PHM Methods

PHM Methods

PHM Approach

• Physics-based models
• Knowledge-based
• Data-driven models

PHM approach methods (sajidDataScienceApplications2021 link)

• Knowledge Based approach
  • Based on experience, it is classified into three types: a rules-based approach, a case-based approach and a fuzzy knowledge-based approach.
  • Knowledge-based rules are in the format of ‘If-then’ and knowledge obtained from previous experience or concrete problem situations.
  • Drawbacks are difficulty in obtaining accurate knowledge from experience and limited access to experts with knowledge. Therefore, it results in low prediction accuracy.
  • Currently used in data mining techniques for extracting required knowledge from databases
  • Examples: expert system, fuzzy system
• Data-Driven approach
  • Use computational power and a large amount of data
  • Data processing and analyses of big industrial data, the degradation of components, the remaining useful life and maintenance can be mined.
  • Sub methods: statistical models, stochastic models, and machine learning models
  • Example: machine learning, deep learning
• Physical Model approach
  • Based on laws of physics and mathematics for assessing degradation of components
  • Without a lot of data collection, this model can reveal the faulty logic of the system
  • Since most equipment is complex mechanical and electrical systems, establishing degradation models is difficult because of the ignorance of degradation.
  • Example: Kalman filter, Particle filter, FEM
• Digital Twin approach
  • Digital twin (DT) integrates multidisciplinary, multi-probability and multi-scale using physical model, sensor and real-time data
  • DT combines data and models, which creates a bridge between the physical world and the digital world

Reference

• sajidDataScienceApplications2021 link DOI
• gawdeScopingReviewMultifault2022 link DOI