Friction properties of three polymers simulated by trained artificial neural network.

The process of developing a tribologically modified polymer materials with excellent tribological performance is a highly time- and material-consuming work. Since the tribological properties of polymer composites are affected by many factors, e.g. surface pressure, slide velocity, ambient temperature, roughness and form of the contact surface, compositions of the material etc. Due to synergetic effects of these factors, it is hard to predict how the tribological properties are altered under distinct load conditions. Therefore, traditional developing process requires a large amount of tribological experiments with different variations of experimental conditions.

In order to elucidate the influences more accurately and intuitively, and more importantly, to reduce the experimental expenditures, artificial neural network (ANN) is utilized. In the first step, ANN was trained and validated by limited amount of measurements for optimizing the network. Afterwards, the optimized artificial neural network enables us to predict the friction and wear properties of the materials studied under distinct load conditions. Thereby, obvious reduction of the experiments can be achieved.