Refrigerator failure early prediction based on warranty data

Abstract

This paper provides a method to predict the failure characteristics of refrigerator components based on the early warranty return data from the field (suspended data). The predication will be used to support the management decision of actions. Very commonly, the company doesn't (or has no capability to) record each unit purchase date by the end customers (not by OEM), but does record all units' manufacture dates and returned units' purchase dates for the warranty claims purpose. This makes it difficult for the analyst to estimate how many units are currently used in the field, which is required in order to do a suspended data analysis. This paper chooses a 'sample population' from warranty claim database and uses Weibull distribution to describe the time duration between the manufacture dates and end customer purchase dates for the units in the sample population. Then, assuming this time duration distribution representing the time duration distribution of the 'objective population', and using integration function described in this paper, how many units of objective population are currently used in the field can be estimated. This makes the reliability prediction possible (by processing the field suspended data). An example is used to demonstrate the method.