基于空间结构隐Markov模型的故障诊断

Fault diagnosis based on spatial structure hidden Markov model

针对机械故障诊断领域所引用的隐Markov模型忽略了机构之间空间结构特性的问题,利用机械设备的空间结构性对经典隐Markov模型进行补偿,建立完整的故障诊断模型。该方法提出行为结构基因表达式和机构基因自相关矩阵对空间结构进行描述,以便对机械设备进行故障溯源诊断。通过基因变异操作将机械空间结构分解为多个隐Markov模型,结合隐Markov模型参数算法和机构状态基因操作,快速定位机械故障源并进行故障排除。对机械空间结构进行分解诊断,使模型更符合实际情况,也提高了计算效率。通过空分设备氧气透平机故障诊断的应用实例,表明了该方法的有效性。

The Hidden Markov Model（HMM） quoted in mechanical fault diagnosis was ignored the spatial structure features among mechanical components.To solve this problem,a complete fault diagnosis model was established by compensating HMM with the spatial structure features among mechanical components.Behavior-structure gene expression and mechanism gene self-correlation matrix were proposed to describe the spatial structure features,so that the original faults of mechanical equipment were diagnosed.Through gene-mutation operation,the mechanical spatial structure was divided into lots of HMMs,and by combining HMM parameters algorithm with mechanism-statue gene expression operations,the original mechanical faults were diagnosed and debugged quickly.The improved model was more suitable for actual conditions and computational efficiency was also increased by decomposing and diagnosing mechanical spatial structure.The proposed fault diagnosis model was validated by a real oxygen turbine example and the result demonstrated its effectiveness.