Improving additive manufacturing through networks of learning 3D printers

Additive manufacturing, also known as 3D printing, offers virtually limitless opportunities in the design of parts and materials, because it enables the control of geometry, material composition, and processing conditions at every point in an object. Novel 3D printed materials could achieve properties far beyond those of their constituents through complex and locally varying structure, enabling exciting combinations of properties and functionality. The printing of such new materials could find uses in a variety of applications from 3D printed houses to soft-robotics and medical devices.
However, this great potential comes with complexity and at present significant specialised knowledge is required to effectively operate and manage even simple 3D printers producing parts of just one material. The machines have a wide range of configurable parameters which affect the quality of the desired output and can suffer from numerous error modes due to incorrect configuration and variations in input material or the environment. To address this challenge, we aim to create networks of learning 3D printers, which learn from each other's successes and failures to improve the accuracy and reliability of 3D printing. Initial areas to explore will be error detection, diagnosis and correction in addition to dynamic path re-planning and automatic parameter optimisation for each print. A range of data, such as images, temperatures and tool paths, will be gathered and analysed during the printing process and this will be shared amongst the machines, improving the system after each print. Subsequently, the network of printers could be used to help realise the development of new complex materials by learning as a group how to manufacture a desired material. This would involve the addition of further sensors and testing equipment to the system in order to measure properties of the output materials.
The project will include building 3D printers and working with them to develop new materials and devices in addition to writing software to analyse prints, detect errors and enable printers to communicate. Throughout the project a broad range of research areas will be explored such as machine learning and artificial intelligence, computer vision, sensors and instrumentation, control, machine design and materials science.