A method for the reduction of the computational cost associated with the implementation of particle-filter-based failure prognostic algorithms. (1st January 2020)

• Record Type: Journal Article
• Authors: Rozas, Heraldo; Jaramillo, Francisco; Perez, Aramis; Jimenez, Diego; Orchard, Marcos E.; Medjaher, Kamal
• Is Part Of: Mechanical systems and signal processing. Volume 135(2019)
• Publication Date: 2020-01-01
• Subjects: Prognostic algorithms -- Time-of-Failure probability distribution -- Online performance assessment; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2019.106421
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: Reduction of computational cost of prognostic algorithm implementations. Novel method determines the need of updates in prognostic algorithm results. Novel fast-computing online prognostic performance metric is presented. Proposed strategy validated on battery discharge time prognosis problem. Abstract: Failure prognostic algorithms require to reduce the computational burden associated with their implementation to ensure real-time performance in embedded systems. In this regard, this paper presents a method that allows to significantly reduce this computational cost in the case of particle-filter-based prognostic algorithms, which is based on a time-variant prognostic update rate. In this proposed scheme, the performance of the prognostic algorithm within short-term prediction horizons is continuously compared with respect to the outcome of Bayesian state estimators. Only if the discrepancy between prior and posterior knowledge is greater than a given threshold, it is suggested to execute the prognostic algorithm once again and update Time-of-Failure estimates. In addition, a novel metric to evaluate the performance of any prognostic algorithm in real-time is hereby presented. The proposed actualization scheme is implemented, tested, and validated in two case studies related to the problem of State-of-Charge (SOC) prognostics. The obtained results show that the proposed strategy allows to significantly reduce the computational cost while keeping the standards in terms of algorithm efficacy.