基于贝叶斯网络的铝型材挤压过程异常检测

针对挤压机设备异常原因复杂、异常检测精度低、检测方法时效性不足等问题,提出基于贝叶斯网络的铝型材挤压机生产过程异常检测方法。充分利用贝叶斯网络对于解决不确定性问题的优点和挤压机设备运行异常时会体现能耗数据异常的特点。分析挤压过程的能流机制,构建逻辑结构清晰、冗余低的贝叶斯网络结构。以挤压机历史能耗数据作为训练样本,进行仿真实验,可以准确地发现异常并定位导致异常发生的原因。实验结果表明,该方法在实际应用场景中有较强的可操作性,对于挤压机异常检测问题有实际的意义。

ABNORMAL DETECTION OF ALUMINUM PROFILE EXTRUSION PROCESS BASED ON BAYESIAN NETWORK

Aiming at the problems of complex reasons for abnormal extrusion equipment, low accuracy of abnormal detection and insufficient timeliness of detection methods, this paper proposed an abnormal detection method of aluminum profile extrusion presser production process based on Bayesian network. We made full use of the advantages of Bayesian network to solve the uncertainty problem and the abnormal operation of the extruder equipment would reflect the characteristics of abnormal energy consumption data. We analyzed the energy flow mechanism of the extrusion process, and built a clear logical structure with low redundancy Bayesian network structure. The historical energy consumption data of the extruder was used as a training sample to carry out simulation experiments, which could accurately find the abnormality and locate the cause of the abnormality. The experimental results show that the method has strong operability in practical application scenarios and has practical significance for the abnormality detection of extruders.