Abstract:

Magnetorheological media in service are subjected to multi-field coupling of temperature rise, magnetic field and sustained shear, whose action induces an unknown and difficult to accurately predict rheological behavior. Due to the problem that back propagation (BP) neural networks tended to converge to local extremes, the prediction model of the rheological properties of magnetorheological grease with high-temperature thermal effect was optimized by establishing the sparrow search algorithm (SSA) optimized BP neural network, which was used for the characterization and prediction of the relationship between experimental temperatures, magnetic fields, thermal effect time, shear rate and the rheological properties of magnetorheological grease. The results showed that at high magnetic fields, the large aggregation of magnetic chains was less impeded by the soap fibers and the magnetorheological grease shear stress increased substantially. The high-temperature thermal effect caused some damage to the magnetorheological grease composite structure, as evidenced by the overall lower shear rate profile of MRG-24h under all experimental conditions. The BP prediction model indicated low generalization performance with negative R² when evaluating data with large local discretization, whereas the SSA-BP prediction model still had high prediction accuracy when evaluating the prediction performance for overall different datasets and data with small and large local discretization. The SSA-BP prediction model can provide a prediction of the rheological properties of magnetorheological grease and data support for the design and development of magnetorheological devices.

Key words: magnetorheological grease, thermal effects, multi-field coupling, shear stress, SSA-BP prediction model

摘要：

磁流变介质在服役过程中受温升、磁场、持续剪切的多场耦合作用，其作用引起流变学行为规律不明并难以准确预测。由于反向传播（BP）神经网络存在容易收敛于局部极值的问题，通过建立麻雀搜索算法（SSA）优化BP神经网络的高温热效应磁流变脂流变性能预测模型，用于表征和预测实验温度、磁场、热效应时间、剪切速率与磁流变脂流变学性能之间的关系。结果表明：高磁场下，磁链大量聚拢受到皂纤维的阻碍较小，磁流变脂剪切应力大幅增加。高温热效应对磁流变脂复合结构造成了一定破坏，表现为MRG-24h在各实验条件下的剪切速率曲线整体较低。BP预测模型在评估局部离散性较大的数据时，出现R²为负值的低泛化性能表现，而SSA-BP预测模型对于整体不同数据集和局部离散性较小、较大的数据进行预测性能评估时，仍具有较高的预测精度。SSA-BP预测模型可以为磁流变脂流变性能预测，并为磁流变器件的设计和研发提供数据支持。

关键词: 磁流变脂, 热效应, 多场耦合, 剪切应力, SSA-BP预测模型