Each year in the US, over 700,000 patients receive total knee arthroplasty to restore joint function and improve quality of life. A major challenge during surgery is achieving proper ligamentous balance. Improperly balanced knees can lead to accelerated wear of the articular surfaces, reduced range of motion, and patient discomfort and pain. Currently, surgeons rely heavily on their experience and their interpretation of the "feel" of a balanced knee. The goal of the proposed research is to investigate the use of piezoelectric materials embedded in total knee replacement bearings in order to sense forces in vivo and convert knee loads into usable electrical energy to power the embedded sensor. This paper presents an investigation of the effects of various geometric properties on the performance of piezoelectric transducers embedded into polyethylene knee replacement bearings. This work takes a combined modeling and experimental approach to investigate the effects of the overall bearing geometry as well as placement of the embedded transducers on the performance of the system. A simple cylindrical geometry is chosen to represent the knee bearing in order to isolate various effects. Specifically, the curvature of the upper bearing surface is investigated to determine the effects of different curvature profiles on the voltage output of the embedded transducer. Designs with the smallest diameter of curvature are found to provide increased load transfer to the embedded piezoelectric and larger generated power. Additionally, the radial placement of the embedded piezoelectric device is investigated to determine the performance of the system as the piezoelectric device is translated from the geometric center of the bearing to the outer edge. Results show that optimal performance is obtained for placement near the center of the geometric curved feature. Lastly, the effects of variations in machining and fabrication are investigated and it is found that tight tolerances must be maintained in order to obtain experimental results that can be accurately predicted by the model developed in this work.