Like anything that’s new, the terms “digital manufacturing”, “Industry 4.0”, “smart factory” are often confused, but in fact, all of these concepts represent the currents of a single trend. I mean the automation of individual machines and separate processes has evolved into full integration of all elements into a single digital ecosystem which involves both manufacturers and their partners. Vertical chains are being built up – from the primary processing of resources to the delivery of goods and the provision of goods-based services.
“Smart manufacturing” differs from, so to speak, “unintelligent” one not by the presence of IoT sensors but by the presence of information systems capable of processing intelligent data from these sensors. As of today, the informatization of manufacturing is slowly but gradually penetrating enterprises across the globe. This is the difference, for example, from telemetry sensors, which were the first IoT devices employed at factories. Back then, those sensors did not always have an access to the Internet – instead, they communicated only over a local network. The term “smart manufacturing” stems from the general trend of digitizing the entire surrounding world (digitizing photos, maps or communications is the same idea). The general plan is to digitize the activities within an enterprise and move from paper workflow to modern digital means of design, production, control and sale. A “smart factory” provides support tools for people and machines which help them carrying out their tasks through the analysis of context-sensitive information. For example, sources of information about orders, products, equipment, available production facilities and technological processes are rarely in a unified database and format.
However, there’s a solution. The idea of a “virtual enterprise” (a so-called “digital twin”) offers a framework that links all this information together, providing a mirror of a real enterprise, and thus paves the way for more innovative production prototypes, optimizing the assembly line, product design, mass modifications and adaptations in accordance with the requirements of the customer or the market.
The Components Of a Smart Factory
In general, the main components of “smart manufacturing” are:
1) Semantic multimodality supporting the presentation of various information occurring in a factory context – for example, attribute trees, relational, sensory, tabular, graph, and entity data.
2) Multi-dimensionality. Information on several dimensions should be presented and recorded. For example, the description of business processes and technological operations performed at various levels of the structure of digital production, the study of spatial hierarchies and assets from the geographical point of view, analysis of the life cycle of equipment and production processes.
3) Multi-granularity. The access to data from sensors and equipment contributes to the assessment of the technological process and, if necessary, provides the possibility of its regulation and control. This way, an organization gets a comprehensive view of all business units, including hierarchies and responsibilities.
4) Transparency and integration. Currently, data and information are distributed between various systems, such as production automation, quality control, and enterprise resource planning systems. It is important to integrate all relevant information from these systems while maintaining the systemic nature of the records.
Today’s factories have to solve the problem of increasing the flexibility and openness of manufacturing processes. Of course, the invention of flexible manufacturing lines didn’t happen yesterday. However, nowadays there is a clear transition to a new level as it has become possible to get all the necessary information about the state of technological processes, equipment, warehouses, new orders in real time. This way, now it is possible to automate the processing of this information.
Corporations like Cisco, IBM, Huawei and Microsoft have played a big role in the development of technologies for smart manufacturing. They promote their cloud platforms, working with large turnkey customers. AT&T, Cisco, GE, IBM and Intel have even created a consortium to solve the problem of protocol compatibility and generally coordinate the “rules of the game”.
But the spread of the industrial Internet of things (aka IIoT) largely depends on the idea and philosophy of open data: open source, open licenses, open innovations, and so on. They changed the thinking and attitude to information, which led to the formation of common approaches, as well as new standards. The use of a single technology coming from a single corporation is not enough to create a “smart factory”.
The cybersecurity issues are closely related to the paradigm of openness. Nowadays, standard means of protecting information (encrypted protocols, passwords, etc.) have limitations and correlate very poorly with the idea of transparency. New research organizations are working on the development of new approaches to cybersecurity.
The attendants of Internet of Things World Forum annually discuss 200–300 stories from industry representatives about the use of industrial IoT technologies. An excellent illustration of this approach is the AW auto component plant in North Carolina with more than 2,000 employees and a 1.3 million square meters production complex. AW North Carolina produces about 600,000 transmitters annually and shows the highest profitability within the Aisin group. How did they manage to do that? At some point, factory managers admitted that they were working using outdated technologies, and covered the entire production area with Wi-Fi networks. This way, they were able to digitize and process data at each production site, as well as keep the statistics data. One more great example from the automotive industry is provided by another international corporation: IBM has equipped the Indian automobile plant Mahindra & Mahindra with sensor networks and services for data analysis. At the factory, it was possible to identify the workers’ overlapping tasks, establish a more effective interaction between the production units, and begin to automatically identify defects.
All in all, everything depends on the equipment’s features. For example, how does an industrial computer differ from a household one? In a nutshell, computers are the same, but the industrial counterpart is more reliable and has more various specialized interfaces. And, of course, they are much more expensive. For example, Daimler has about 3000 databases, which need to be integrated within the framework of smart manufacturing. For obvious reasons, smart houses do not face such a problem.
The most serious technological difficulties in the formation of IT systems at smart factories are connected with the creation of the unified environment with various data coming from sensor networks. The researchers of smart networks mention two concepts: “smart factory” and “data lakes”. The term “data lake” has appeared in recent years to describe the storage of data sets that are provided for processing and analysis in their original formats. It is often viewed as the opposite concept of a “data warehouse” – this kind of data is available for analysis only after the phase of mandatory data reorganization, which is not required when using data lakes. Another significant difference between the “data lake” and the “data warehouse” is the type of data processing that is performed in each specific case. A “data lake”, in accordance with its definition, contains data that is open to any kind of processing: natural language processing, machine learning, specialized (semi) structured queries, and so on.
The ontology of a new type of factory assumes that employees, equipment, location of assets, their relations with each other should be consolidated into a single digital field, and constantly recorded in a single “account book” to provide managers with a holistic view of a digital enterprise. Operational data — for example, order information or data from equipment sensors — are also included in the enterprise ontology. All this data should be dynamically downloaded from the corresponding databases.
Today, production managers are beginning to understand that modernization measures will cost, but the final effect of automation and the introduction of new IT solutions should not only compensate but bring returns on the funds invested in innovation. The problem is that if the modernization is not completed, the return on investment is unlikely to be high, and possibly negative. But this is not a problem of IoT technologies – this is a problem of a long-term management strategy.