This thesis discusses ongoing research regarding the merging of a reliability assessment model based on Genetic Algorithms (GAs) namely GenRel with statistical maintenance analysis. To verify the feasibility of this thesis methodology, failure data from a fleet of 13 Load-Haul-Dump (LHD) vehicles operating in a typical underground mine in the Sudbury mining region, in Ontario, Canada was used to test the methodology.

The purpose of this thesis research is to investigate whether GAs can be applied to predict future failure data of the LHD vehicles from an initial population of failures. Data collected from the LHD fleet is in the form of Times Between Failures (TBF) and Times To Repair (TTR). The reason for choosing GAs for failure pattern prediction is that:​ GAs are a class of evolutionary algorithms which imitate the biological evolution procedure such as reproduction, selection, crossover and mutation in search for optimum solution. The reliability of mining equipment changes over time due to several covariates/factors (e.g. equipment age, the operating environment, number and quality of repair). These factors affect the equipment's failure patterns and they create complex impacts on the equipment's reliability characteristics. These impacts are assumed to follow a biological evolution process.

A statistical maintenance analysis procedure was used first to evaluate the maintenance characteristics of the LHD vehicles. After the maintenance analysis, a Genetic Algorithm based reliability assessment model was used to perform failure pattern prediction for the LHD vehicles. GenRePs preliminary results were based only on the use of the Exponential and Lognormal distribution functions (Vayenas and Yuriy 2007). In this thesis, an improved computational logic in GenRel is described and tested to assess the model's applicability.

The final results demonstrate a successful application of the thesis methodology for reliability assessment of the LHD vehicles. These results can contribute to improvements in mine maintenance scheduling and planning, and at the same time, generate future failure patterns which can be used in simulation modeling of an entire mining system. Future studies should involve testing this thesis methodology's applicability for several types of mining equipment, and applying GenRePs results in discrete-event simulation modeling of entire mining systems.