Scholarlink

An aircraft and many mechanical machinery have multiple components crucial for proper functioning, out of which printed circuit boards (PCB) play a pivotal role. A PCB has several components mounted on it, each playing an essential role in the overall functioning. Predicting the remaining useful life (RUL) of printed circuit boards (PCBs) is one of the most important things. In the present study, the RUL prediction on a PCB without any components and a PCB with a capacitor fixed is performed, varying the input acceleration values sinusoidally and randomly. The RUL of PCB is estimated by combining a physics-based approach. This Koopman operator is a mathematical tool to transform non-linear dynamic systems into linear ones and a data-driven approach to enhance RUL prediction accuracy. Performance of various models using LSTM network with sigmoid as activation function, Tanh as activation function, and different types of Koopman operators were performed and compared. A hybrid approach combining Koopman operators and LSTM networks effectively predicted the RUL of PCBs in aircraft systems, providing a robust and accurate method for addressing the complexities of RUL estimation in electronic components.

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

Remaining Useful Life Prediction Using Physics-Based Approach and Machine Learning Techniques

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Venkata Krishna Manohar Bulusu,
  • Chethana R. Gowda,
  • Akashjyoti Barman,
  • Rahul Kumar,
  • Sudha Radhika,
  • Sabareesh Geetha Rajasekharan

摘要

An aircraft and many mechanical machinery have multiple components crucial for proper functioning, out of which printed circuit boards (PCB) play a pivotal role. A PCB has several components mounted on it, each playing an essential role in the overall functioning. Predicting the remaining useful life (RUL) of printed circuit boards (PCBs) is one of the most important things. In the present study, the RUL prediction on a PCB without any components and a PCB with a capacitor fixed is performed, varying the input acceleration values sinusoidally and randomly. The RUL of PCB is estimated by combining a physics-based approach. This Koopman operator is a mathematical tool to transform non-linear dynamic systems into linear ones and a data-driven approach to enhance RUL prediction accuracy. Performance of various models using LSTM network with sigmoid as activation function, Tanh as activation function, and different types of Koopman operators were performed and compared. A hybrid approach combining Koopman operators and LSTM networks effectively predicted the RUL of PCBs in aircraft systems, providing a robust and accurate method for addressing the complexities of RUL estimation in electronic components.