基于标准齿轮减速箱的故障混合预测分析模型

Fault Mixing Prediction Analysis Model Based on Standard Gear Reduction Box

在制药生产线中,搅拌器是主反应器的主要组件,体积最大,结构相对复杂,与其他设备存在串联关系。如果发生故障,一方面需要停掉整个生产线对其进行故障诊断与维修,从而造成非计划停机的设备闲置损失;另一方面,突发性的非计划停机,会造成设备中的化学材料反应物质浪费。针对此问题,提出一种基于标准齿轮减速箱的故障混合预测分析模型。通过对振动信号进行特征提取以及变分模态分解,将转换的频域特征数据和幅值数据按照时间序列融合,并将数据分为故障数据、设备带病运行数据以及设备健康运行数据3类,通过自注意力网络层进行特征提取。实验表明,该模型可以准确预测设备的故障,且在测试集的准确率达到83.61%,验证了实验的有效性与优越性。

In the pharmaceutical production line,agitator is the main component of the main reactor,it has the largest volume and relatively complex structure,and tandem relationship with other equipments.If a fault occurs,on the one hand,it is necessary to stop the entire production line for fault diagnosis and maintenance,resulting in idle loss of unplanned shutdown of equipment,on the other hand,sudden unplanned shutdown also causes the chemical material reaction substances in the equipment to be wasted due to insufficient reaction.In order to solve this problem,a fault mixing prediction analysis model based on standard gear reduction box is proposed.It fuses the converted frequency domain characteristic data and amplitude data according to the time series by feature extraction and variational mode decomposition of vibration signals,and divides the data into three categories:fault data,equipment sick operation data and equipment health operation data.Features are extracted through the self-attention network layer.Experiments show that the model can accurately predict the failure of the equipment,and the accuracy rate in the test set reaches 83.61%,which verifies the effectiveness and superiority of the experiment.