3D printing is not a simple process. It calls for precision positioning and control algorithms that requires mathematical models to make an exact representation of a design. A miscalculation or a wrong input of data and change the outcome. The mistake is costly, not just monetary-wise but also with time. Setting printing parameters can take weeks of computations, formula and adjustment. Everything must be in place before the printing process starts. However, despite careful preparation, printing errors can still happen.
To minimize printing errors, scientists at the Laboratory of Lightweight Materials and Structures of Peter the Great St. Petersburg Polytechnic University in Russia developed a neural network for the 3D printer.
Scientists and engineers have long brought Artificial Intelligence (AI) to the manufacturing sector. We now see collaborative robots, software that are capable to think and smart factories. Taking it a step further, scientists are looking at neural networks to make AI more adept in 3D printing.
A neural network picks up its cue from the brain neurons. It is capable for deep machine learning. Through communication and feedback, a machine will be capable to self-learn. Neural networks differ from AI algorithm by not using task-specific rules.
When a computer uses a neural network, it can develop image recognition abilities, as well as other abilities. For instance, it can recognize the image of a “tree” and “not-tree”. With this ability, the machine can develop a list of differences between a tree and a not-tree. It can use this list in future image recognitions. Later on, it can build a library of data to build its knowledge. This ability is actually how CAPTCHA works.
3D printing helps in the development of neural networks. Recently, researchers in St. Petersburg Polytechnic University in Russia developed a neural network for metal 3D printing. This will allow a much faster and more efficient 3D printing. The neural network allows the machine to learn previously entered data thereby taking away the need for mathematical adjustments for new structures. To add to this, the machine can anticipate flaws during the printing process and make necessary adjustments to amend the flaws.
With the use of the new neural network, the team from the St. Petersburg Polytechnic University was able to develop printing modes in making ship mastheads. The scientists tested the neural network to measure the quality of laser melting. They also measured the quality of the parts as well as the strength of the welding process. The next step in this endeavor is to create an online system based on the neural network that will automatically input data that can create new parameters. This way, the system will have a continuous learning process. In the manufacturing sector, this will help a lot in developing the quality of parts made as well as speed up the process in making these parts.