Prediction of Tribological Characteristics of Biomaterials Using Artificial Neural NetworksPages 1-9
M. Arulkumar Abstract:
Biomaterials are increasingly important in orthopaedic implants, particularly in Total Joint Replacement (TJR) procedures. Research has been conducted to identify suitable biomaterials for different joint replacements, with Total Hip Joint (THJ) and Total Knee Joint (TKJ) being the most common. Promising biomaterials for TJR include metals (such as Ti6Al4V, CoCrMo, and 316L SS), ceramics (like Al2O3), and polymers (such as UHMWPE). These materials are valued for their high strength-to-weight ratio, biocompatibility, corrosion resistance, and wear resistance. In this study, artificial intelligence was used to predict the tribological properties of these biomaterials. Experimental trials were conducted using a pin-on-disc tribometer to validate the artificial intelligence-based prediction of tribological properties, varying parameters such as applied load (40 N, 60 N and 80 N), sliding distance, and sliding velocity. With actual properties determined through machine runs. Experimental trials through a pin-on-disc tribometer were used to find the tribological properties of these biomaterials by varying applied load Sliding distance (1000 m, 1500 m, 2000 m) and sliding velocity (1.25 m/s, 2.1 m/s, and 3.2 m/s). Response Surface Methodology (RSM) was employed to predict the frictional behaviour of biomaterials by varying these input parameters. The predicted values were then compared with experimental results, showing that RSM could accurately predict the friction behaviour of the biomaterials with 95% accuracy. These findings demonstrate the potential of RSM as a predictive tool in biomaterials research, which can contribute to optimising tribological properties in medical implants and devices. This could lead to advancements in the design and application of orthopaedic implants.
Keywords: Machine learning,
Biomaterials,
Response surface methodology,
Sliding dry wear test,
Coefficient of friction
|
PDF 