Optimization of Rolls-Royce gas turbine components machining using artificial intelligence

¹ University of Derby, College of Science and Engineering, Markeaton Street, Derby, DE22 3AW, UK.
² Rolls-Royce plc, Derby, UK

^* Corresponding author: Sam.kemp@rolls-royce.com

Abstract

Industry 4.0 has changed the ways in which Small to Medium (SME’s) and Large Enterprise (LE’s) manufacturers and businesses operate. Artificial Intelligence (AI), Bigdata, Edge Computing, Cloud Computing, Internet of Everything (IoE), Fifth Generation (5G) and Information Communication Technology (ICT) allow processes to be optimized, controlled, and monitored in close real-time. These enabling technologies allow manufacturing facilities to collect enormous amounts of data from process lines, such as real-time measurement data, machinery state of health, and cycle time, to accurately plan and report the state of both processes, machinery, and final products. In this work, the research programme focuses on the collected data from process input variables that can be monitored to ensure process outputs and final components conform to design specifications. Current methods of analyzing data, especially in aerospace manufacturing environments, require engineers or process operators with a precise and high skill set to be able to map the results into the appropriate chart and interpret these results. Furthermore, the quantity of data analysis required to monitor process inputs in real-time renders conventional analysis techniques unfeasible. Current development using AI has shown the potential and the capability to detect trends in large data sets using machine and deep learning. It also enables a more precise and automated analysis without human intervention. This paper focuses on determining the feasibility of using a deep neural network (i.e., deep neural networks -DNN) to predict process outputs based on process inputs. A simplified example is presented using data collected from firing a ‘statapult’ with varied configurations, training a deep neural network and predicting future results based on process inputs. Initial results are presented, and the results show promising estimating and prediction capabilities.