Pattern Analysis of Failure Detection Based on the Machine Performance Data (Case Study: Siemens Locomotives of Iran Railway)

Failure detection of complex industrial equipment such as locomotives is a challenging tack in maintenance process. Failure to timely diagnose the cause of locomotive failure, in addition to reducing access time, will cause disruption to the rail network, increase excess train stops and other adverse events. With the advances made in recent years, the amount of data stored in new locomotives is increasing. In different ways, the knowledge contained in the data can be extracted and used to increase the productivity of the organization. By pattern analysis of failure detection of locomotives, the cause of many failures can be discovered and repair time can be reduced. Among the available methods, data mining techniques can be mentioned. The present study uses data mining and Apriori algorithm to discover meaningful rules from the data in locomotives, with the aim of improving the efficiency of the failure detection process. The result of this study is the discovery of 20 frequent occurrence in passenger locomotives, 18 two-component laws and 2 three-component laws, and its key achievement is the improvement of locomotives maintenance in a short period of time in the Railway of the northeastern region of Iran. Use of the discovered laws in the future can eliminate many out-of-schedule stops of passenger trains and also significantly reduce the cost of locomotive maintenance. It is estimated that using only one of the mentioned laws alone can reduce the cost of Iran's railways by about ten billion rials per year.