基于改进狼群算法-深度置信网络(IGWO-DBN)模型的旋风分离器压降预测

针对目前旋风分离器压降计算模型在准确性和实用性上的不足，为更好地指导旋风分离器的结构设计和性能优化，采用深度学习方法对其压降进行了预测。选取了影响压降的7个几何参数，采用深度学习中的深度置信网络（deep belief network，DBN）对旋风分离器压降数据进行预测，并利用改进的狼群算法（improved grey wolf optimizer，IGWO）对DBN模型的初始化权重和偏置参数进行寻优，构建IGWO-DBN组合模型，同时与几种传统计算模型和机器学习模型的预测结果进行对比。结果表明，IGWO-DBN模型在计算精度上优于Shepherd-Lapple模型、Casal模型等传统计算模型，并优于反向传播神经网络（back propagation neural network，BPNN）、支持向量机（support vector machine，SVM）、极限学习机（extreme learning machine，ELM）等机器学习模型，计算效率大幅提升，且具有较好的泛化性和鲁棒性，可用于旋风分离器压降参数的预测。

Pressure drop prediction for a cyclone separator based on an improved grey wolf optimizer-deep belief network(IGWO-DBN) model

Existing pressure drop calculation models of cyclone seperators have insufficient accuracy and practicabitily. In this work, the design parameters of a compressor unit have been determined in order to more accurately predict the pressure drop of a cyclone separator. We chose seven geometric parameters which influence the pressure drop, used the depth of deep belief network (DBN) to predict cyclone pressure drop data, and the improved algorithm of wolves (GWO) initial weights and bias of the DBN model parameters optimization to construct an IGWO-DBN combination model. Comparison of the model prediction results shows that the IGWO-DBN model is superior to the Shepherd-Lapple model, the Casal model and other traditional computational models, as well as to the back propagation neural network (BPNN), support vector machines (SVM), extreme learning machine (ELM) and other machine learning models. The IGWO-DBN model has good generalization ability and robustness, and can be used to predict the pressure drop parameters of a cyclone separator.