Fault Diagnosis and Prognosis Based on Lebesgue Sampling

Bin Zhang and Xiaofeng Wang
Submission Type: 
Full Paper
AttachmentSizeTimestamp
phmc_14_049.pdf497.42 KBSeptember 24, 2014 - 1:12pm

Traditional Riemann sampling based fault diagnosis and prognosis (RS-FDP) algorithms adopt period sampling where samples are taken in a periodic manner. With increases in the volume of sensor data as well as the complexity of algorithms, the bottleneck of RS-FDP is mainly the limited computational resources, which is especially true for applications with embedded technology. On the one hand, since the sampling period is determined according to the worst-case operating scenario, the fault diagnosis and prognosis (FDP) algorithm might be executed even if there is little new information actually present in the measurements. In other words, the algorithm may take greater utilization than it actually needs. This will result in significant over-provisioning of the real-time system hardware. On the other hand, when the fault grows very fast, it is expected to assign more resources to the FDP algorithm so that it can takes more frequent actions to provide accurate fault information, which obviously cannot be met by periodic sampling. There is a great need of cost-efficient FDP approaches where computation can be executed on an “as-needed” basis and this motivates the LS-FDP design in this paper.

This paper introduces the concept of Lebesgue sampling in FDP and creates the fundamental knowledge of Lebesgue sampling based FDP (LS-FDP). Contrast to periodic sampling-based approaches, the computation in LS-FDP will be triggered only when an event takes place, and the prognosis will be executed based on the Lebesgue sampling-based model whose states are predefined according to the quantization level. With the feature of “execution only when necessary” in Lebesgue sampling, the computation efforts in LS-FDP can be significantly reduced by eliminating unnecessary computation when fault growth is slow. On the other hand, when fault grows rapidly, it tracks fault growth more accurately by executing FDP more frequently. This paper discusses the theoretical foundation and systematic methodologies for LS-FDP synthesis and analysis. Experimental results will be presented to verify the efficiency of the proposed approach.

Publication Year: 
2014
Publication Volume: 
5
Publication Control Number: 
049
Page Count: 
11
Submission Keywords: 
Diagnosis and fault isolation methods
Submission Topic Areas: 
Model-based methods for fault detection, diagnostics, and prognosis
Submitted by: 
  
 
 
 

follow us

PHM Society on Facebook Follow PHM Society on Twitter PHM Society on LinkedIn PHM Society RSS News Feed