards intelligent data analysis and real-time supervision for industry 4.0

Highlights

• Proposes a generic framework for data analysis and real-time supervision for Industry 4.0 predictive manufacturing systems.

• An initial implementation is presented, focusing on the integration of the different modules.

• The implementation is deployed in a 4-node cluster and applied to a simulated test.

Abstract

The manufacturing industry represents a data rich environment, in which larger and larger volumes of data are constantly being generated by its processes. However, only a relatively small portion of it is actually taken advantage of by manufacturers. As such, the proposed Intelligent Data Analysis and Real-Time Supervision (IDARTS) framework presents the guidelines for the implementation of scalable, flexible and pluggable data analysis and real-time supervision systems for manufacturing environments. IDARTS is aligned with the current Industry 4.0 trend, being aimed at allowing manufacturers to translate their data into a business advantage through the integration of a Cyber-Physical System at the edge with cloud computing. It combines distributed data acquisition, machine learning and run-time reasoning to assist in fields such as predictive maintenance and quality control, reducing the impact of disruptive events in production.

Introduction

With the advancements in the Information and Communication Technology field and the exponentially increasing volumes of data being generated every day, a new set of possibilities has been presented to improve the efficiency and the characteristics of production processes. Adding to this is the transformation from a saturated seller’s market into a customer-driven one, with its growing demand for highly customized products accompanied by decreasing product lifecycles and smaller lot sizes, pushing companies towards a paradigm shift in order to leverage their data to attain a business advantage in such a competitive and dynamic market [1].

As such, the currently ongoing 4th industrial revolution, usually referred to as Industry 4.0 in Europe [[2], [3], [4]] and Industrial Internet in the US [5], aims to introduce and take advantage of the interconnected world along the entire value chain, allowing the sharing and processing of the data available in all of the its actors to generate relevant knowledge and optimize the overall process. The adoption of the Industry 4.0 paradigm encompasses the following three characteristics [6]:

• • “Horizontal integration across the entire value network”: By integrating the overall value chain it is possible to optimize it beginning with the suppliers, materials, logistics, etc. In this sense, all of the value chain’s actors must be connected and coordinated among each other based on their individual requirements, creating a very dynamic ecosystem.

  • “End-to-end engineering across the entire product life-cycle”: The integration and digitalization across all phases of the product’s life-cycle is crucial to ensure that data can be collected, stored and processed to generate new knowledge from the product’s inception to its end of life. This knowledge can be particularly relevant for the product’s improvement, not only regarding its production, but also the for instance its design or material suppliers.

  • “Vertical integration and networked manufacturing systems”: At the shop-floor, the integration among the different components and actors (such as resources, humans and Manufacturing Execution Systems) should be done through a Cyber-Physical System (CPS). This system will allow not only the internal integration and optimization but also a harmonized and smooth integration with the two previously presented functionalities.

With the vertical integration emerges the concept of Smart Factory (SF). According to [7] these factories must have some characteristics including among others the capacity to deal with mass customization [4], flexibility [8] and new maintenance strategies [9]. However, one of the main characteristics is the ability to generate new and relevant knowledge based on data processing [10]. As previously stated, SFs are implemented and deployed via CPS. The concept of Cyber-Physical Production System (CPPS) [11] merges the functionalities and benefits of CPS applied to the industrial context. The main objective in a CPPS is to create an abstraction layer where each of the shop-floor’s actors is represented by a cyber entity, as shown in Fig. 1.

The communication between the heterogeneous components is now made at the cyber level, allowing a smooth and effective integration of all the components and actors avoiding the usual problems related to vendors’ specifications and standards. In [12] the authors present a comparison between today’s factories and the now emerging Industry 4.0 based SFs, implemented through CPPS. With all the resources integrated, sharing information and their behaviours among each other, the shop-floor can adapt and organize in runtime to optimize at different levels (production, maintenance, energy consumption, etc). Moreover, with the advances in the Industrial Internet of Things (IIoT) and the increasingly large number of sensors and other data sources available on the shop-floor, the amount of extracted data is growing and the traditional algorithms are no longer able to process these volumes of data. Hence, the big data analysis field is becoming more and more important in several areas as a way to tackle this challenge [13].

This is often coupled with the usage of Machine Learning (ML), allowing manufacturers to obtain insights regarding their factory which would have been otherwise missed. ML can be defined as a system’s capacity to improve its performance on a given task or set of tasks over time based on previous results [14]. Therefore, ML algorithms can be used to predict a system’s behaviour and/or improve its overall performance, enabling the development of tools capable of analysing data and perceive what are the underlying trends and correlations. Thus, ML-based approaches can be used to predict abnormal events (failures, degradation, energy consumption, etc), generate warnings and advise the system and/or the operator regarding which course of action to take, assisting in diagnosis and maintenance tasks [15].

In line with this, the work detailed in the coming sections aims to provide an integration of this real-time and historical analysis with self-improvement mechanisms under a single framework for Predictive Manufacturing Systems (PMS), along with an initial implementation of its core functionalities. It is structured as follows: Section 2 presents a summary of related work that can be found in current literature pertaining to CPPS-based PMS. Section 3 introduces the proposed framework and formalizes some core concepts, requirements and functionalities. Section 4 details the initial implementation of the framework and the integration of its modules. Afterwards, Section 5 describes the tests performed to validate the implemented solution, with the results being discussed in Section 6. This is followed by Section 7 in which a brief summary of the developments is presented, along with some conclusions and remarks regarding future work.