铝基复合材料高速干摩擦行为的遗传神经网络预测模型

Prediction Modeling of Friction Behavior of Aluminum Matrix Composites Using Neural Network and Genetic Algorithms under High Velocity and Dry Sliding Condition

对4种S iC颗粒增强铝基复合材料在5种速度和4种压力条件下进行了销-盘摩擦磨损试验,运用遗传神经网络技术建立了铝基复合材料在高速干滑动过程中的摩擦行为预报模型,并用该模型对铝基复合材料进行预报.结果表明,蓄热能力较大的铝基复合材料在服役条件下具有较高的摩擦系数,与实际情况相一致.用遗传神经网络建立的铝基复合材料摩擦行为预测模型为服役条件下提供了简便、可靠的优选材料方法.

Ever increasing application of discontinuous reinforced aluminum （DRA） composites in braking materials arises from their specific properties. However, it is difficult to describe exactly friction behavior of such composites for reasonable selecting. In the present study, using genetic algorithms and radius basis function neural network （GARBF）, prediction modeling of friction behavior was established based on a measured database for DRA composites under high velocity and dry sliding condition. Friction tests with pin- on-disc arrangement had carried out at five sliding velocities （40, 55, 70, 85, and 100 m/s） and four different nominal pressures （0. 1333, 0. 4667, 0. 60, and 0. 7333 MPa）. Modeling results confirm the feasibility of GARBF network and its good correlation with experiment results. Using GARBF modeling data to predict analysis, results show that friction coefficients of composites increased with increasing stored heat capability. It is proposed that a well-trained GARBF modeling is expected to be very helpful for selecting composite materials under different working conditions, for prediction dynamic tribological properties.