A hybrid modelling framework for the description of friction forces in autoinjector devices

Autoinjectors (AJs) are medical devices enabling precise subcutaneous self-administration of medicines. An important challenge in modelling these devices is accurately predicting injection time, which significantly impacts device usability, drug delivery and patient experience. Existing literature models for injection time are based on the balance of forces acting on the stopper. These typically include a driving force, resistance due to the flow of liquid through the needle, and friction between the stopper and syringe wall. In some cases, incomplete mechanistic understanding can lead to an oversimplified treatment of friction resulting in poor predictions of delivery time. This study presents a hybrid modelling approach that integrates physics-based and data-driven components including Artificial Neural Networks and Gaussian Processes. By integrating mechanistic modelling with surrogate models, the proposed approach overcomes key limitations of existing models and provides an accurate dynamic representation of the friction forces involved in the injection process. The hybrid models developed demonstrate enhanced descriptive capabilities for friction forces acting between the plunger and the syringe inner walls, offering a more comprehensive understanding of the complex physical phenomena involved in the AJ.