Scholarlink

<p>In the context of industrial intelligence, equipment faults are becoming increasingly complex. Accurate diagnosis is difficult to achieve with a single sensor, making multi-sensor data fusion a key technology. Dempster–Shafer theory (DST) is widely used in multi-source information fusion due to its advantages in processing uncertain information. However, traditional DST methods are mostly based on static modeling assumptions, making it difficult to dynamically capture the variation of sensor reliability in complex environments. This can easily lead to evidence conflicts, thus reducing diagnostic accuracy. To address this issue, this paper combines the modeling capability of Markov chains for dynamic stochastic processes with the advantages of DST in evidence fusion, proposing a novel fault diagnosis algorithm based on an <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(m\)</EquationSource> <EquationSource Format="MATHML"><math> <mi>m</mi> </math></EquationSource> </InlineEquation>th-order Markov information model. Firstly, by constructing an <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(m\)</EquationSource> <EquationSource Format="MATHML"><math> <mi>m</mi> </math></EquationSource> </InlineEquation>th-order Markov information model, the Markov permutation set is defined, and the transition probability together with the stationary distribution of the Markov permutation set are derived to characterize the dynamic evolution pattern of faults. Furthermore, Markov–Rényi divergence is introduced to dynamically evaluate sensor reliability and assign weights, so as to correct conflicting data. Finally, fault diagnosis is achieved by fusing the weighted stationary distribution using a novel connection sum. Experimental results on three typical industrial scenarios, namely rotating machinery fault diagnosis, drilling process fault diagnosis, and permanent magnet synchronous motor bearings fault diagnosis, demonstrate that the proposed algorithm outperforms traditional diagnosis methods and classical DST-based methods in terms of fault recognition accuracy, leading to significantly improved diagnostic reliability.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A multi-sensor fault diagnosis algorithm based on an mth-order Markov information source model and its applications

  • 期刊论文

  • 发表日期：

    2026年6月1日

查看全文
  • Junshuang Liu,
  • Xiaojian Ma

摘要

In the context of industrial intelligence, equipment faults are becoming increasingly complex. Accurate diagnosis is difficult to achieve with a single sensor, making multi-sensor data fusion a key technology. Dempster–Shafer theory (DST) is widely used in multi-source information fusion due to its advantages in processing uncertain information. However, traditional DST methods are mostly based on static modeling assumptions, making it difficult to dynamically capture the variation of sensor reliability in complex environments. This can easily lead to evidence conflicts, thus reducing diagnostic accuracy. To address this issue, this paper combines the modeling capability of Markov chains for dynamic stochastic processes with the advantages of DST in evidence fusion, proposing a novel fault diagnosis algorithm based on an \(m\) m th-order Markov information model. Firstly, by constructing an \(m\) m th-order Markov information model, the Markov permutation set is defined, and the transition probability together with the stationary distribution of the Markov permutation set are derived to characterize the dynamic evolution pattern of faults. Furthermore, Markov–Rényi divergence is introduced to dynamically evaluate sensor reliability and assign weights, so as to correct conflicting data. Finally, fault diagnosis is achieved by fusing the weighted stationary distribution using a novel connection sum. Experimental results on three typical industrial scenarios, namely rotating machinery fault diagnosis, drilling process fault diagnosis, and permanent magnet synchronous motor bearings fault diagnosis, demonstrate that the proposed algorithm outperforms traditional diagnosis methods and classical DST-based methods in terms of fault recognition accuracy, leading to significantly improved diagnostic reliability.