Reliability Analysis And Optimization Design Of Load-sharing Redundant System

Redundancy technique is widely used to improve the reliability of the system inengineering applications. And the prediction, evaluation and analysis of the reliability ofredundant systems become the top priority for redundancy design. Traditional reliabilitytheory of redundant systems fails to consider the mutual influence between the lives of theunits. Therefore, it will inevitably bring relatively larger deviations when predicting thereliability of load-sharing redundant systems, and is unable to meet the engineeringrequirements. At present, reliability theory of load-sharing redundant system is still in thedevelopment stage, and there are many problems to be solved for the load-sharing redundantsystems in engineering. Considering the engineering requirements, this dissertation mainlyinvestigates the reliability analysis theories of load-sharing redundant system includingcontroller, load-sharing redundant system consisting of units with degradation failurefollowing random process, load-sharing redundant system containing common-cause failure,and the optimization design of series-parallel system containing load-sharing redundantsubsystems. The main contributions of this dissertation are summarized as follows:1. Reliability model of load-sharing redundant system including controller(1) For the non-repairable system consisting of units following exponential lifedistribution, reliability models are built according to the properties of Homogeneous MarkovProcess considering the controller following exponential life distribution or having a constantreliability. The Monte Carlo simulation method is presented considering the controllerfollowing arbitrary life distributions.(2) For the non-repairable system consisting of units following arbitrary life distributions,reliability model is built using Events Space method according to the theory of conditionalprobability and total probability, and the Cumulative Damage model. The Monte Carlosimulation method is then presented. The reasonableness of the model and the effectiveness ofthe simulation method are demonstrated by an application example of a pump drainagesystem.(3) For the repairable system consisting of units with lifetime and maintenance timebeing both exponentially distributed, system state transition diagram is made and numericalcalculation of Markov Chain is used to work out the transient reliability characteristics ofsystem. The reasonableness of the model is demonstrated by an application example of aconverter valve system.2. Reliability model of load-sharing redundant system consisting of units withdegradation failure following random processFor the system consisting of units with degradation failure following Gamma Process orWiener Process, using Events Space method, reliability models are built separately based onthe independent increments properties of Gamma Process or Wiener Process. The Monte Carlo simulation methods are presented separately to simulate system failures and validate thereliability model.The reasonableness of the model and the effectiveness of the simulation method based onGamma Process are demonstrated by an application example of a V-belt transmition system.The reasonableness of the model and the effectiveness of the simulation method based onWiener Process are demonstrated by an application example of a fan ventilating system.3. Reliability model of load-sharing redundant system containing common cause failure(1) For the system consisting of units following exponential life distribution, reliabilitymodel considering β factor common-cause model is built according to the properties ofHomogeneous Markov Process; after system state transition diagram being made thereliability index of system considering binomial failure rate (BFR) common-cause model isobtained using the numerical calculation of Markov Chain.(2) For the system consisting of units following arbitrary life distributions, reliabilitymodel considering β factor common-cause model is built using Events Space methodaccording to the theory of conditional probability and total probability, and the CumulativeDamage model, and the Monte Carlo simulation method is then presented; reliability index ofsystem considering binomial failure rate (BFR) common-cause model is obtained using theMonte Carlo simulation method.The reasonableness of the models and the effectiveness of the simulation methodspresented are demonstrated by an application example of a power system consisting oflithium-ion batteries.4. Optimization design of series-parallel system containing load-sharing redundantsubsystems(1) According to system reliability requirements of multiple times, optimization modelsof reliability redundancy allocation are built separately for series-parallel system containingsubsystems consisting of same type of unit and system containing subsystems consisting ofdifferent types of units.(2) Considering the characteristics of reliability model of load-sharing redundantsubsystems, one order neighborhood heuristic algorithm and Genetic algorithm (GA) arestudied to solve the optimization models presented.The examples demonstrate the comprehensiveness of the optimization models and thefeasibility of the optimization algorithms.