基于组合模型的滚动轴承故障预测研究

针对轴承故障预测可使用的样本数据少、特征参数信息贫乏且呈现非线性、不确定性等特点,提出一种基于改进灰色GM(1,1)和遗传算法优化的BP神经网络的组合预测模型。首先,根据各单一模型在当前时段的预测误差,通过最小二乘法确定出在未来时段中两种单一模型的权重,然后将预测结果进行加权求和,得到最终的组合模型预测值。该模型既能实现灰色GM(1,1)模型处理小样本的轴承振动数据预测的目标,也能发挥BP神经网络解决非线性拟合问题的优势。最后,将组合模型与各单一模型进行实例数据分析,结果表明组合模型的预测精度为96.63%,比上述子模型的预测结果分别提高了7.84%和6.13%。

Research on Rolling Bearing Fault Prediction Based on United Model

According to the characteristic of fault prediction of bearing containing the data of small sample and effective information is poor,nonlinear and uncertain,a combined forecasting model which is based on the GM(1,1)and BP neural network optimized by genetic algorithm is proposed. Firstly,the weights of each model in the future period can be determined by LSM(least square method)which the prediction error of each model is needed in the current period. Then,the two individual forecast results are united by using the obtained weights to get the predictive value. The model can not only give full play to the advantage of GM(1,1)model with fitting the vibration data of small sample for bearing,but also can make the advantage of BP neural network to solve the non-linear problem. The result indicated that the accuracy of combined model was 96.63%,which was increased by 7.84% and 6.13%,compared with the single model through the practical case.