By The Business Analyst Team
Aerospace Manufacturing / Quality Improvement
Profile: Worldwide leading manufacturer in aircraft propulsion and equipment, space and defence.
The aerospace industry is continuously searching to reduce weight. Jet engine designs are increasingly using composite materials, as this provides important benefits such as reduced fuel consumption, noise levels and maintenance costs.
The level of expertise required to produce parts from composite materials takes years of training and specialised people, not to mention advanced technologies and processes. Our client had been making aircraft parts out of composite materials for years. For one part in particular, a high replacement part used to channel fresh air to the engines of a fighter jet, our client had reached only a 50% first-time-yield rate - even after years of painstaking efforts to improve the production process. This meant that half of all these parts either had to be entirely scrapped or reworked, which generated significant costs in addition to missed revenues due to the resulting delivery delays.
Rejection or acceptance of this part is decided based on the results of a non-destructive ultrasound analysis. The part is produced by layering resin-impregnated sheets of fiber-reinforced fabric that are cut and formed to the desired shape and then baked under vacuum at a succession of different temperatures for varying lengths of time. After allowing the part to cool, the part is again baked before undergoing quality control.
The challenge is to find the right combination of process parameters to avoid delamination of the sheets and formation of internal blistering. It’s a complex combination of temperatures, durations, and heating and cooling rates that must be individually adjusted to factors such as resin composition and the resin/fibre ratio.
Working with the MondoBrain team, the client quickly secured the data sources from past production runs, including whether the part had passed quality control or not, information on the parameters under their control (for example, baking durations and temperatures, cooling and heating times) as well as information on parameters not under their control (for example, supplier supplied data on the composition of the raw materials).
The team brought this data into the MondoBrain platform, manually added their known constraints and leveraged the AI-powered software to generate prescriptive recommendations on what the operating parameters should be for their next production run. Rather than predicting whether or not given courses of action will be successful, MondoBrain allowed the team to precisely define what is needed to attain success - based on the historical data available.
By using this process of refining results and updating new production data to refine further, the team is increasingly learning how to produce these composite parts for fighter jets at higher and higher success rates. To date, they have improved the acceptance rate of those parts from 50% to 88% - an enormous gain in performance.
They continue to improve on this achievement as a result of the increased understanding they’ve gained working with MondoBrain. Not only does MondoBrain suggest precise process parameters but each time they implement them and observe the results, they are challenged to go further in understanding what contributed to the results observed, thereby improving their ability to boost performance even further.