Artificial intelligence (AI) refers to computer-connected devices that mimic human intelligence. There are currently additive manufacturing (AM) applications in the food, chemical, aerospace, automotive, and healthcare industries, among others. Perhaps the most significant advantage of 3D printing is that even complex objects can be created based on the customer’s specifications. It is better suited for small-scale production in the current state of affairs. The stages of additive manufacturing are 3D model preparation, component prototyping, and component production. The goal of the prefabrication stage is to determine whether it is technically possible and feasible to print a given 3D model. When 3D printing with artificial intelligence is used, it is also referred to as smart manufacturing. As a result of smart manufacturing, productivity would rise. The global 3D printing market is expected to reach $6 billion by 2022, with the most growth opportunities for businesses in the home improvement and life sciences industries. Although the additive manufacturing process has made significant advances in recent years, there are still a number of challenges to overcome before it is widely adopted by the industry. To achieve a reasonable level of print accuracy in additive manufacturing, for example, numerous and complex variables must be monitored and controlled throughout the process. Experimenting with different lattice positions or designing appropriate support structures is not a long-term or time-efficient way of determining the best configuration. Machine learning is currently being used to solve this problem in the pre-fabrication stage, through generative design and testing, to increase printing efficiency and cost savings while also improving print quality. Artificial intelligence is currently being used in 3D printing and additive manufacturing to develop intelligent service-oriented production processes for the industrial sector. Because of the ever-expanding data repository, it offers a diverse set of algorithmic, theoretical, and methodological options that have the potential to improve current manufacturing standards. Artificial neural networks, machine learning, adaptive neuro-fuzzy inference systems, evolutionary algorithms, and so on are examples.
2.1 Artificial Neural Network Algorithm
The 3D printing process includes model design, material selection, printing, and part evaluation and characterization. This section combines the use of ANN for 3D printing process monitoring, design, and correlation between process parameters and final component characteristics.
2.1.1 Process Surveillance
Process monitoring via various sensors provides direct information regarding quality supervision and control during the printing process. Three distinct types of data sources can be distinguished: (a) one-dimensional data sources such as spectra; (b) two-dimensional data sources such as graphs and images; and (c) three-dimensional data sources such as morphologies. While one-dimensional data is faster to process, it contains less information than two- and three-dimensional data.
2.1.2 Designing
Design is a critical area of research that necessitates a thorough understanding of the capabilities and limitations of 3D printing technologies. It is the initial and most critical step in the workflow process. A well-designed CAD model not only ensures printability but also minimizes the amount of support material required. The design process, on the other hand, is typically iterative and time-consuming. A data-driven approach to 3D printing design would aid designers in their work.
2.2 Machine Learning (ML)
As shown in Figure 1, it is beneficial to break down the applications into the pre-processing, process, and post-process for the application of AI in AM, given the complexity of the process. Design space can use ML during the pre-processing stage (geometrical design, topology optimization, raw materials design, and powder properties). Predicting material properties is now possible thanks to advances in machine learning (ML) in raw material design. Designing new novel materials is also made easier with it, and it can take advantage of AM’s unique manufacturing capabilities to bring designs to life that otherwise would not be possible. Even though machine learning has been applied to design space (geometric design and topology optimization), it has not been applied to powder properties, which are the least explored areas. The process itself categorizes ML applications for experimentation in-process monitoring and optimization, and simulation work in the same area. One of the areas that have seen much research is experimental process monitoring and optimization.
Figure 1. AM powder bed manufacturing process
2.3 Adaptive Neuro-Fuzzy Inference System (ANFIS)
One of the soft computing techniques, adaptive neuro-fuzzy inference system (ANFIS), plays a significant role in accurately modeling input-output matrix relationships. Because of the nonlinearity of the 3D printed PLA process, ANFIS is an excellent tool for predicting weight gain based on input variables. ANFIS has also been utilized in the modeling and evaluation of metallic slurry erosion.
2.4 Evolutionary Algorithms (EA)
EAs are widely used in the AM domain to solve complex multiobjective design, process planning, and machine setup problems. Future applications are expected to benefit from increased computational power, allowing for more sophisticated and precise new algorithms. The increased exploration of EAs in the AM domain is sparked by the complexity created by combining additive and subtractive technologies. Finally, machine learning-based closed-loop process control and optimization could be critical in the drive to industrialize additive manufacturing. If additive manufacturing is going to be widely adopted in manufacturing, it must produce high-quality, repeatable parts. Engineering analysis (EA) tools are critical in the optimization processes required to achieve this goal, with collaboration between academics and industry serving as the final link in the chain.
Summery:
Artificial intelligence (AI) implementations have a lot of potential for additive manufacturing. Despite some progress, AI applications in streamlining additive manufacturing to be integrated into other manufacturing techniques or become a commodity for users are still a long way off. AI can assist additive manufacturing in a variety of ways, including design correlation, design improvement, defect reduction, and microstructural design. The main stumbling block right now is the availability and reliability of data needed for training. The available experimental data has a wide range and is not always open to the public, thanks to the research community or AM manufacturers; the available experimental data has a wide range and is not always open to the public, thanks to the research community or AM manufacturers. The integrity of the data is critical in the development of AM-specific AI algorithms. Experiments have been carried out or are currently being carried out with a wide range of observations and outcomes. As a result, establishing a data storage facility is critical for the manufacturing industry. To work properly for AI algorithms, the data generation condition must be disclosed along with the data. To recreate label the data, information such as process parameters, exact details of raw materials such as powder or feedstock, composition, raw material properties such as flowability of powder, particle size, and so on, as well as the type of machine, should be disclosed and shared. Furthermore, the difficulties of the process, such as high temperatures or high speeds, make it difficult to monitor and measure. The majority of currently available technologies rely on thermal or optical imaging of the material’s surface, and depth information is rarely available. Companies and academics must collaborate to develop tools that track a wide range of process conditions and parameters accurately and quickly. It’s still not possible to detect defects, create a 3D image of the build while it’s being built, or keep track of the microstructure and grain orientations, for example. These areas would undoubtedly present some challenges that a motivated and willing community could overcome.
For more details to check out the Book, Applications of Artificial Intelligence in Additive Manufacturing
https://www.igi-global.com/book/applications-artificial-intelligence-additive-manufacturing/271276
