Classification of Seemingly Random Failures Using Similarity Analysis

Companies in the commercial vehicle industry use various quality methods to reduce the appearance of failures in the field. However, increasing reliability, continuous improvement of the products, and additionally, the growing number of possible failure patterns of commercial vehicles lead to the appearance of seemingly random failures. These failures and their causes are not comparable with others considering obvious features at the first sight, but a closer look with expert knowledge enables a grouping of similar failure cases.This paper presents a concept for the classification of seemingly random failures by using the general ideas of case-based reasoning (CBR). Occurred field failures are described with defined features in a so called “failure case description”. All these case descriptions are collected in a data base. The data base has to be analyzed to get an impression of the general failure distribution and the extent of random failures. On this basis, the failure data can be assigned according to logical correlations and similar failure patterns. Further comparison algorithms can be used to assign the pre-grouped failure cases depending on what the user wants to know. By using this concept, the companies can avoid high investment of time and money in the troubleshooting process and realize a more effective way of avoiding random failures.