Advances in Sensor Technologies in the Era of Smart Factory and Industry 4.0 †

The evolution of intelligent manufacturing has had a profound and lasting effect on the future of global manufacturing. Industry 4.0 based smart factories merge physical and cyber technologies, making the involved technologies more intricate and accurate; improving the performance, quality, controllability, management, and transparency of manufacturing processes in the era of the internet-of-things (IoT). Advanced low-cost sensor technologies are essential for gathering data and utilizing it for effective performance by manufacturing companies and supply chains. Different types of low power/low cost sensors allow for greatly expanded data collection on different devices across the manufacturing processes. While a lot of research has been carried out with a focus on analyzing the performance, processes, and implementation of smart factories, most firms still lack in-depth insight into the difference between traditional and smart factory systems, as well as the wide set of different sensor technologies associated with Industry 4.0. This paper identifies the different available sensor technologies of Industry 4.0, and identifies the differences between traditional and smart factories. In addition, this paper reviews existing research that has been done on the smart factory; and therefore provides a broad overview of the extant literature on smart factories, summarizes the variations between traditional and smart factories, outlines different types of sensors used in a smart factory, and creates an agenda for future research that encompasses the vigorous evolution of Industry 4.0 based smart factories.


Introduction
With its introduction in Germany in 2011, Industry 4.0 instantly became the focus of a global world that promoted the computerization of manufacturing [1]. Industry 4.0 has revolutionized the manufacturing process, leading to intelligent manufacturing promising self-sufficient manufacturing processes by using machines and devices that communicate with each other through digital connectivity [2,3]. Although most of the design philosophies and technologies of Industry 4.0, such as the internet of things (IoT), cyber-physical systems (CPSs), and artificial intelligence are already in use, most firms lack insight into the wide set of technologies offered by Industry 4.0 that provide devices with seamless connectivity, interoperability, visibility, and intelligence capabilities. in production processes. With enhancing the complexity of cyber physical systems, the physical machines and business processes have combined with automation to give rise to complex optimization decisions that were made traditionally by humans [5,21,22]. This has enabled manufacturers to integrate the floor decisions and perceptions with the supply chain, giving birth to what we now call a smart factory. The introduction of mechanical manufacturing equipment marked the first industrial revolution, followed by the development of the mass production of goods [13]. The digital revolution was considered to be the adoption of increased automation and control in manufacturing processes by using electronics and IT [12]. The adoption of IoT in these processes has given rise to the deviation of the centralized factory system to a decentralized system [12][13][14]. This technology has enabled machines and industries to go through a self-optimization and reconfiguration, to adapt their behavior to changes in orders and operating conditions [23,24]. The core of smart factories is the technology that makes data collection possible. This technology includes intelligent sensors, motors, and robotics which are employed on the production and assembly lines of the manufacturing industry [17].

Smart Factory vs. Traditional Factory
In order to meet drastic changes in customer demands, the manufacturing process requires abilities that help in adjusting product type and production capacity, to enable the handling of multiple product varieties [25]. Manufacturing should have adequate functionality, scalability, and connectivity with customers and suppliers to meet such challenges. Traditional factories lack capabilities that allow them to monitor and control automated and complex manufacturing to enable efficient production of customized products [26]. Traditional factories have stand-alone and segregated applications with less integration of the production system, product life cycle, and value chain. Consequently, there is poor reuse of systems and integration between real and virtual systems in a traditional set-up. A general concept of a smart factory can be seen in Figure 1.
Sensors 2020, 20, x FOR PEER REVIEW 3 of 25 physical machines and business processes have combined with automation to give rise to complex optimization decisions that were made traditionally by humans [5,21,22]. This has enabled manufacturers to integrate the floor decisions and perceptions with the supply chain, giving birth to what we now call a smart factory. The introduction of mechanical manufacturing equipment marked the first industrial revolution, followed by the development of the mass production of goods [13]. The digital revolution was considered to be the adoption of increased automation and control in manufacturing processes by using electronics and IT [12]. The adoption of IoT in these processes has given rise to the deviation of the centralized factory system to a decentralized system [12][13][14]. This technology has enabled machines and industries to go through a self-optimization and reconfiguration, to adapt their behavior to changes in orders and operating conditions [23,24]. The core of smart factories is the technology that makes data collection possible. This technology includes intelligent sensors, motors, and robotics which are employed on the production and assembly lines of the manufacturing industry [17].

Smart Factory vs. Traditional Factory
In order to meet drastic changes in customer demands, the manufacturing process requires abilities that help in adjusting product type and production capacity, to enable the handling of multiple product varieties [25]. Manufacturing should have adequate functionality, scalability, and connectivity with customers and suppliers to meet such challenges. Traditional factories lack capabilities that allow them to monitor and control automated and complex manufacturing to enable efficient production of customized products [26]. Traditional factories have stand-alone and segregated applications with less integration of the production system, product life cycle, and value chain. Consequently, there is poor reuse of systems and integration between real and virtual systems in a traditional set-up. A general concept of a smart factory can be seen in Figure 1. The smart factory is an upgrade from old-fashioned automation to a linked and flexible system, and which constitutes a continual data stream through highly connected operations and production systems which can learn and adapt to changing demands [26][27][28]. These factories can assimilate data from physical, operational, and human assets to drive manufacturing, maintenance, inventory tracking, digitization of operations, and other activities in manufacturing systems [29]. The main aim of smart factories is to use intelligent production systems and suitable engineering methods for the successful and interconnected implementation of production facilities [25,30]. It is an engineering The smart factory is an upgrade from old-fashioned automation to a linked and flexible system, and which constitutes a continual data stream through highly connected operations and production systems which can learn and adapt to changing demands [26][27][28]. These factories can assimilate data from physical, operational, and human assets to drive manufacturing, maintenance, inventory tracking, digitization of operations, and other activities in manufacturing systems [29]. The main aim of smart factories is to use intelligent production systems and suitable engineering methods for the successful and interconnected implementation of production facilities [25,30]. It is an engineering system that operates on interconnection, collaboration, and execution. Interconnecting devices in smart factories allow the exchange of information, recognize and assess situations, and integrate the physical world with the digital world, making smart factories adaptive in nature [23,31]. In other words, the smart factory integrates physical and cyber technologies and makes the involved technologies more accurate, enhancing the performance, quality, controllability, management, and transparency of the manufacturing processes. In such a smart factory environment, the manufacturer has the ability to meet customer requirements by changing the production specifications and other settings of the machines at the last minute. This ability is not present in traditional factories [32] ( Table 1). The true feature of a smart factory lies in its capability to readjust and evolve along with the growing needs of the organization [32,33]. These needs can be categorized into: changing customer demands, emergence of new markets, development of new products and services, enhanced productive approaches to operations, and use of advanced technologies in maintenance processes [13,34,35]. The ability to tailor and learn from real-time data makes smart factories more receptive and predictive, to avoid operational downtime and other possible failures in processes [12,29]. Table 1. Key differences between the traditional manufacturing factory and smart factory.

Traditional Factory Smart Factory
Manual and isolated processes, operations; no integration with different systems and tools.
Digitized and integrated processes, operations; complete integration with existing systems, new systems and tools.
Legacy systems with frequent machine failures and increased maintenance costs.
Smart systems with improved machine utilization and reduced maintenance costs.
Tied to systems or machines for data, therefore zero or limited data for decision making; process-driven decision making.
Update or receive data on the go, therefore complete data for faster decision making; data-driven decision making.
Zero or limited visibility on operations, productivity data.
Increased transparency, visibility on operations and production data.
Limited innovation in production development. Smart and intelligent products.
Inaccurate asset tracking process and poor resource utilization.
Accurate asset tracking using IoT, RFID; improved resource utilization.
Poor interoperability. High interoperability.
The production line is fixed unless manually reconfigured by people with system power down.
When switching between different types of products, the needed resources and the route to link these resources should be reconfigured automatically and online.
A smart factory is characterized by four intelligent features: • Sensors: these are devices that have the ability to self-organize, learn, and maintain environmental information to analyze behaviors and abilities. Therefore, sensors can make decisions that enable them to adjust to changes in the environment. • Interoperability: through interconnection between different devices, coordination between them can be enhanced, allowing flexibility in configuration protocols of the production system. • Integration: robots and artificial intelligence (AI) allow smart factories to have a high level of integration among processes. AI, along with the integration of human intellectual capabilities, enables factories to perform analysis and decision making.

•
Virtual reality (VR) techniques: as one of the high-level components of smart factories, VR facilitates human-machine integration by virtualizing manufacturing processes using computers, signal processing, animation technology, intelligent reasoning, prediction, and simulation and multimedia technologies.

Key Sensing Technologies in a Smart Factory
Smart factories are comprised of intelligent machines, devices, and control equipment that monitor vital parameters of the manufacturing processes [5]. These improvements have not only altered the factory floor infrastructure, promoting steady and precise collaboration between machines, but have also altered machinery requirements, increasing demand for reliable sensors [10]. This section gives brief information about the key sensors used in a smart factory.

Passive Sensors
The current manufacturing system is defined by different technologies, however the main technologies used are sensors, actuators, effectors, controllers, and control loops [36]. A vital role is played by sensors in a smart factory, as they collect and implement accurate data into the manufacturing processes to enhance product quality. Sensors are electrical, opto-electrical, or electronic devices consisting of sensitive materials that help to determine the presence of a particular entity or function [37]. In many cases, a physical stimulus is transformed into an electrical signal using sensors, which then can be evaluated and analyzed for making decisions about the operations being carried out [5,36]. Recent developments in sensor technology have enabled manufacturers to control and acquire data like never before.
Sensors can operate either actively or passively [36,38]. When operating actively, a particular physical stimulus is required for the sensor to work. For example, color identification sensors are active as they need visible light to illuminate the object so that the sensor can receive a physical stimulus. In the passive instance, the physical stimulus is already present and does not have to be provided [38]. For example, infrared devices are passive as the stimulus is already being generated from infrared radiation that is linked with the temperature of a body.
Several types of sensors have been developed and used successfully in industrial process control. Smart factories use a variety of sensor types, from basic temperature to humidity monitoring to sophisticated position and product sensing [39,40]. These sensors make manufacturing efficient by helping in advance factory operations, such as moving products, controlling robotic and milling processes, and sensing environmental factors. The main measurement and control parameters in a factory environment are temperature, position, force, pressure, and flow [40].

Temperature Sensors
As temperature directly affects material properties and product quality, it is one of the crucial parameters to be measured and controlled in industrial plants. A temperature sensor is a device that has the ability to collect temperature concerned information from a resource, and then changes it into information that can be understood by another device [41]. These sensors have the ability to measure the thermal characteristics of gases, liquids, and solids. Several temperature sensors have been developed in recent years which can be used in electrically and chemically hostile environments. These sensors can be divided into two groups: (1) low-temperature sensors, with a range of −100 to +400 • C, using sensing materials such as phosphors, semi-conductors, and liquid crystals; and (2) high-temperature sensors with a range of 500 to 2000 • C, based on blackbody radiations [25,40,42]. Table 2 shows the different sub-types of temperature sensor, along with their key features.

Pressure Sensors
Pressure sensors have the ability to capture pressure changes and transform them into an electrical signal, where the applied pressure defines its quantity. These are electro-mechanical devices that identify force in gases or liquids and provide control signals to display devices [25,37,43]. These sensors can also be used to detect atmospheric changes [51]. For example, barometric pressure sensors have the ability to detect changes in the atmosphere that are helpful for the prediction of weather patterns and changes. Another example is vacuum sensors, which are used when pressure in a vacuum is below atmospheric pressure levels, which can be difficult to detect using mechanical methods. Table 3 shows the different sub-types of pressure sensors along with their key features.

Position Sensors
These sensors are used to sense the positions of valves, doors, throttles etc. These sensors are equipped with location tracking abilities that help to determine the precise positions of work-in-progress, tools, and other production-relevant items within the facility [57,58]. Motion sensors (which trigger actions such as illuminating a floodlight by detecting movement of an object) and proximity sensors (which detect that an object has come within the range of a sensor) are worth mentioning as they serve functions similar to position sensors [59,60]. Table 4 shows the different sub-types of position sensors along with their key features. Sensitive to environmental changes such as humidity, temperature etc.

Magnetostrictive Linear Position
Material changes its size or shape when in the presence of a magnetic field to detect the position of an object.
Ferromagnetic materials such as iron, nickel, and cobalt.
Used in the controlling of gaps between rollers, hydraulic or pneumatic cylinders, in automotive industry and electric actuators.
Non-contact; ability to detect position in the presence of a barrier between magnet and sensing rod; ability to measure multiple magnets with a single sensing rod.
Dead band on each side of the sensor cannot be reduced to zero.

Eddy Current based
Work with induced currents that occur in a conductive material in the presence of a changing magnetic field using Faraday's law of induction.
Conductive material such as copper, aluminum, titanium alloy etc.
Widely used in automation applications, machine tool mounting, final assembly of delicate machinery and monitoring drive shafts.
Functional in dirty environments; less expensive; unaffected by most contaminants.
Omnidirectional, can only determine the distance of the object not the direction of the object from the sensor.

Optical
Work two ways: (1) light is transmitted from an emitter and sent to a receiver at the other end of the sensor; (2) emitted light signal is reflected from the monitored object towards the light source. Change in light characteristics are used to determine the position.
Glass or plastic disc used as an encoder, with LED used as a light source, and a photodetector as light receiver.
Widely used in deadbeat galvanometers, induction motors, induction furnaces, electric brakes, and speedometers.
Both linear and rotational movement can be detected.
Large amount of heat is produced in the soft core of transformers, induction coil, electric motors, etc., reducing the efficiency of these machines.

Force Sensors
Force sensors are designated to translate applied forces (such as tensile, compressive force, etc.) into electric signals which reflect the degree of force [63,64]. These signals are then sent to indicators, controllers, or computers that inform operators about the processes, or serve as inputs that help to achieve control over machinery and processes. A variety of force sensors are being used in smart factories depending on the type of force being measured [65]. For instance, load cells measure compressive forces, strain gauges measure the internal resistance forces, and force sensing resistors measure the rate of change of an applied force. Table 5 shows the different sub-types of force sensors, along with their key features.

Flow Sensors
These sensors have the ability to sense the movement of gases, liquids, or solids within a pipe or a conduit. These sensors have extensive uses in processing industries, and allow operation of the machinery at an optimum performance level [67,68]. A flow sensor can be electronic, using ultrasonic detection of the flow, or partially mechanical [69]. For instance, flow sensors in automobiles measure air intake in the engine and adjust fuel delivery to the fuel injectors in order to provide optimum fuel to the engine. Flow sensors are also used in medical ventilators, where the correct rate of delivery of air and oxygen to patients is needed for respiration [70]. Table 6 shows the different sub-types of flow sensors along with their key features.

Smart Sensors
Among other recent advances in technology, smart sensors have been in the spotlight in terms of their potential significance and wide range of application areas. With the integration of computing and IoT in industrial processes, ordinary sensors have been transformed into smart sensors providing them with abilities to carry out complex calculations with collected data [37,40]. Apart from increased capabilities, smart sensors have also become remarkably small and exceedingly flexible, turning bulky machines into high-tech intel. Equipped with signal conditioning, embedded algorithms, and digital interfaces, smart sensors have become devices with detection and self-awareness capabilities [25,72]. These sensors are built as IoT components that convert real-time information into digital data that can be transmitted to a gateway [36,73]. These abilities allow smart sensors to predict and monitor real time scenarios and take corrective actions in an instant. Complex multi-layered operations such as collecting raw data, adjusting sensitivity, and filtering, motion detection, analysis, and communication are the main functions of intelligent sensors [38]. For instance, wireless sensor networks (WSNs) are one of the applications of smart sensors, whose nodes are connected with one or more other sensors and sensor hubs, making a communication technology of some kind. In addition, information from multiple sensors can be combined to deduce conclusions about an existing problem; for instance, temperature and pressure sensor data can be used to infer the onset of a mechanical failure. Figure 2 shows the building blocks of a typical smart sensor [41], while Figure 3 summarizes the key features of smart sensors [51,72].

Smart Sensors
Among other recent advances in technology, smart sensors have been in the spotlight in terms of their potential significance and wide range of application areas. With the integration of computing and IoT in industrial processes, ordinary sensors have been transformed into smart sensors providing them with abilities to carry out complex calculations with collected data [37,40]. Apart from increased capabilities, smart sensors have also become remarkably small and exceedingly flexible, turning bulky machines into high-tech intel. Equipped with signal conditioning, embedded algorithms, and digital interfaces, smart sensors have become devices with detection and self-awareness capabilities [25,72]. These sensors are built as IoT components that convert real-time information into digital data that can be transmitted to a gateway [36,73]. These abilities allow smart sensors to predict and monitor real time scenarios and take corrective actions in an instant. Complex multi-layered operations such as collecting raw data, adjusting sensitivity, and filtering, motion detection, analysis, and communication are the main functions of intelligent sensors [38]. For instance, wireless sensor networks (WSNs) are one of the applications of smart sensors, whose nodes are connected with one or more other sensors and sensor hubs, making a communication technology of some kind. In addition, information from multiple sensors can be combined to deduce conclusions about an existing problem; for instance, temperature and pressure sensor data can be used to infer the onset of a mechanical failure. Figure 2 shows the building blocks of a typical smart sensor [41], while Figure 3 summarizes the key features of smart sensors [51,72].

Smart Sensors
Among other recent advances in technology, smart sensors have been in the spotlight in terms of their potential significance and wide range of application areas. With the integration of computing and IoT in industrial processes, ordinary sensors have been transformed into smart sensors providing them with abilities to carry out complex calculations with collected data [37,40]. Apart from increased capabilities, smart sensors have also become remarkably small and exceedingly flexible, turning bulky machines into high-tech intel. Equipped with signal conditioning, embedded algorithms, and digital interfaces, smart sensors have become devices with detection and self-awareness capabilities [25,72]. These sensors are built as IoT components that convert real-time information into digital data that can be transmitted to a gateway [36,73]. These abilities allow smart sensors to predict and monitor real time scenarios and take corrective actions in an instant. Complex multi-layered operations such as collecting raw data, adjusting sensitivity, and filtering, motion detection, analysis, and communication are the main functions of intelligent sensors [38]. For instance, wireless sensor networks (WSNs) are one of the applications of smart sensors, whose nodes are connected with one or more other sensors and sensor hubs, making a communication technology of some kind. In addition, information from multiple sensors can be combined to deduce conclusions about an existing problem; for instance, temperature and pressure sensor data can be used to infer the onset of a mechanical failure. Figure 2 shows the building blocks of a typical smart sensor [41], while Figure 3 summarizes the key features of smart sensors [51,72].

Calibration Capability
The ability of a sensor to determine its normal function is termed calibration capability [39]. Self-calibration is simple in many cases, and different calibration techniques are available for different types of sensors: • Sensors with an electric output carry out calibration by using a known reference of voltage level. • Sensors such as load cells used for weighing systems can adjust their output to zero when no force is being applied [27].

•
Other sensors can use look-up tables for calibration. However, to carry out calibration using look-up tables, a huge amount of memory capacity must be available to store correction points due to the large volume of data gathered during the process. On the contrary, an interpolation method is preferable in which a small matrix of correction points is required [35].

Self-Diagnosis of Faults
Smart sensors carry out self-diagnosis by observing internal signals for evidence of faults. Differentiating between normal measurement deviations and sensor faults can be a challenge for some sensors [72]. This challenge is overcome by storing multiple measured values around a set-up point and calculating minimum and maximum values for the measured quantity [51]. In order to measure the impact of sensor fault on the measured quantity, uncertainty techniques are used. This enables the continuation of using a sensor after the fault has arisen.

Nuclear Sensors
Nuclear sensors are very uncommon [37,74] due to two reasons: they are costly and have strict safety regulations for their use. Recent developments have made the availability of low-level radiation sources for safe use of these sensors [37]. Table 7 summarizes the pros and cons of these sensors. Have low capacitance; output signals prone to noise contamination; produce output signals of very low magnitude. [74,75] Nano-sensors Vary in size from 1 to 1000 mm, using nanotechnology.
Thin layers of metal films or semiconductors; more advance than MEMS using special forms of etching, optical lithography or electron beam lithography.
Used as accelerometers, biological sensors and sensors for airborne chemicals.
Lower production costs; reduced power consumption; smaller size.

Micro-Sensors (MEMS Sensors)
Micro-sensors contain two and three dimensional micro-machined structures that are a part of microelectromechanical system (MEMS) devices [74,75]. These sensors can be regarded as small sized transducers, as they convert mechanical signals from an energy source into electrical form. Currently, sensors that measure temperature, pressure, force, speed, sound, magnetic field, optical, biomedical, and chemical features are being used successfully by industries [37,74]. Table 7 summarizes the main characteristics of these sensors.

Nano-Sensors (NEMS)
Nano-sensors, based on nanotechnology, are the most recent development in sensing technology [72,75]. These are a part of nano-elctromechanical system (NEMS) devices which include nano-actuators as well. Table 7 gives a summary of these sensors.

Conclusions
This paper has discussed the different types of sensor technology used in the manufacturing industry, specifically in smart factories. An extensive review of the extant literature is also presented on the technology lead smart factory. Key differences between a smart factory and traditional factory have been highlighted. Use of sensors, interoperability of different IoT devices, integration of robotics and AI, and use of VR techniques were found to be the key intelligent features of a smart factory. As sensors are an important component of a smart factory, the main sensor types and their sub-types, along with their characteristics, materials, uses, application areas, advantages and disadvantages have been identified in this paper. Sensors and Wi-fi are changing the way people communicate with the surrounding world, bringing a new era of connectivity, termed the internet of things. This technology has the potential to provide virtually boundless opportunities to businesses and communities, with enhanced connectivity and use of collected data. With the help of these technologies, data flow is integrated between partners, suppliers, and customers, as well as organizations to develop a finalized product according to customer demands. Manufacturers around the world are beginning to realize the importance of sensors, and the benefits of merging traditional operations and IoT. Therefore, a developing trend for the smart factory is human-machine collaboration. This paper attempted to provide a broad overview of the extant literature on Industry 4.0, summarizing the variations between traditional and smart factories, and the main sensors used within a smart factory setup.
While aspects of the smart factory have been successfully implemented in several key industries (e.g., the chemical industry), it has been found that there are technological, cost, and knowledge barriers to the broader implementation of sensors across manufacturing [20]. These barriers also include standardization, cyber security, and risk concerns, which may enhance organizational resistance to automation within the factory. In terms of standardization, two Industry 4.0 reference architectures have been standardized, namely, Reference Architecture Model Industry 4.0, from the International Electrotechnical Commission, and Industrial Internet Reference Architecture, from the Object Management Group. A discussion on these standardized architectures and the compatibility of each Industry 4.0 proposal to these standards is needed. Moreover, the concept of digital twin in the manufacturing system configuration stage, which can enable the validation of system performance in a semi-physical simulation manner, is gaining interest. The digital twin conducts a direct test and validation that can quickly locate the problem and malfunctioning part, rectify the design mistakes, and test the workability of equipment in the system execution [76,77]. In addition, a deficiency of labor skills and manufacturer technical willingness relative to available technologies can be limiting. It has been revealed that there is a scarcity of studies that address these issues. Furthermore, artificial intelligence, block chain, cloud computing, and big data analytics (also known as ABCD) are known for the generation of information technology [76]. The employment of blockchain in smart factories is known to enable increased security, enhanced traceability, and reduced costs [76,77].
These features along with standardization and cyber security issues need a detailed study and therefore, are identified as a future extension of this work.
Funding: This research received no external funding.

Conflicts of Interest:
The authors declare no conflict of interest.