Vacuum extraction (VE) is an important obstetric intervention used to assist vaginal delivery during prolonged or complicated labour. Safe and effective use of VE requires technical skill and experience, yet in many low- and middle-income countries (LMICs), access to suitable training opportunities and realistic simulation models remains limited. Existing training simulators are often too expensive, insufficiently durable, or unable to reproduce the anatomical and biomechanical conditions of assisted vaginal delivery. Building on earlier work by Wang, this thesis aimed to improve the VE training simulator by focusing on three main objectives: modularity, durability, and biomechanical realism.To support the redesign, a clinical validation of the existing simulator was first considered together with stakeholder input and previous recommendations by Wang to define the design requirements. Based on these findings, several concepts for the muscle-bone interface were developed and evaluated to improve modularity. A silicone bulb-end connection combined with a sliding PLA cover plate was selected as the most suitable solution, as it enabled disassembly and replacement of individual soft-tissue components without permanent bonding.To improve durability, repeated mechanical robustness testing was performed to identify structural weaknesses during simulated VE procedures. These tests showed that failure was governed by the interaction between geometry, fixation, reinforcement, and loading conditions. Based on the identified failure mechanisms, targeted design changes were implemented in the final prototype. To improve biomechanical realism, different silicone material configurations were evaluated based on the traction forces generated during VE. A hybrid configuration with a DragonSkin ™10 first-layer pelvic muscle and a DragonSkin ™20 levator ani provided the best balance between force realism and repeated-use performance.The final prototype completed 100 consecutive VE cycles without functional failure and generated an average peak pulling force of 92 N. Overall, this thesis demonstrates that a redesign strategy centered on modularity, durability, and biomechanical realism can substantially improve the functional performance of a low-cost VE training simulator.