Abstract:Performance degradation assessment (PDA) is an intersection technology of machine condition monitoring, degradation evaluation, and fault diagnosis in condition-based maintenance (CBM). A commonly used strategy of PDA is to use a health index (HI) to detect an incipient fault for machine condition monitoring and subsequently to track monotonic performance degradation trajectories. Sparse representation, as a classical signal processing method, has been widely studied for machine fault diagnosis while its extension to PDA is seldom reported. Based on the sparsity property of fault signals in the frequency domain, a novel sparse representation degradation modeling methodology that integrates sparse representation and monotonic degradation modeling is proposed in this paper to enrich PDA. At first, the sparsity property of square envelope spectra is thoroughly investigated. Based on the sparse representation theory, a HI is designed as the sum of weighted atoms. Besides, the amplitudes of square envelope spectra are used as a dictionary. Then, a sparse representation degradation model is proposed to realize the monotonicity property of a HI and the sparsity of weight coefficients. Experimental verification shows that HIs constructed from the proposed methodology are more suitable to PDA compared to existing HIs. Moreover, a prominent advantage of the proposed sparse degradation modeling is that no prior knowledge and complicated signal processing algorithms are required. Moreover, optimized weights are sparsely located at informative frequency bands for subsequent fault diagnosis, which provides informative frequencies for many blind fault diagnostic scenarios.

Performance Degradation Diagnosis and Solution for OpenMP Programs

Hybrid Performance Modeling And Analyzing Of Parallel Systems

Analysis of Factors Affecting Execution Performance of OpenMP Programs

Performance Modeling of OpenMP Program Based on LLVM Compilation Platform

Machine Learning Framwork for Performance Anomaly in OpenMP Multi-Threaded Systems

Using Hardware Counter-Based Performance Model to Diagnose Scaling Issues of HPC Applications.

A Load Monitoring Method and Balancing Strategy at Source Level of OpenMP Programs for Synchronization Segments

Pattern-driven performance optimization at runtime: experiment on JEE systems.

Extended Overhead Analysis For Openmp Performance Tuning

Pattern-driven Performance Optimization at Runtime

Heartbeat Diagnosis of Performance Anomaly in OpenMP Multi-Threaded Systems

Distributed Predictive Modeling Framework for Prediction and Diagnosis of Key Performance Index in Plant-Wide Processes

Novel sparse representation degradation modeling for locating informative frequency bands for Machine performance degradation assessment

On the Parallelization Optimization Strategy for High Performance Computing Software

An Automatic Framework for Detecting and Characterizing Performance Degradation of Software Systems

Understanding Performance Problems in Deep Learning Systems.

Performance Analysis and Optimization of a Hybrid Distributed Reverse Time Migration Application

A Study of Performance Programming of CPU, GPU accelerated Computers and SIMD Architecture

Taming the "Monster": Overcoming Program Optimization Challenges on SW26010 Through Precise Performance Modeling

VAPRO: Performance Variance Detection and Diagnosis for Production-Run Parallel Applications

Proactive bottleneck performance analysis in parallel computing using openMP