1. Introduction

Industry 4.0 is a new platform for modern technologies^[1] in sensors-driven factories. One of the key components of Industry 4.0 is a smart factory, otherwise also a smart or digital factory. A smart factory represents the future state of fully interconnected production systems, without a significant amount of manpower^[2]. All elements of the intelligent factory are interconnected, exchanging information, and recognising and evaluating situations. Thus, physical and cyber technology is integrated, which results in improved controllability, control, transparency of production processes, maximises value for the customer, and in addition, there is communication between the factory and the market itself^[3].

The core technologies of Industry 4.0 include intelligent sensors, IoT, cloud computing, and high-volume data analysis. IoT represents the integration of sensors and computer technology in the field of wireless communication, and cloud services allow access to the network respectively, as a shared pool of computing resources. The combination of these technologies allows the involvement of all devices in the concept of a smart factory, but the collection of huge amounts of data requires another technology, which is the analysis of high-volume data. With the help of analytical tools—data mining or machine learning—this technology is one of the most important elements of the entire concept of Industry 4.0^[4]. In the modern concept, traditional centrally controlled processes are replaced by decentralised control, which builds on the ability of individual elements of a smart factory to communicate with each other. In self-regulatory production, people, machines, equipment, and products communicate with each other ^[5]via intelligent sensors.

2. Results

The factories of the future combine the efficiency of mass production with custom production and optimize the supply chain in real-time thanks to the Internet connection^[6]. These factories handle fluctuations in demand, which are fully automated and fault-durable. The smart factory is connected to the Internet, however it has advanced security against cyberattacks that would jeopardize production.

We summarised the intelligent sensor advantages in Figure 1 based on Reference^[7]. Sensitivity is defined as the relation unit change between output and input. Smart sensors such as IoT devices are wireless, using the internet and usually cloud. Intelligent sensors have low power consumption, automatic diagnostics, calibration, and the ability to process and share data in real-time. Robust means good durable material, solid welds, seals, potting, chemical compatibility, secured wires, and other situational protection. Automatic diagnostics are related to the possibility for making decisions or proceeding action-based actions for control. Some authors emphasise low cost as a feature for smart sensors, but we think that it depends on user experience and sensor value added.

Figure 1. Main characteristics of intelligent sensors.

The transformation of a traditional factory into a smart one brings with it a higher integration of physical production with digital technologies. Sensors and actuators bring factory communication capabilities and data collection and analysis capabilities^[8]. The intelligent factory brings a change from traditional automation to a fully connected and flexible system that can use a continuous flow of data from connecting operations and production systems to learn and adapt to new requirements. The production system in smart factory is different—with more resources for small-lot products, dynamic routing of production line, comprehensive connections with high-speed network infrastructure, deep convergence of physical and digital world (digital twins), self-organisation control system, and big data analytics^[9]. A flexible conveying system of the production lines is designed for the main production loops (cycles), with storage loops on the production line and branch loops for customizing products. The smart factory can integrate data from corporate assets to manage production, maintenance, inventory tracking, digitize operations through the digital twin, and other technologies. In the enterprise infrastructure, smart logistics, smart grids, smart buildings, and smart distribution are interconnected. Project management is important for the successful implementation and sustainability of these systems in smart factories^[10].

Due to the frequent occurrence of extraordinary situations caused mainly by external elements, there is a need to deploy more demanding control systems. Management in smart factories is decentralised. Decentralisations can offer the ability to make decisions at process locations, independent of any central entity^[11]. The complexity of these environments with many simultaneous processes, most of which follow each other, as well as the enormous amount of data that sensors generate in production, can no longer be served by existing control systems based on the simple technology of recording or processing transactions. Therefore, multi-agent systems come to the light, where intelligent information agents form a network of decentralised and distributed intelligence^[12][13]. Beside the existing solutions, these systems are not based on centralised control but are capable of collective self-configuration. These systems interconnect individual autonomous agents (or their digital twins) to communicate, coordinate, and cooperate to achieve a set common goal. Individual communication elements collect data as needed, which they later use to improve and optimize production.

In smart factories, thanks to intelligent sensors, each product actively participates in the production process. The components to be processed contain digital information on how to process them. They, therefore, communicate directly with robots and production machines. With the help of a chip with radio frequency identification or other sensor technology, it can control its production flow. A smart product has access to knowledge related to its structure, composition, or purpose^[14]. On the other hand, thanks to this connection, the customer uses the user interface and intervenes in production in real-time. The sensors allow the customer to obtain information for creating the product specification, and its adjustment according to needs and requirements^[15]. Autonomous vehicles powered by electricity are also connected to the system, ensuring the transport of stock and final products around the factory. Vehicle control is provided by a sophisticated system of sensors. Parts, materials, and components needed for production are transported to the factory when they are really needed for production (Just-in-Time system). Sensors and possibly drones constantly check stock in a smart factory^[16].

2. Conclusions

The fourth industrial revolution is permeating the industry, enabling an increasing number of enterprises to have an incomparably greater overview of their production and maintenance activities than ever before. The deployment of highly reliable and low-maintenance devices contributes to the precise planning of production capacity and equipment’s associated maintenance.

The number of papers discussing the key terms sensors, smart factories, and preventive maintenance increased over time, mostly in the last years. We found that the contemporary burst trend is related to Industry 4.0 technology. Predictive maintenance, smart factories, and intelligent sensors publications, together concerned topics mainly related to deep machine learning, Internet of Things, and big data analytics. The maintenance process in smart factories is based on digitisation, data-driven manufacturing, agent-based systems, and digital twins. Intelligent sensors in such factories use edge, fog, and deep learning methods for control of manufacturing processes. In the future, Internet and blockchain will be important for predictive maintenance.

Four different types of maintenance used in smart factories—Industry 4.0 for predictive maintenance, smart manufacturing for condition-based maintenance, fault diagnosis for maintenance and prognostics, and remaining useful life analysis. The importance of predictive maintenance is also growing due to the growing number of robots, digitisation, and artificial intelligence introduced into production lines to automate routine activities.

We can state that the three types of sensors are mainly used for predictive maintenance in smart factories. Firstly, intelligent sensors which have the potential to connect to higher-level systems. Furthermore, there is a possibility to connect these intelligent sensors to the internet—to build up the IoT devices. Finally, we can use the gathered data in cloud-related technologies. The most prevalent methods used for collecting and monitoring machines and devices are vibration analysis^[17], SCADA systems, CNC machine sensors, and PLCs. Based on the deep analysis, we conclude that the current trend, insights, and future research issues are characterised by:

• Usage of multisource wireless networks of sensors in predictive maintenance.
• Dominance of vibration and temperature sensors for predictive maintenance.
• Challenges of the big data analytics and deep learning techniques.
• Challenges of interoperability of multiple sensors and maintenance systems.
• Decentralisation of maintenance control systems.
• High potential of virtual sensors and nanosensors for the future.
• Challenge of availability and reconfigurability of sensors.
• Security and safety of sensor data.

Based on the results synthesis, we proposed the Smart and Intelligent Predictive Maintenance (SIPM) system for smart factory concerning four major subsystems: production, monitoring, planning, and maintenance. These subsystems communicate and collaborate through modern IoT and cloud-based technologies. Their main advantage is real-time management and planning to reduce the economic costs caused by production downtime^[18].

From a managerial point of view, the predictive maintenance system in smart factories is an early warning, especially in high-risk industries. The ability to detect weak signals of potentially significant strategic impact is a welcome positive in a turbulent business environment. The system of predictive maintenance does not reduce the responsibility or the possibility of personal development of employees, but it must be stimulated by responsible managers. It offers the possibility to reduce the number of hierarchical levels between managers and ordinary employees, so that you can bring about higher employee autonomy and help other innovation processes to be implemented effectively. The challenge for managers today is to select criteria based on which they will be able to select intelligent sensors for smart factories. There is a wide range of sensors on the market and the authors most often state the criteria of sensor sensitivity, sensor cost, flexibility, and size (miniaturity).

Further research may comprise the advanced machine learning methods as deep learning, data-driven algorithms. The new concept called “Machine as a Service” (MaaS) takes over the software as a service (SaaS) model and is another trend suitable for future research. An interesting direction of future research concerns building performance model evaluation related to the reasonable cost. The cost/benefit analysis of using preventive tools in contrast to sustainability requirements is challenging for research.

This entry is adapted from the peer-reviewed paper 10.3390/s21041470