Similarity Based Methods for Faulty Pattern Detection in Predictive Maintenance

Scalability, Explainability and Performance of Data-Driven Algorithms in Predicting the Remaining Useful Life: A Comprehensive Review

“…As a result, adequate infrastructure [270], and highly optimized algorithms that can handle such large volumes of (potentially imbalanced) data [17] are required to address this challenge. Creating a database of prototype patterns for healthy and faulty states of the system can address this issue and increase the accuracy and performance of identifying faulty patterns [272,273,274,275].…”

“…Detecting a similar fault in instances that belong to different loading conditions or in systems of different ages might show dissimilar signals. As a result, using the complete data on all instances with different degradation behaviors can reduce the model's accuracy [272]. The PdM needs to be accurate enough to minimize the number of false positives (i.e., instances where the system is functioning properly, but a fault is flagged) or false negatives (i.e., instances where there is a credible fault, but it is gone unnoticed) [6,15,269].…”

“…Creating Prototypes Based on Large Datasets. Most similarity-based algorithms consider point-to-point similarity rather than a curve-to-curve comparison, or better yet, pattern-to-pattern comparison, to account for the temporalinter-dependency of sensor data [272]. Choosing the proper similarity method is highly dependent on the type of data being handled.…”

“…Identifying the cascading effect of multiple faults in complex systems is one of the future challenges of maintenance evolution. As an example, a vehicle's recorded data, with its global structure differing from one driving session to the next, will vary significantly; this means that each driving session will contain a combination of stochastic patterns dictated by the path, road conditions, and the style of driving [272]. In such a system, the data captured from different sub-components may indicate faults, but isolating these faults is a challenging feat.…”

Scalability, Explainability and Performance of Data-Driven Algorithms in Predicting the Remaining Useful Life: A Comprehensive Review

“…As a result, adequate infrastructure [270], and highly optimized algorithms that can handle such large volumes of (potentially imbalanced) data [17] are required to address this challenge. Creating a database of prototype patterns for healthy and faulty states of the system can address this issue and increase the accuracy and performance of identifying faulty patterns [272], [273], [274], [275].…”