Study On Maintenance Decision Making Theory And Its Application To Rotor System Of Centrifugal Compressor

Maintenance is the performed actions to keep and restore the system in a well condition. The objective of maintenance decision making (MDM) is to scheme the optimal maintenance plan for a system to make the system in a good condition usually, improve system reliability and availability, decrease the deterioration rate, prolong the service life, guarantee the safety as much as possible during system operating, protect the environment and so on under the condition of the minimization of the spent cost, time and other resource and the maximization of the restoration ratio. The research content of MDM consists of the maintenance policy decision making and making the optimal maintenance plan for a system under a certain maintenance policy. In this dissertation, aiming at some problems of the decision making of maintenance policy and condition-based maintenance which have not been solved and are combined with the rotor system of centrifugal compressor, the following contents are explored:(1) Study on maintenance policy decision making with mix data and unknown weight of every attribute. The qualitative pairwise compassion matrix among attributes which is presented by decision-makers is transformed into the fuzzy number matrix before the fuzzy preference programming method is employed to calculate the crisp weights of all attributes. To solve the ploblem of mix data, the ELECTRE III based stochastic dominance and the distance between the centroid point of a fuzzy number and coordinate origin (DCPFNCO) is proposed. The approach constructs the local harmonious and non-harmonious index (LHI and LNHI) based on DCPFNCO firstly in order to solve the problem of the comparision of any two maintenance polices under fuzzy evaluation, and they are combined with LHI and LNHI based on stochastic dominance and ELECTRE III to solve the ploblem of mix data. The proposed approach is applied to maintenance policy decision making of the rotor system of centrifugal compressor. The results of the numerical example show that the optimal maintenance policy for a system could be determined by the proposed approach.(2) Study on condition assessement of the rotor system of centrifugal compressor. The model-based approach is employed to assess the unbalance quantity, the degree of misalignment and the quantity of bending deformation of the rotor system; the preprocessing approach based the improved EMD is presented to reduce the effect of noise on the assessement accuracy; then, finite element analysis is used to verify the two approaches.The condition assessment approach of the stage seal of centrifugal compressor based on the axial thrust is proposed, where the BP neural network is applied to express the relationship between the axial thrust and the stage seal condition and the trained BP neural network is used to assess the stage seal condition; the axial thrusts are calculated by FLUENT under the different condition of the stage seal to verify the proposed approach. According to the results of the numerical examples, it could be seen that the unbalance quantity, the degree of misalignment and the quantity of bending deformation of the rotor system could be assessed accurately by the preprocessing approach based the improved EMD and the model-based approach; the stage seal condition could be estimated exactly by the condition assessment approach of the stage seal of centrifugal compressor based on the axial thrust.(3) Study on condition-based maintenance optimization for the system of which the condition is determined by multi-deterioration condition variables monitored continuously and perfectly. Suppose each deterioration condition variable follow a stationary Gamma process with two certain parameters which are estimated by Bayesian estimation method. The conjugate prior distribution of single parameter of the stationary Gamma process could be obtained and is proved in this section. Then, the expected time interval model of the preventive maintenance of the system whose condition is determined by only one deterioration condition variable is derived and that of the considered system is given subsequently. The effect of the imperfect preventive maintenance on the system condition is formulated according to the idea of the effective age model and the expected restoration time period model of the imperfect preventive maintenance in some literatures is used. The expected operation cost per unit time and the expected availability of the considered system are modeled. On these bases, the condition-based maintenance optimization model for the system whose condition is determined by multi-deterioration condition variables monitored continuously and perfectly is formulated and solved by the improved ideal point method which is proposed in this dissertation. The proposed approach is applied to two cases that the rotor system of centrifugal compressor of which the condition is determined only by unbalance quantity and by unbalance quantity, stage seal condition and the quantity of bending deformation of the shaft, respectively. The results of the numerical examples show that the proposed maintenance optimization model accords with practical engineering and could solve the problem of condition-based MDM for the system of which the condition is determined by multi-deterioration condition variables monitored continuously and perfectly.(4) Study on condition-based maintenance optimization for the system subject to sudden failure and multiple deterioration failure simultaneously. Each deterioration condition variable is assumed to follow a stationary Gamma process with two certain parameters; then, the reliability of the system whose condition is determined by only one deterioration condition variable is derived on the basis of the reliability definition, whose parameter is estimated real-timely by the Bayesian estimation method. Suppose the sudden failure time follow the Weibull distribution with two parameters; then, the reliability of the system subject to only the sudden failure is modeled whose parameters is estimated real-timely by Bayesian estimation method with the conjugate prior distribution of two parameters. Subsequently, the considered system reliability is formulated on the basis of the above two reliability models. It is used for the index of the condition of the system. The models of the interval between two preventive maintenances under the condition without sudden failure, the expected uptime and downtime of the system in a cycle and the expected operational cost per unit time are modeled respectively. On these bases, the condition-based maintenance optimization model for the system undergoing the sudden failure and multi-deterioration failure simultaneously is formulated. The proposed approach is applied to the condition-based MDM of the rotor system of centrifugal compressore which suffer from sudden failure due to occurrence of the crack of the shaft or impeller and deterioration failure due to increase of unbalance quantity. The results of the numerical examples show that the presented condition-based maintenance optimization model for the system undergoing the sudden failure and multi-deterioration failure simultaneously is effective.To sum up, some feasible and effective approaches to solving the not solved problem of decision making of maintenance policy and condition-based maitiance are proposed in this dissertation and they may provides some references for the deep research on MDM and the application of MDM to the rotor system of centrifugal compressor or other system.