Special aspects of estimating the probability of fire occurrence on diesel locomotives of various types

Abstract

Aim. The paper considers the problem of estimating the probability of fire occurrence on diesel locomotives of various types and the ways to solve it. The problem arises due to JSC RZD locomotive fleet special aspects. Thus, the operating fleet presents diesel locomotives designed and constructed in the 20th century as well as in 21st century, and this accounts for different causes of fire owing to design differences. The biggest contribution to differences in fire numbers on new and old type locomotives is made by the construction of a diesel engine as well as the fire resistance of cables. Researches show that substantial difference in fire statistics for diesel locomotives of various types for the same period of observation are caused by the volume of operating diesel locomotive fleet. For instance, volumes of operating fleet for some types of diesel locomotives amount to thousand units (loco-days), while for other types they make up just a couple of hundreds. This raises questions about whether a period of observation and a volume of operating fleet are enough for estimating the probability and what methods should be used to estimate it. Furthermore, we need an interval estimation of probability which is caused by reliability considerations, by getting “the worst scenario”. Again, this is influenced by above differences in types of diesel locomotives. The paper also analyzes the necessity of estimating “the worst scenario” and problems arising in reference with its calculation. To solve the problem of enhancing the reliability of calculations is to calculate the upper boundaries of probabilities. In this case some types of diesel locomotives will have a lower boundary of probability rather than “the worst scenario” as interval estimation. The necessity of such estimation is specified for diesel locomotives of specific designs with materials complying with modern standards in terms of reliability and fire resistance or having scarce statistics for applying approximation methods of calculation because of limited operating fleet. Methods. Researches into statistics of fires on diesel locomotives of types 2TE10, 3TE10, 2TE116, 2M62, TEP70, ChMEZ, TEM2 made by the authors began with application of a “classic” statistics tool, i.e. check of statistical hypotheses about a law of distribution of a random value “fire” belonging to known discrete laws. While at this, a minimum amount of tests was defined for making sure that targeted estimates of probability are of certain reliability. The condition of a diesel locomotive during operation is not stationary, so a classic estimation of probability of fire occurrence would lead to uncertainty in applying the results of estimation for the purposes of planning and prediction. To evaluate “the worst scenario”, we used both precise and approximate methods for defining confidence boundaries based on “double approximation”. Further, to enable transition from estimation of probability of fire occurrence on diesel locomotives of a certain type to estimation of fire probability for certain units, a sufficient amount of rolling stock was researched. The authors have found that the amount of operating fleet should be not less than 610 loco-days to ensure the precision of probability calculation with an error not exceeding ε. The authors have also identified the method and the necessity of separately estimating the probability of fire on locomotives with operating fleet less than 610 loco-days. Results. Conclusions. In fine, for each type of locomotive we have defined a law of random value occurrence, calculated interval estimates of probability of fire occurrence considering an amount of operating fleet. Tools of statistical analysis for calculating probabilities of fire occurrence on diesel locomotives of various types have been also identified. We have determined methods for calculating interval probability estimates taking into account an available amount of observations with an error not exceeding a specified value ε at the level of 0,2p* j . This research and related calculations have enabled us to obtain one of the primary elements for estimating a fire risk, i.e. the probability of fire on diesel locomotives of various types.