基于机器学习的液压摆缸叶片密封性能预测模型

Machine learning based prediction model for sealing performance of hydraulic swing cylinder blades

摘要:

液压摆缸运行时若液压油从叶片与缸体的间隙泄露，会导致压力下降，减少输出扭矩，无法实现预期操作效果。由此，提出一种基于机器学习的液压摆缸叶片密封性能预测模型。利用敏感性与疲劳性指标组建叶片密封性能指标体系，收集密封性能相关指标的样本数据，计算密封泄漏率与密封环热量，利用雷诺方程计算油膜厚度与密封间隙油膜压力的耦合关系；基于机器学习方法中的BP人工神经网络训练数据，将连续S型函数转化成神经元激活函数，全局误差低于设定的极小值，即可获得密封性能预测结果。实验结果表明:所建模型液压摆缸叶片密封性能预测精度高、效率快，为相关领域机械设计工作提供可靠参考。

Abstract:

If hydraulic oil leaks from the gap between the blades and the cylinder during the operation of the hydraulic swing cylinder, it will lead to a decrease in pressure, reduce output torque, and fail to achieve the expected operating effect. Therefore, a machine learning based prediction model for the sealing performance of hydraulic swing cylinder blades is proposed. Establish a blade sealing performance index system using sensitivity and fatigue indicators, collect sealing performance sample data, calculate sealing leakage rate and sealing ring heat, and use Reynolds equation to calculate the coupling relationship between oil film thickness and sealing gap oil film pressure; Based on the training data of BP artificial neural network in the machine learning method, the continuous S-type function is converted into the neuron activation function, and the global error is lower than the set minimum value, then the sealing performance prediction results can be obtained. The experimental results show that the built model has high prediction accuracy and fast efficiency for the sealing performance of hydraulic swing cylinder blades, providing reliable reference for mechanical design work in related fields.