
并行计算的文章已经完成,等会议7月开完了发上来。
可以先发个摘要。
ROLE OF PARALLEL COMPUTING IN DATA MINING FOR CONTAMINANT SOURCE IDENTIFICATION IN WATER DISTRIBUTION SYSTEMS
ABSTRACT
Data mining is demonstrated as a rapid and efficient methodology to identify location(s) of contaminant source(s) in a water distribution system. The key to the method is populating a database containing the array of intrusion events in a reasonably short time, and setting an “m” value for a query sentence for each sensor. To analyze the “m” value used in the query sentence, parallel computing, implemented in the SHARCNET platform, provides an efficient way to simulate the intrusion events of different scenarios under uncertainty in parallel. The “m” value for each sensor is determined as the 95% quantile of the offset values of every intrusion event in the scenarios considered. As demonstrated in the case study for the Goderich water distribution system, parallel computing can reduce the required simulation time significantly from 42 days to only 15 min allowing extensive investigation; the ability to store more data scenarios can reduce both false negative rates in efforts to identify the correct intrusion node, as well as the false positive rates.
Keywords: parallel computing, contaminant source identification, uncertainty, EPANET, false positive, false negative
Do everything simple, but not simpler.