The Decision-making On Scheduled Maintenance Of Civil Aircraft Based On The Reliability

Maintenance is an important measurement to ensure aircrafts continuous airworthiness.So,in the course of aircrafts design,manufacture and operation,maintenance is a remarkable issue that needs special attention.For example,we should prepare to edit and compile maintenance review board report and maintenance manual during design and manufacturing of aircrafts;in operation of aircrafts,we must implement maintenance activities and record maintenance information.Rational maintenance can reduce cost,which not only be helpful for airplane maker to win market but also reduce cost of airlines.Thus,maintenance is an important logistics and service-oriented work.In the paper,we discuss aircrafts maintenance from reliability evaluation,maintenance policy and optimal combination of maintenance tasks perspective.Firstly,precise reliability evaluation is a foundation of aircraft maintenance.We hardly make correct decision under no exact reliability information circumstances.The maintenance policy means selection and determination optimal repair period of maintenance item according to its reliability,character and requirements.At last,in practice of maintenance,it’s necessary to optimally combine maintenance task and form maintenance work package,which not only is requirements of management of maintenance activities,but also reduce maintenace cost.All kinds of censoring small samples often occur in maintenance of civil aircrafts,which makes reliability evaluation much more difficult.Considering that true lifetime of censoring sample has identical distribution with complete sample,if censoring sample is transformed into complete sample,conversion frequency of censoring sample can be estimated according to frequency of complete sample.On the one hand,standard life table estimation and product limit method are improved on the basis of such con-version frequency,enabling accurate estimation of various censoring samples.On the other hand,by taking such fre-quency as one of the weight factors and integrating variance of order statistics under standard distribution,weighted least square estimation is formed for accurately estimating various censoring samples.Probability distribution function or reliability function contained in likelihood function of censored data results in performance reduction of MLE.In this paper,on the basis of constructing censoring event and its probability density function,probability density function of censored data is formed.Probability distribution function in traditional MLE is substituted with probability density function of censored data;then likelihood function is constructed by combining PDF of censored data and that of complete sample.Thus,likelihood function is completely comprised of probability density functions,forming improved MLE which fully satisfies MLE definition.Large amount of simulation and experiments have proved that,despite that improvedMLE increases calculation,it often presents more accuracy in estimation than traditional MLE;moreover,when proportion of censored data in samples is high or sample size is small,improved MLE shows more prominent superiorities.According to instance in evaluating aileron actuator and elevator actuator reliability,compared with traditional methods,improved methods tends to be more conservative and accurate in reliability estimation.Usually,when we evaluate reliability of aircrafts,we assume failure follows one and same distribution.So,an identic identical selfsame reliability mode should be adopted.In fact,the hazard of most equipment in civil aircrafts shows different variation and tendencies,and even becomes a multi-peak trace with their operation time,for example,bathtub curve.As a result,traditionally fitting the reliability of equipment with the one and same model may lead to relatively larger error.However,if we divide the entire lifetime cycle of the equipment into several different intervals according to the trend and characteristics of its hazard,the segmented fitting can more accurately approximate its reliability.Therefore,in this paper,the hazard is regarded as a piecewise function,and two kinds of segmented reliability distribution are put forward,one of which is continuous,the other of which is discontinuous.In order to estimate the parameters of the segmented reliability model based on clustering of equipment lifetime samples,regression estimation and maximum likelihood estimation of the segmented reliability distribution are discussed in this paper.Finally,since the usual information criterion is not suitable for the segmented distribution model,we propose a improved information criterion to test and evaluate the segmented reliability model in the paper.After massive testing and verification,the segmented reliability model and its estimation methods in the paper proved to be more efficient and accurate than that of the traditional single model,especially when the change of hazard is time-phased and multi-peak.The segmented hazard model behaves well in reliability evaluation of engine bleed air system and proximity sensors of leading-edge flap in civil aircraft,which means the model and method brought forward in the paper is highly applicable and accurate.After accurate reliability evaluation,legitimate maintenace policy and optimized maintenance decision is key of aircrafts maintenance.Although functional inspection is an important maintenance policy,its practical application has two problems: firstly,it may be inaccurate under certain conditions;secondly,samples of potential defect and delay time are deficient in practice.Most of current functional inspection models disregard restrictions on the practical degeneration process.As a result,many theoretically good inspection policies are inaccurate in practice as they can’t discover defects accurately.According to the practical degeneration process,a constraint model of unequal intervalinspection based on cost ratio is created and the indicator measured the accuracy of inspection policy is discussed.Assuming that failure rate is obviously different before and after occurrence of potential defect,then a segmented distribution and a new hierarchical clustering of time series based on KS test are proposed to estimate potential defect time.After that,delay time samples can be extracted from lifetime sample and then delay time distribution can be naturally inferred.Finally,because lifetime is summation of potential defect time and delay time,delay time distribution can be easily estimated.The experiment result shows that unequal interval inspection policy is more economic and efficient than the period inspection while the constraint model is more accurate and accordant with engineering practice than the unconstraint model.In additon,single object model in aircrafts maintenance may be monotonous and no exist optimum points,which cann’t solve optimal decision-making.Thus we comprehensively consider multiple object and create multi-objective programming model,a optimal policy which can approximate multiple objects can be obtained.Experiments imply maintenance model can make high-efficiency and precise decision.Optimal combination of maintenance tasks not only is convenient for management maintenance activities but also can reduce maintenance cost.The optimal combination of maintenance task under all kinds of restraints not only is a numerical optimization,but also is a NP-hard combinatorial problem.Firstly,we put forward a novel genetic algorithm--adaptive genetic algorithm based on cluster search.Then take block replacement strategy for example,we optimally combine maintenance tasks.Only when maintenance interval of tasks and its float ranges are known,maintenance tasks are optimally combined by a improved fuzzy c-mean clustering in the paper.In my improved fuzzy c-mean clustering,the variables of class centers are removed,so the dimension of clustering model reduces.My novel algorithm and model behave excellent.