Analysis of randomly occurring high injection forces in an insulin delivery device

Abstract

During manufacturing scale-up of a new product, new failure modes often surface which require corrective action. However, as production numbers of an insulin injection device pass 200 million per year, testing continues to find sub-assemblies with too-high injection forces, seemingly at random. Up until now, no corrective action has been effective in preventing these problems. These non-conforming sub-assemblies cause batches to be rejected, reducing the production yield at Sanofi's Site Frankfurt Devices (SFD) production facility. This thesis describes the current state of rejected batch problem solving and explores the application of new methods to better understand the root problems and improve the process. Frequency spectra analysis of testing data using the Fast Fourier Transform (FFT), combined with device physics, identified the key interaction points within the sub-assemblies. This model of part interactions has been verified through testing of purpose-built defective pieces and examination of defective parts. The verified model was then used to identify which components within sub-assemblies cause non-conformances. The root causes of several failure codes were determined, and results were further confirmed by rebuilding and retesting subassemblies with the identified problem components. These results confirm the usefulness of this novel application of frequency analysis to a new field of industrial troubleshooting. Various improvement and control methods are explored and next steps recommended for Sanofi to further improve quality control processes and thereby the production yield. The opinions expressed herein are solely those of the author and do not necessarily reflect those of Sanofi S.A.