The Application Of Two Value Image In Fault Fuzzy Diagnosis In Electrical Equipment

Fuzzy diagnosis, which is one of the common diagnosis method of electricequipment fault diagnosis, is on the basis of fuzzy set theory to build a sort ofmapping relationship between Fault symptoms space and Fault state space anddiagnose fault according to those symptons. However, for fuzzy set theory hasnot mature yet, there are no unified way to confirm Element membership andmapping relationship between two fuzzy sets, those elements are confirmedaccording to lots of tests and experience. Meanwhile, because of the fuzzyinformation (eg.dependence is large and complex)and the uncertainty of thesystem itself, along with the determination of bound and appropriateMembership Function of every symptom and characteristic parameter ,themethod has boundedness when applying. This research provided a new methodof membership calculation in fuzzy diagnosis, which combined characteristic ofbinary image, advantages of radar map and image processsing tool in Matlab,used graphical analysis and calculational methods to work out elementmembership of fault symptons, then realized on line monitor and fault diagnosisof electrical equipment.Example 1 is aiming at unviewable elements(eg. Temperature, gas content,current,etc.) of electrical equipments fault characteristic. Nine model faultcharacteristic of transformer in industrial standard has been researched onmembership, radar map is used to display multi elements characteristics in falutcharacteristic synthetically. Draw data of 5 charateristic gas in transformer oilon the radar map, and then do the image similarity calculation between detectingradar map and 9 model fault charateristic, take the similarity as the membershipto do fuzzy diagnosis research.Example 2 is aiming at the viewable elements(eg. Icing thickness of line)of electrical equipments fault characteristic. The adoptive image is an ordianryicing colour RGB image, then use Matlab image processing tools to dobackground correction, Gausian smooth, contrast increase, edge detection andimage segmentation ,etc. to pick up the icing information and transform it intobinary image to calculate its membership between the image and model icingfault. The results of the study shows that the two methods both have certainfeasibility, and can be taken as examples for electrical equipment fuzzydiagnosis; and research result of example 2 can apply to icing on line monitor.