Temperature Monitoring in Metal Additive Manufacturing in the Era of Industry 4.0

Abstract

The field of metal additive manufacturing has witnessed significant growth in recent years, with technology offering the ability to produce complex geometries that are challenging to manufacture using the traditional methods. In situ monitoring and control of the manufacturing process are crucial for increasing the production capacity and improving the quality of manufactured parts. This article provides a comparative analysis of computational, indirect, and direct methods for in situ temperature monitoring during additive manufacturing of metal alloy components. Furthermore, it discusses the current status, recent improvements, and perspectives for in situ temperature measurements. The basic principles of thermal imaging, two-color pyrometry, and millimeter-wave radiometry are explored, highlighting their limitations for addressing challenges related to material emissivity and rapid changes in building material composition. Overcoming the challenges related to the inaccessibility of the chamber where the parts are formed, direct temperature measurements would allow for the integration of collected information into big data systems. Within the framework of Industry 4.0, this approach offers a viable alternative to the conventional metal shaping processes, improving the production capacity and part quality. This research aims to contribute to ongoing advancements in metal additive manufacturing and its potential to completely replace traditional metal casting practices in the Industry 4.0 era.

1. Introduction

Metals are the most widely used engineering materials [1,2]. Recently, many metallic components have been manufactured using additive manufacturing techniques, where aluminum, titanium, stainless steel, and other materials have been the primary constituents [3,4]. Metal additive manufacturing is relatively new, and there is still much to be developed, as the fabricated parts have yet to meet the industry’s expectations [5,6]. The main challenges to be resolved are maintenance costs, improved automation, an enhanced production speed, improved tensile behavior, reduced fatigue, increased hardness, an improved surface quality, and a homogeneous microstructure [7,8]. Provided it overcomes current barriers through improvements in material science and manufacturing processes [9], in the Industry 4.0 era, metal additive manufacturing may replace traditional casting by leveraging real-time sensor data for improved process control and part quality [10,11,12,13].

Since the beginning of the 21st century, rapid surfacing of advanced technologies and processing methods has driven astonishing growth in metal additive manufacturing. As per recent industry reports, the number of additive manufacturing systems sold for metal part production reached 3793 in 2023, a remarkable increase of 24.4% compared to the previous year’s figure of 3049 (Figure 1), while for the same period, the additive manufacturing industry’s expansion resulted in an 11.1% rise in its market value [14]. Accordingly, the number of companies producing industrial-grade additive manufacturing systems is also increasing at a comparable pace; for example, in 2017, the number of companies rose to 135 from 97 in 2016, accounting for more than one thousand system producers and users nowadays [15].

This review aims to address the significant challenges associated with in situ monitoring of additive manufacturing processes at high temperatures. It evaluates the existing metal additive manufacturing techniques, highlights the sensors used for temperature monitoring, and offers insights into improving process control within the framework of big data and Industry 4.0 principles.

Following the introduction in Section 1, Section 2 addresses the primary challenges. Section 4, Section 5 and Section 6 outline the crucial research questions, focusing specifically on metal additive manufacturing (Section 4), sensors for temperature monitoring in additive manufacturing (Section 5), and the role, challenges, and outlook of metal additive manufacturing within the context of Industry 4.0 (Section 6).

The review protocol methodology involved several steps for the systematic collection, sorting, and removal of data from the literature. Different search terms and strings were entered into the Inspec and ScienceDirect databases. The primary search phrase used was ‘metal additive manufacturing’, combined with the terms ‘materials’, ‘methods’, ‘sensors’, ‘temperature control’, ‘data control’, and ‘Industry 4.0’, using the ‘AND’ string to refine the query. To focus specifically on high-temperature applications relevant to metal additive manufacturing, exclusion terms such as “bio” and “polymers” were applied. Once the relevant literature had been gathered, the selection process prioritized more recent publications. Additional emphasis was placed on research concerning advanced applications in areas such as space and medicine, design automation and optimization, smart manufacturing, artificial intelligence, cloud manufacturing, big data systems, and other approaches that can overcome the inherent drawbacks related to direct temperature measurements. Finally, the data were condensed to include up to one hundred of the most relevant references.

2. Background and Challenges

The field of additive manufacturing has experienced constant growth over the past three decades, significantly accelerating in the last five years [16]. Figure 2 illustrates the remarkable rise of the metal additive manufacturing industry, as revealed by the worldwide consumption of metal powder and wire feedstock. Titanium alloys have become the most processed feedstock material for the metal additive manufacturing industry, mainly used in aviation due to their light weight and in the medical industry because of their excellent biocompatibility [17,18]. The consumption of nickel-based alloys is increasing for use in components exposed to high temperatures, in space, and in energy industries. With a decrease in the price per part for additively manufactured components, the most significant increase in usage for the coming year is expected for copper alloys and stainless steel [19]. Copper is increasingly used in electric vehicles and its systems, heat conductors, and heat sinks in the engine parts of space components, while stainless steel components are finding service in the automotive, tooling, and mechanical engineering industries. The total consumption of metals in additive manufacturing applications is estimated to reach 4614 tons in 2025.

Additive manufacturing allows parts with intricate geometries to be rapidly produced [21]. Considering commercialized applications for metal shaping, additive manufacturing provides numerous benefits, as parts are rapidly produced without typical castings or other expensive and energy-intensive processes [22]. However, a significant concern is the quality of the parts produced compared to those produced using the traditional methods [23]. In most cases, parts made through additive manufacturing are more susceptible to containing voids (ranging from the nanometer scale to several microns in size), which affects the surface roughness and causes local structural incoherency that is only revealed during tensile testing [24]. These challenges are caused by process conditions such as the powder’s purity and particle size variations, narrow energy sources of laser light or electron beams, and unknown process temperatures. The repeatability between each run is also a concern, where the quality of the products significantly varies for similar manufacturing process parameters [25]. In practical applications, in situ monitoring requires highly qualified personnel and a separate quality control system. For this reason, the primary quality control tools currently employed are ex situ techniques such as X-ray diffraction [26], X-ray computed tomography [27], and focused ion beam cross-sectioning and electron microscopy [28], utilized to examine the quality of the parts produced after the entire production process is completed. These methods provide insights into the final part quality, but the details of how the material and part have evolved during production are inexistent. With rapid development of the additive manufacturing industry and demands for ever smaller particles on both sides, in the feeding material and the final product, if such challenges are not addressed, the uniformity and consistency of the parts produced through metal additive manufacturing will remain in question.

3. Metal Additive Manufacturing

Additive manufacturing has advanced from using dyes to create 2D forms to fusing organic polymers and plastics to build complex 3D structures and finally producing high-strength structural materials from metal powders. As given in

Table 1. Commercialized additive manufacturing methods for generating metallic structures.

Method	Advantages	Disadvantages	Feeding Material
Selective Laser Melting (SLM)	Binder-free, rapid processing, high powder recyclability, high geometric precision, enhanced material utilization, localized microstructure control, allows for automatization, allows for monitoring	High equipment costs, high operational costs, time-intensive, thermal residual stresses, build volume and processing defects	Aluminum alloys, titanium alloys, stainless steels
Electron Beam Melting (EBM)	Excellent structural properties, processing of reactive metals, rapid processing, high processing speeds, good thermal management, the ability to process refractory metals, finer microstructural features, good tensile properties, oxide-free parts	Higher costs, limited material options, preheating required, requires a vacuum, limited build size	Aluminum alloys, titanium alloys, nickel-based superalloys, cobalt–chrome alloys
Direct Energy Deposition (DED)	High recyclability, on-site part repair, allows larger components to be printed, high deposition rates, the ability to fabricate large parts, in situ alloying, high energy efficiency, rapid prototyping and production	Surface finish, high heat input, high residual stresses, prone to processing defects, post-processing required	Aluminum alloys, titanium alloys, stainless steels, nickel-based alloys cobalt alloys, high-entropy alloys
Binder Jets	Cost-effective, high production speeds, no residual stresses, controlled structure	Post-processing required, binder residue issues, fewer material options	Aluminum alloys, titanium alloys stainless steels, nickel alloys
Ultrasonic Additive Manufacturing (UAM)	Joining dissimilar materials, low temperature, low residual stresses, cost-effective feedstock	Limited material compatibility, low deposition rates, limited geometrical complexity	Aluminum alloys

, the processes used to handle metallic powders at high temperatures include selective laser melting, direct metal laser sintering, electron beam melting, and direct energy deposition [29,30].

Selective laser melting (SLM) involves depositing a layer of metallic powder onto a print bed, which is then heated by a laser beam to slightly above the melting point to create a molten pool. The molten metal cools rapidly and solidifies. The building platform is then lowered to the thickness of one layer, and the process is repeated until the printing is complete. After cooling, the building platform is raised to reveal the final part, while the remaining powder is retrieved for reuse [31,32]. After removing the support structure, the final part is cleaned and, if necessary, surface-machined to improve the surface quality. The typical layer thickness in commercial systems ranges from 20 to 100 μm, while the particle size varies between 20 and 50 μm [33]. Single-component metals such as titanium, aluminum, and steel are the most suitable materials [34].

Direct metal laser sintering (DMLS) is a powder bed fusion (PBF) process similar to selective laser melting (SLM), which uses combined metallic powders. A metal with a lower melting point will liquefy the powder particles of the metal with a higher melting temperature. This results in a high-density product that can be produced in complex shapes with intricate geometries [35]. This maximizes the tensile strength and allows light but structurally sound parts to be created using metals such as aluminum.

Electron beam melting (EBM) is a process that occurs in a vacuum chamber. In this process, a high-energy electron beam fuses the metal powder particles layer by layer. The electron beam moves according to a predetermined route, which results in selective melting of the powder pool. This process is repeated layer by layer until the final part is entirely built [36].

Through Direct Energy Deposition (DED), metal powder or the tip of a wire is melted as it is deposited onto a substrate or a surface using focused electron beam energy. The material is delivered via a nozzle that moves in multiple axes, allowing for the creation of complex geometries. Figure 3a shows the powder bed fusion process, one of the most widely used techniques for creating metal parts [37,38]. This technique utilizes an electron beam or laser energy for direct metal laser sintering, while Figure 3b illustrates the use of wire as the feed material [39,40,41]. Typically, these processes are guided by a computer-aided design system that directs the nozzle’s path and builds the desired shape layer by layer. By accurately applying the energy and the material, parts are produced without porosity or structural gaps [42]. While Direct Energy Deposition is primarily used for metals, due to its high energy density in targeted spatial areas, ceramics and composite materials can also be built, although its application in these areas is less widespread [43].

Regardless of the metal additive method, the extreme thermal gradients encountered during part manufacturing processes pose difficulties in mitigating the segregation and ensuring consistent elemental stoichiometry. Typically, the laser scan speed, laser power, and beam overlap are set at the beginning of the process and can only be adjusted between each powder dispensing but cannot be changed during melting and solidification periods. The lack of feedback control often results in unmelted powder particles being concealed in the melted powder, causing the produced parts to have worse mechanical properties [44].

The binder jetting technique involves using binders to join metallic powders. This solid-state process addresses issues such as shrinkage, segregation, and unevenness, as the material remains solid throughout fabrication. As a result, it is more accommodating when dealing with variability in the raw powder.

In Ultrasound Additive Manufacturing (UAM), bonding is achieved by applying ultrasonic vibrations, typically at a frequency of 20 kHz, laterally (along the width of the sheet) using a sonotrode that moves along the length of the sheet under a constant normal force. This lateral vibration causes the top sheet to rub against the bottom sheet, facilitating direct metal-to-metal contact. Following this, dynamic shearing occurs at the newly formed metal asperities, resulting in their eventual bonding. Although bonding occurs in solids, when dissimilar materials are bonded, severe plastic deformation typically predominates in softer materials. Regardless of the technique used, temperature is the most critical parameter to measure during additive manufacturing processes for metal alloy shaping. Controlling the temperature mitigates the challenges mentioned above, allows for the proper formation of the desired parts, and allows for control over selected solidification parameters, which significantly influence the properties of the final product.

4. Sensors for Monitoring Temperature in Additive Manufacturing

Sensors have become essential tools in modern manufacturing focusing on high-quality products. The manufacturing process sometimes becomes so complex that it relies entirely on sensors for automatic monitoring and control. The longevity and accuracy of these sensors, particularly in additive manufacturing, remain a challenge because of their lifecycle costs and data consistency. The latter can often be the weakest link in the reliability of complex systems. Assuming sensor technologies are becoming more affordable, communication between manufacturing facilities and remote decision-making centers remains a significant concern. Interestingly, the most promising approach to addressing the challenges mentioned above may be novel techniques for sensor production based on additive manufacturing [45,46,47].

Optical and radiation pyrometers can measure high temperatures but have significant drawbacks. They are highly subjective, meaning that although sensors properly detect a temperature change, two different devices may report different temperatures. Sensor conversion systems for translating recorded signals into temperature values are typically accurate within narrow conditions. Therefore, the algorithms that convert sensor readings into the actual temperature must be appropriate for the actual conditions.

The treated material’s inaccessibility in additive manufacturing makes it challenging to implement large-scale and widely available sensing systems for direct temperature measurements to monitor the system performance effectively. Moreover, measuring high temperatures is another challenging task due to environmental factors affecting any device or sensor in such conditions [48]. In extreme temperature environments, only two methods are generally reliable: radiation pyrometry (including optical pyrometry) and thermocouples for higher temperatures [49].

Regarding promising measuring techniques, Dedyulin et al. [50] collected data on a large number of emerging technologies for monitoring temperatures in highly harsh environments, including Johnson noise thermometry; optical refractive index gas thermometry; Doppler line broadening thermometry; optomechanical thermometry; fiber-coupled phosphor thermometry; fiber-optic thermometry based on Rayleigh, Brillouin, and Raman scattering; fiber Bragg grating thermometry; Bragg waveguide grating thermometry; ring-resonator thermometry; and photonic crystal cavity thermometry.

All of these methods are in the earliest research or metrological testing stages. The success of emerging methods will largely depend on two factors: their performance and the cost of their adoption. This cost encompasses both financial expenses and the need to change established industry standards and practices. A key contributor to the financial cost is the integration of sensors into the additive manufacturing system, in particular multiple lasers, wavelength references, photodetectors, and related electronics. Generally, all emerging techniques will require complex and costly integration procedures [51].

4.1. Indirect Temperature Measurements

All objects, regardless of their size, shape, and composition, emit energy as electromagnetic waves if their temperature is above zero Kelvin. The energy emitted by an object is directly proportional to its temperature, which means the more energy an object emits, the hotter it is. By measuring the intensity of the electromagnetic radiation emitted by an object, its temperature can be determined. This concept is widely used in many fields, including engineering, physics, astronomy, and many others.

Even before Maxwell introduced the classical theory of electromagnetic radiation in 1862, blacksmiths and foundrymen used the light emitted from hot objects to assess their temperature. In the late 18th century, Josiah Wedgewood invented the first pyrometer. He used pieces of clay heated at known temperatures to establish a color scale that could be compared to the kiln’s temperature. Later, Wedgwood developed a more accurate technique for measuring temperature based on the dimensional shrinkage related to the starting clay.

Nowadays, the Planck distribution is a tool for understanding how much light an object emits at a specific temperature by quantifying the amount of photons emitted at different wavelengths. Figure 4 shows the Planck distribution for different temperatures of black body objects and how it compares to the distribution in gray-body objects.

The greatest challenge in indirect temperature measurement methods is predicting the emissivity with reasonable accuracy. Although a large amount of data is available in databases, these are usually representative values for generalized conditions. As shown in Figure 5, the infrared emissivity of typical refractory or structural materials with a high melting point changes significantly with temperature, creating additional challenges for additive manufacturing applications. These challenges arise due to the wide range of temperatures encountered over short temporal and spatial scales.

Most often, conventional optical sensors, like photodiodes or infrared cameras, are used to monitor the brightness of the melt pool in metal additive manufacturing processes [54,55]. The data collected, such as the size of the melt pool, is used to control and alter the electron beam or laser parameters. This helps to improve the uniformity of parts by reducing the power distortion and increasing the part density. Since the material’s emissivity during these processes is unknown, only the intensity of the melt pool can be used as a measured property (not the temperature).

4.2. Direct Temperature Measurements

In traditional metal casting, thermocouples are used to measure temperature directly. A thermocouple is a temperature sensor that detects the differences in temperature at the junctions of two conductive or semiconducting materials. The electrical current generated by the thermocouple is proportional to the difference in temperature at the junctions and the properties of the materials. This means that the difference in temperature at the two junctions creates an electromotive force (E_emf), which can be calculated using Equation (1), where S is the Seebeck coefficient, and ∇T is the temperature difference between the two junctions.

E_emf = −S∇T

(1)

The Seebeck coefficient depends on the material’s composition and varies with temperature. In typical conducting materials at room temperature, the Seebeck coefficient is expressed in μV/K and can range from −100 to +1000 [56]. As material properties vary significantly with temperature, thermocouples of different materials will have a different sensitivity and accuracy at different temperatures [57,58].

The most common types of commercial thermocouples are T, J, E, K, S, and B, where each type consists of wires of different conductive materials. The typical thermocouple types, conductor materials, and suggested operating temperature ranges are given in

Table 2. The most common types of thermocouples and recommended operating temperatures [62,63].

Thermocouple Type	Wire Material	Operating Range, °C
T	Copper and Copper–Nickel	−250 to 350
J	Iron and Copper–Nickel	0 to 750
E	Nickel–Chromium and Copper–Nickel	−200 to 900
K	Nickel–Chromium and Nickel–Aluminum	−200 to 1250
S	Platinum–10%Rodium and Platinum	0 to 1450
B	Platinum–30%Rodium and Platinum–6%Rodium	0 to 1700

. Type K thermocouples are the most widely used because of their wide temperature range and durability due to their significantly higher chemical inertness than that of, e.g., T-types containing copper or J-types containing iron. T-type thermocouples have the highest accuracy of all base metal thermocouples, with values below 0.75% for the difference from the electromotive force reading [59]. This is followed by E-type thermocouples and then J, K, and N types when compared to the standard margin of error. Type E thermocouples have the highest electromotive force (E_emf) values compared to those of any other standard metal thermocouple, which enable high measurement sensitivity [60]. If the process conditions allow, non-insulated thermocouples would be the best solution for applications requiring a swift response [61].

4.3. In-Industry Practices for Temperature Monitoring

Commercialized mid-wave infrared (MWIR) high-speed thermographic cameras have become a standard tool for assessing temperatures in additive manufacturing processing. These cameras use arrays of sensors that detect infrared radiation by collecting data over a range of wavelengths. InSb cameras detect mid-wavelength infrared radiation in the 3–5 μm band at rates exceeding 1000 Hz, with large fields of view, a spatial resolution of 20 μm per pixel or better, and low noise-equivalent temperature differences (~20 mK) when using cryogenic cooling. However, these cameras are still susceptible to significant errors due to variations in object emissivity. Furthermore, the extreme temperature changes observed in additive manufacturing processes pose challenges for commercial mid-wave infrared cameras.

Infrared hyperspectral imaging provides more accurate temperature measurements in additive manufacturing. Snapshot hyperspectral imaging has a capability for predicting process states and defects with the help of machine learning algorithms, where the characteristic advantage is weighed against the drawbacks of the low temporal resolutions and reduced spatial resolutions of other in-process data acquisition and subsequent data processing methods. While the conventional assessment methods focus more on the physical integrity of forming pieces rather than the material composition, hyperspectral imaging can also detect minor temperature differences related to variations in chemical composition. This provides another important piece of information, as contamination can also lead to the formation of cracks and malfunctions in the final products [64]. The National Institute of Standards and Technology (NIST) developed an 11-wavelength and high-speed thermographic camera. Instead of using the ratio of 2 intensities, the 11 intensity values are used to back-fit the Planck distribution for the object temperature and emissivity [65,66]. However, this camera is expensive and has limited acquisition rates and fields of view (about 50 Hz, with an 80 × 80 pixel sensor array). While limited to laboratory use, as they advance, hyperspectral imaging technologies could offer a viable solution to the emissivity problem in indirect temperature measurements [67]. A comparison of different sensors for temperature monitoring in metal additive manufacturing is given in

Table 3. Key advantages and disadvantages of indirect sensors used for process monitoring in metal additive manufacturing [68].

Sensor	Advantage	Drawback
Photodiode	High temporal resolution Easy to integrate	No spatial resolution No spectral resolution
High-speed camera	High spatial resolution High temporal resolution	Amount of data No spectral resolution
Bolometer (resistance change)	LWIR sensitivity Good spatial resolution	Low temporal resolution Require specialized optics
Spectrometer	Excellent spectral resolution Good temporal resolution	No spatial resolution Prone to chromatic aberrations
Snapshot hyperspectral camera	Good spatial and spectral resolution	Correction/calibration necessary Amount of data

.

An innovative way to correct the temperature during additive manufacturing is by making an in situ black body correction inside the build chamber. By measuring the emittance of a perfect black body under the same conditions as those for the object of interest (with the same temperature, view factor, etc.), it is possible to calculate the emissivity. Rodriguez et al. [69] found temperature errors of up to 300 °C using a thermographic camera, which further required considerable corrections to obtain accurate temperature values. After correction by building a black body in the build chamber, the temperature uncertainty was decreased to approximately 3.7% over the temperature range examined. Nonetheless, implementing this approach in practice is challenging since creating a perfect black body is complex. When implementing such a method in additive manufacturing, as the emissivity varies with temperature, surface roughness, and other factors, black body correction must be continuously performed during the buildup of the structure to ensure accurate results. A single-point black body correction would not be sufficient, particularly over a wide temperature range and sharp temperature gradients at short distances, both typical for additive manufacturing of metallic structures.

Various sensors can measure temperature but only provide point information or the 2D temperature distribution. Reliable temperature readings are crucial for monitoring additive manufacturing processes and computer modeling, where using the fewest sensors while ensuring that all necessary information is exchanged is critical [70]. Computer modeling can identify critical parameters that cannot be obtained from sensors, provided that the input data is accurate and comprehensive [71]. Obtaining accurate data for metal additive manufacturing is challenging due to the inaccessibility of the treated materials during melting and solidification and the pace of the process. The latest advancements in digital twin technology present a potential opportunity to enhance temperature assessments in additive manufacturing settings. Integrating current visual technologies with emerging virtual, augmented, and mixed reality platforms could open a new avenue for exploring next-generation quality control methods [72]. A breakdown of the most common temperature measuring methods is given in

Table 4. Comparative list of most common temperature measuring methods in metal additive manufacturing.

Sensor Type	Method Type	Operational Principle	Advantages	Applications	Commercialized
Thermocouples	Direct	Generates voltage between two dissimilar metal junctions	Wide temperature range, fast response, diverse applications	Monitoring build plate temperatures	Yes
Pyrometers	Indirect	Measures the intensity of infrared radiation emitted by an object	Accurate temperature readings without contact	Monitoring high-temperature processes (laser sintering or directed energy deposition)	Yes
Infrared cameras	Indirect	Provides a spatial representation of the temperature distribution	Visualizes temperature variations across the build area	process monitoring, quality control, in situ troubleshooting	Yes
Thermal imaging	Indirect	Measure the infrared radiation emitted by an object	Non-contact monitoring of moving objects or inaccessible areas	Monitoring melt pool temperatures, thermal gradients, and heat distributions	Yes
Radiometry	Indirect	Measures the intensity of electromagnetic radiation	Safe non-ionizing radiation, penetrates inside the examined object	Monitoring wide areas consisting of various materials	Yes
Mathematical modeling	Computational	Uses conservation equations and discretizes complex geometries into smaller elements	Enables a comprehensive data analysis of complex systems	Controlling industrial processes where the traditional sensors might fail	n/a

.

5. Metal Additive Manufacturing and Industry 4.0

Due to their bottom-up emergence, the capacities of existing systems constrain the development of large-capacity smart factories. However, due to its scalability and customization potential, additive manufacturing could emerge as a crucial element of Industry 4.0 [73,74]. Currently, additive manufacturing is finding applications in diverse industries, such as aerospace, biomedical, and manufacturing, where specific properties and shapes are required [75,76]. Despite doubts about its suitability for mass production, the use of additive manufacturing in industry is on an upswing thanks to new technological advancements [77,78]. As a developing technology that can create accurate, strengthened, and intricate objects at an increased production speed, metal additive manufacturing, fully integrated into big-data systems, offers a viable alternative to the conventional metal shaping techniques.

The information gathered from sensors that monitor the quality of products or equipment activity generates a significant amount of data. This data must be analyzed along with other big-data-related information [79,80]. Cloud computing is a crucial technology for processing vast amounts of information which presents significant potential for the industrial world [81,82]. In particular, large-scale processing using sensors feeding information from the production line may allow process parameters that would otherwise be executed with significant delays to be automatically adjusted. These technologies help ensure that existing information is effectively stored, processed, shared, and utilized to enable intelligent manufacturing of future metal-based products [83].

It is crucial to develop physics-based, scientifically proven models for temperature evaluations to enable cloud computing and intelligent manufacturing. Physics-based computer modeling significantly reduces the amount of data exchanged between sensors and control units, allowing for temperature evaluations in every part of the forming object at any time during the forming process. Unlike one- or two-dimensional temperature control, which relies on physically measured and analyzed values that require numerous sensors, physics-based 3D modeling with fewer sensors allows for temperature assessments in each build layer as the object forms. However, scientific knowledge of the physical processes inside the printing machine is a prerequisite for effective process design and reliable sensing protocols.

Lu and Wang [84] proposed and experimentally tested a physics-based compressive sensing model that reconstructs a printed object’s complete 3D temperature distribution with only a few sensors monitoring the additive manufacturing process. Figure 6 depicts their findings, which show a minor difference between the actual sensor values and the model-predicted values, with the maximum error being well below 10%, an error that could be lowered by repositioning the sensors. Similar physics-based computer modeling will significantly improve additive manufacturing process monitoring while reducing system costs. In large-scale manufacturing systems with various sensors, these monitoring approaches can reduce the numbers and types of sensors without losing much process information. In this manner, undetected faulty sensor risks can be downsized. Multi-physics models that consider the thermal expansion and shrinkage during manufacturing processes will improve the accuracy of the modeling further. An optimization scheme developed to find the best sensor locations, particularly with high gradients, would reduce the prediction error further.

Computer modeling provides a reliable data feed with minimal installed sensors [85]. Additive manufacturing processes aiming towards automatic process control require a physics-based comprehensive computer model adjusted by a reliable data feed. This allows the treated materials’ properties to be modeled in spaces where sensors cannot be installed. This approach relies on in-depth engineering knowledge of specific processes and an equally comprehensive understanding of the observed phenomenon, which would result in the design of accurate sensing protocols for efficiently producing metallic parts using additive manufacturing technology.

6. Challenges and Outlook

In the 2010s, quality control was mainly focused on gravimetric measurement of the weight of the melt pool, which was derived from intensity data [86] rather than monitoring temperatures. In some cases where the density significantly changes with temperature changes, the thermogravimetric techniques developed for monitoring the melt pool size can be modified to include temperature data. Nowadays, direct temperature measurement methods during metallic structure layering are an intriguing challenge. This data would identify potential irregularities during the process and allow for remote monitoring and process control.

For indirect temperature measurements to be reliable in additive manufacturing using metal alloys, it is necessary to calibrate infrared (IR) cameras in a closed-dark chamber to avoid emissivity disturbances. Then, calibration should be performed by directly measuring the temperature of the contact surface and correlating the results with the measured emissivity. It is crucial to calibrate as closely to the build platform as possible, as any variation in temperature, pressure, or suspended powder can affect the apparent surface temperature. Once the absolute temperature is measured, the data can be utilized to ensure that a consistent temperature is maintained during the manufacturing process. This temperature control can help to reduce the segregation and improve the surface conditions and density while minimizing defect concentrations and making materials’ stress–strain characteristics more tightly controlled and repeatable [87]. For producing large parts, tight and rapid temperature control may be used to control the solidification conditions, making additive manufacturing of metallic components more widely used throughout industry, reducing costs and processing times.

When considering temperature measurements only, the thermal history during the additive manufacturing process has a significant impact on the performance of the resulting structures [88]. Numerous studies from various fields have focused on developing thermal prediction models to enhance quality control and minimize defects in additive manufacturing processes. For example, temperature monitoring may be considered from the machine learning perspective [89,90,91,92], using physics-informed deep learning [91,92,93,94,95], or using AI-enhanced models [96,97].

Rapid advancements in the other components of smart manufacturing may overcome the inherent limitations regarding the physical placements of temperature sensors inside forming parts [98]. In recent years, digital transformation has had a significant impact on metal additive manufacturing, shifting it toward smart manufacturing, where artificial intelligence and the integration of cyber-physical systems with physical objects play key roles. In sustainable economies, innovative technologies must be energy-efficient, green, responsible, and sustainable. Transformation changes are intensely involved in global energy policy and others, and therefore, digitization, artificial intelligence, and cyber-physical systems are unavoidable components of global megatrends [99].

7. Conclusions

The field of metal additive manufacturing has seen substantial growth, with increasing utilization in various industries such as aviation, medical, automotive, and mechanical engineering fields. In 2024, there was a 24.4% increase in the number of systems sold, while the market value of the metal additive manufacturing industry rose by 11.1%. The ability of metal additive manufacturing to produce intricate geometries and the continuous development of advanced technologies and processing methods have significantly contributed to its expansion.

However, challenges still need to be overcome to meet industry expectations, including maintenance costs, production speed and capacity, and the homogeneity of the microstructures of manufactured parts. In situ temperature monitoring during the additive manufacturing process is crucial to improving the quality and reliability of the fabricated components. Due to the tightness and inaccessibility of the build chamber where the parts are formed, direct temperature measurements of the formation of these parts are practically inconceivable. With the inherent difficulty installing sensors for direct temperature measurements, for the time being, temperature measurements will rely on indirect measurements using commercialized mid-wave infrared cameras, where the optimal solutions involve manufacturing parts in a controlled environment and utilizing calibrated emissivity values in conjunction with direct temperature measurements.

As additive manufacturing continues to evolve within the Industry 4.0 framework, advancements in material science, manufacturing processes, and data control through sensor technology will be essential in realizing the full potential of metal additive manufacturing, addressing the dual challenges of production capacity and part quality. The ultimate solution for addressing the challenges in temperature monitoring within metal additive systems relies not only on temperature sensors but also on a broad framework that encompasses other components of smart manufacturing, such as digitalization, artificial intelligence, the integration of cyber-physical systems with physical objects, and others.