Research Of Fault Prognostic Of Primary Air Fan Based On Data Mining

Fan is one of the most important auxiliary equipment of coal-fired power plants,and its running state directly decide whether or not a power plant can be safe and economicly operate.Poor working conditions of fan lead to be prone to fault.The current management of the fan is limited to low level stage,such as regular inspection,downtime maintenance and post-processing.Combined with the influence of economic environment,power plants are pressed for improving efficiency and reducing costs of equipment maintenance.Therefore,this paper proposes a method to fault prognostic basing on data mining technology.This method can find fault in early time,at that time,unit operation is not affected by fault.So,win time to make maintenance plan in a timely manner.In this paper,a primary air fan as a case study,obtain the historical data from PI database of SIS system for data mining.First of all,interpolation and discretization was carried out on the field data processing.Create a database under normal condition and prepare for data mining.Next,mining association rule of Boolean type with algorithm,establish a rule base as a fan model under the normal state,and select fault data for verifying availability of rule base.For defects in the process of dividing interval,introduce the fuzzy method to blur mining process,so as to avoid the previous information loss resulted from division of intervals.Also,through mining the fuzzy association rules,the establishment of fuzzy association rules library as the early warning model advance the warning time,play the optimization effect.Association rules based on the operation data,can accurately reflect the relationship among the measurement points by experimental verification.When the fault began to form,the original relationship among points gradually be broken.Its appearance is that the degree of test data matching for association rules is lower.The method can find problem in early time before traditional monitoring system give overrun alarm so that have a good warning effect.