The improvement of additive manufacturing through artificial intelligence, machine learning, and deep learning

Additive manufacturing (AM) has made it possible to manufacture complex personalized items with minimal material waste, leading to significant changes in the manufacturing industry. However, optimizing and improving additive manufacturing processes remains challenging due to the complexity of design, material selection, and process parameters. This review explores the integration of artificial intelligence (AI), machine learning (ML), and deep learning (DL) technologies to improve and innovate the AM field. The AI driven design optimization program provides innovative solutions for 3D printing of complex geometric shapes and lightweight structures.

By utilizing machine learning (ML) algorithms, these programs can analyze large amounts of data from previous manufacturing processes, thereby improving efficiency and productivity. ML models can learn and identify complex patterns from historical data that human operators may overlook, thereby promoting design and production automation. Deep learning (DL) utilizes complex neural networks to manage and interpret complex information, and gain a deeper understanding of manufacturing processes, thereby further enhancing this ability.

Integrating artificial intelligence, ML, and DL into AM can produce optimized lightweight components, which is crucial for reducing fuel consumption in the automotive and aviation industries. These advanced artificial intelligence technologies can optimize design and production processes, and enhance predictive modeling for process optimization and defect detection, thereby improving performance and reducing manufacturing costs. Therefore, integrating artificial intelligence, ML, and DL into AM can improve the accuracy of component manufacturing, achieve advanced material design innovation, and open up new possibilities for product design and material science innovation.

This review discusses and emphasizes the significant progress in applying artificial intelligence, ML, and DL to AM, and identifies future development directions. By utilizing these technologies, AM processes can achieve unprecedented levels of precision, customization, and productivity in analysis and modification.

1.Introduction
3D printing or additive manufacturing (AM) is a revolutionary industrial production method that can produce more robust and lightweight systems and components. Additive manufacturing technology is classified according to the standardized framework defined by ISO/ASTM 52900. This framework organizes the additive manufacturing process based on material models added or created during the construction process. The widely accepted classifications include the following seven categories:

① Material extrusion: Selective distribution of materials through nozzles or openings, such as fused deposition modeling (FDM).
② Adhesive spraying: Selective deposition of liquid adhesive to connect powder materials, such as metal adhesive spraying.
③ Powder bed melting: Thermal energy selectively melts various regions of the powder bed, such as selective laser sintering (SLS) and direct metal laser sintering (DMLS).
④ Restoration photopolymerization: Selective solidification of liquid photopolymers in large barrels through photopolymerization, such as stereolithography (SLA).
⑤ Material spraying: selectively depositing material droplets using PolyJet and MultiJet.
⑥ Sheet lamination: bonding material sheets together to form an object (e.g. in laminated object manufacturing (LOM)).
⑦ Directed Energy Deposition (DED): During the deposition process, focused thermal energy is used to melt materials (such as LENS).

This standardized classification provides people with a clearer understanding of technology, which is classified based on the core processes used rather than specific brand names or technologies such as SLA or FDM.

Artificial intelligence (AI) is a general field that includes the development of machines and systems capable of performing tasks that typically require human intelligence, such as decision-making, problem-solving, and language understanding. Machine learning (ML) is a specific subset of AI that enables systems to learn, recognize patterns from data, and make decisions with minimal human intervention. Deep learning (DL) is another major in ML that involves using neural networks with multiple layers (deep neural networks) to model complex data representations and improve the accuracy of tasks such as image recognition and natural language processing.

Figure 1 shows the application of data-driven ML in 3D printing. The 3D printing design rules are shown in Figure 2.

Fig. 1. Application of data driven ML in the 3D printing process.

Fig. 2. Designing rules for 3D printing.

This review aims to provide a comprehensive overview of the current research status on the use of ML, DL, and AI in AM. By analyzing the relationship between additive manufacturing and these advanced methods (AI, ML, and DL), this study aims to demonstrate the potential for developing smarter, more flexible, and more efficient manufacturing systems. Table 1 categorizes the relationships and applications of different methods in additive manufacturing processes, design, and integration.

Table 1. The relationships and applications of different methodologies in AM processes, design, and integration.

2.Design Optimization

Artificial intelligence driven optimization techniques, such as ML and DL, are increasingly being used to enhance design process optimization during additive manufacturing. These methods are capable of generating optimized geometric shapes and material distributions that meet specific performance standards and often go beyond traditional design methods. The goal of this multidisciplinary strategy is to improve the capability, quality, and efficiency of additive manufacturing processes. In addition, artificial intelligence systems can generate ideal design solutions given a set of parameters and specifications.

The integrated ML framework can be used to determine the relationships between processes, structures, and properties, as shown in Figure 3 (a), and the subsequent design from structure to program can utilize the learned ML model. As shown in Figure 3 (b), ML can be used to construct further connections between processes, structures, and attributes in any direction.

Fig. 3. ML-integrated design for AM.

DL is a subfield of machine learning that is particularly suitable for additive manufacturing design optimization due to its ability to handle complex patterns and high-dimensional data. Table 2 provides an overview of the current research status and applications of AI and ML in additive manufacturing processes, covering various fields such as design optimization, process optimization, quality control, defect detection, and predictive maintenance.

Table 2. The status and applications of AI and ML in AM process across various areas like design optimization, process optimization, quality control, defect detection, and predictive maintenance.

3. Processes Optimization

AI and ML algorithms optimize the efficiency and quality of additive manufacturing by predicting optimal process parameters (such as temperature, speed, and layer thickness) and analyzing large amounts of data, reducing defects and material waste. Deep learning methods can also simulate the 3D printing process, predict the results of different parameter settings, reduce dependence on physical experiments, while improving sustainability and energy efficiency, and reducing environmental impact.

4. Quality Control and Defect Detection

Quality assurance is a key link in additive manufacturing, especially in the automotive, aerospace, and medical fields. AI, ML, and DL algorithms significantly improve component characteristics such as dimensional accuracy, surface roughness, mechanical properties, and microstructure control by optimizing process parameters such as laser power and scanning speed. ML models based on historical data can predict printing errors, while DL driven computer vision systems can detect defects in real-time (such as porosity and size errors), reducing the need for manual inspection and rework. Real time monitoring and closed-loop systems have further improved manufacturing consistency and quality, with typical cases including GE Aviation's AI optimized fuel nozzle design, which successfully reduced weight and improved performance. Figure 4 shows a real-time defect detection strategy for additive manufacturing processes based on DL and machine vision technology.

Fig. 4. A real-time defect detection strategy for AM processes based on DL and machine vision technologies.

Digital twin technology creates virtual copies of the metal printing process and combines AI, ML, and DL technologies to achieve real-time monitoring and precise control of key factors such as temperature, pressure, and material flow. This synergistic effect not only optimizes process parameters and predicts potential problems, but also improves the quality of printed metal parts, while reducing waste, lowering costs, and accelerating the development of innovative products, promoting the accuracy and efficiency of additive manufacturing. Figure 5 shows the digital twin in 3D printing.

Fig. 5. Digital twin in 3D printing process.

Source: Yangtze River Delta Laser Alliance