Reliability evaluation of propulsion system of tunnel boring machine

Abstract

The propulsion system, comprising four hydraulic cylinders, is an important power supply assembly in tunnel boring machine, and fatigue is the major failure modes to concern for the hydraulic cylinders. To evaluate the reliability of the propulsion system, i.e. the hydraulic cylinder assembly, a reliability model capable of completely reflecting the effects of stochastic load environment and material property degradation is required. For such a purpose, this paper presents a method to describe the uncertainty of random load history at two levels, i.e. the macro-level characterizing the variation in the whole load history intensity over individual load historys, and the micro-level describing the fluctuation of load peaks inside one load history. Starting from random load history expression, a multi-layer, multi-variate dynamic (time-dependent) reliability model is developed through system-level load-strength interference analysis. By the multi-layer configuration, the model hierarchically expresses the role of the uncertainty in the intensity of load history, the randomness in load peaks within a specific load history, the uncertainty of load action number, and the uncertainty of component strength. By utilizing discrete time variable (i.e. load action number), product property degrading with load cycles can be exactly incorporated in failure criterion. The model, taking on either the form of multi-layer integral or the form of multilayer summation, can synthetically reflect the effects of the uncertainty of random load environment including statistical risk over operation period, the dependence among component failures, the uncertainty of load action number in a specified time interval, the uncertainty of material strength and its deteriorating path. Besides, this model incorporates both fatigue criterion and ultimate tensile failure criterion uniformally, and is applicable to both mechanical component and series system. Since no assumption on the independence between component failures, the reliability model can naturally reflect the effect of failure dependence, and the (conditional) probability distributions of the individual variates involved in the models can be easily determined.