Research On Life Prediction Methods Of Electrical Product

The working life is an important reliability characteristic of the electrical product. And in this paper, the life prediction methods of electrical products were researched by taking relay as an example. After the life test data of relay was analyzed,a fact was found that there are some relations between the initial working state performance and the working life: Better performance of the initial state, longer life expectancy of the individual relay, and vice versa. Based on this, applying the rough set theory and the grey theory, the relations between the initial state information and terminal life for electrical products were extracted. Then, two life prediction methods for electrical products were put forward. The first method is based on rough set theory and the second one takes advantage of both rough set theory and grey theory.The first life prediction method which is based on rough set theory can be achieved by comparing the life decision rules with the initial performance of the product to be predicted. To get the life decision rules, a few steps are required to be carried out: Firstly, build life decision table using initial state performance parameters as condition attributes and final life value as the decision attribute. Secondly, extract the initial state parameters associated with working life based on the attribute reduction in rough set theory. Thirdly, establish the life decision rules based on the extracted parameters through the rule generation algorithm. In this method, the life of the electrical product to be predicted can be predicted qualitatively.In the second life prediction method which is based on rough set theory and grey theory, two steps are included: establishing a grey forecast model and using the model to realize life prediction. The grey forecast model is based on the first method, taking the reduced performance parameters as input variables and final life value as output variable. When using the model to realize life prediction, the life level and the sorting order in the corresponding life level of the electrical product to be predicted should be firstly determined by using its performance parameters and the life decision rules that established in the first method. After that, the life prediction can be achieved just by putting the initial performance parameters in the grey forecast model. This method achieves quantitative prediction other than "grading forecast" which refers to the first method.Both life prediction methods are analyzed by using relay life experiment data, and the result suggests the established life prediction methods are effective.