基于可靠性的民用飞机计划维修的决策方法

发布时间：2018-01-16 01:32

  本文关键词：基于可靠性的民用飞机计划维修的决策方法　出处：《南京航空航天大学》2016年博士论文　论文类型：学位论文

  更多相关文章： 可靠性评估 可靠性折算 环境因子 截尾数据 小样本 分段分布 计划维修 维修决策 聚类 单纯形

【摘要】：维修是保证民用飞机持续适航的重要手段,故在飞机设计、制造和运营的整个过程中,维修都是必须重点关注。例如,在民用飞机设计制造时就开始编制维修大纲、维修手册等,在飞机运用中又必须实施维修活动、记录维修信息。合理的维修降低维修成本不仅可以为飞机制造商赢得市场而且还能为航空公司节约成本,因此维修是飞机的重要保障性工作。本文主要从可靠性评估、维修策略和维修任务组合优化三个角度讨论了民用飞机的维修。首先,准确的可靠性评估是民用飞机维修决策的基础,没有准确的可靠性信息很难对维修做出正确的评判。维修策略是根据维修项目的可靠性、特点和需要确定每个维修项目合理的维修方式。最后,在实际执行维修活动时,往往还要对维修任务进行优化组合形成维修工作包,这不仅是组织管理维修活动的需要,而且还能降低维修成本。民用飞机使用过程中经常产生各种截尾小样本,这增加了可靠性评估的困难。鉴于截尾样本的真实寿命与完全样本有相同的分布,若把截尾样本转化成完全样本,则可按照完全样本的频数的比例估算截尾样本折算的频数。一方面在该频数的基础上改进了标准寿命表估计和乘积限估计,使得它们能准确地估计各类截尾样本。另一方面,把该频数作为权重因素之一,并结合顺序统计量在标准分布下的方差,形成能精确估计各类截尾样本的加权最小二乘法。由于截尾数据的似然函数中含有概率分布函数而导致MLE性能下降,本文在构造截尾事件及其概率密度函数的基础上,形成了截尾数据的概率密度函数。用截尾数据的概率密度函数替代传统MLE使用的概率分布函数,再与完全样本的概率密度函数共同构造似然函数。这样似然函数就全部由概率密度函数组成,故改进的MLE符合MLE定义的。大量的仿真实验证明本文改进的方法可以很好评估截尾小样本的可靠度,比传统MLE估计得更加准确。尤其是当样本中截尾数据的比例比较高的情况下,本文改进方法的优势更加明显。从评估升降舵作动器和副翼作动器的可靠性的实例中可以看出,和传统方法相比,本文改进方法评估的可靠性倾向于保守而且更加精确。通常在衡量民机可靠性时,都假设在整个寿命周期内风险有相同的规律并且故障服从同一个分布,故在整个寿命周期内使用统一的可靠度模型。事实上,民用飞机中许多设备的故障风险都很复杂,往往随运行时间有不同的特点甚至呈现多峰性,例如浴盆曲线,因此在整个寿命周期内用单一模型拟合设备可靠度的方法往往有较大的误差。如果根据风险变化的特点把整个寿命周期分成不同的区间,则分段拟合就可以更准确地逼近可靠度的变化。故本文把风险当作分段函数,提出了两类分段可靠度分布:一类风险函数是连续的;另一类风险函数可以是间断的。为了估计分段可靠度模型的参数,在对寿命样本进行聚类的基础上,本文讨论了分段可靠度分布的回归估计和极大似然估计。最后,因为通常的信息准则不适合分段分布模型,本文改进了信息准则以评价分段可靠度模型的优劣。经过大量的实验和仿真验证,本文的分段可靠度模型和它的估计方法比传统的单一模型更精确,尤其是在风险具有阶段性和多峰性时。分段风险模型在评估发动机引气系统和前缘襟翼接近传感器的可靠度中表现良好,说明了分段可靠度模型有很好的实用性和准确性。在准确评估可靠性后,合理地选择维修策略制定优化的维修决策是飞机维修保障的关键。目前,尽管功能检查是飞机维修中最常用的维修策略,功能检查的实际应用遇到两大问题:在某些条件下检查的准确度不够高;缺少潜在缺陷和延迟时间的实际样本。目前的功能检查模型多数没有考虑维修项目实际退化过程对维修策略的限制,许多理论上很好的检查策略在实际中却不很准确,不能准确发现故障。根据退化过程对检查间隔的约束,针对时间延迟模型的功能检查,建立有约束的不等间隔检查的费用率优化模型,并且提出了衡量检查准确性的指标。本文假设故障率在潜在缺陷发生前后是明显不同的,然后提出一个分段分布和基于KS检验的时间序列层次聚类方法来估计潜在缺陷的发生时间。然后,能从寿命样本中提取出延迟时间样本,进一步估计出延迟时间分布。最后,因为寿命是延迟时间和潜在缺陷时间的总和,这样在寿命分布被估计出后潜在缺陷时间可以轻松估计出。最后把本文的维修决策模型应用到民用飞机的减速控制系统中取得了良好的效果。实例和仿真结果都显示:不等间隔的检查策略要比等间隔的检查策略更加经济;和无约束的检查模型相比,有约束的检查模型虽然费用率数值可能不是最优,但是其确定的检查策略更加高效准确并符合工程实际。另外,在民用飞机的维修领域,单目标模型有时是单调的,不存在最优值,不能求出最优维修策略。故本文综合考虑多个目标,为维修策略建立多目标规划模型,就可以求出同时接近多个目标的最优策略。并且提出一个新的中心对称的单纯形优化方法来求解该多目标优化问题。实验证明本文的维修模型能取得更高效更准确的决策。维修任务优化组合不仅便于组织管理维修活动而且还能降低维修成本。但是在各种约束下实现维修任务的最优组合,不仅涉及数值计算而且还是个NP难问题。本文讨论了用遗传算法和聚类模型两个方法实现维修任务的组合优化。首先,本文提出了一种适合求解组合优化问题的遗传算法--基于簇遍历的遗传算法,以定时更换策略的块替换为例子,实现了维修任务的优化组合。当只已知维修任务的间隔及其浮动范围时,本文改进了模糊C均值聚类的模型以实现了维修任务的优化组合。改进FCM模型中消除了代表聚类中心的变量,降低了维数。在试验中,新的算法和模型都取得了良好效果。
[Abstract]:Maintenance is an important means to ensure the airworthiness of civil aircraft, the aircraft design, the whole process of manufacturing and operation, maintenance must be the focus of attention. For example, began compiling maintenance program in civil aircraft design and manufacture, maintenance manuals, application must be implemented in the aircraft maintenance, maintenance records reasonable information. The maintenance can not only reduce the cost of maintenance for the aircraft manufacturer to win the market but also for the airline cost savings, so maintenance is an important guarantee of aircraft work. This article mainly from the reliability assessment, maintenance strategy and maintenance task combination optimization three discussed in terms of civil aircraft maintenance. First, reliability is based on accurate assessment of civil aircraft maintenance the decision, no reliability information accurately is very difficult to make a correct judgement. Maintenance maintenance strategy is based on the reliability of maintenance projects, characteristics and Each project needs to determine the reasonable maintenance mode. Finally, in the actual execution of maintenance activities, often also for maintenance tasks to optimize the combination of the formation of the maintenance work package, which is not only the need of organization and management of maintenance activities, but also reduce the cost of maintenance. The censored small samples often produce civil aircraft in use process, which increases the reliability assessment of the difficulties. In view of real life censored samples have the same distribution and complete sample, if the truncated sample into a complete sample, may be in accordance with the proportion of the estimated frequency of complete sample frequency conversion censored samples. On the one hand based on the frequency of improvement on the standard life table method and product limit estimation, making them can accurately estimate various types of censored samples. On the other hand, the frequency as the weight factor, and variance of order statistics in the standard distribution, The formation of the weighted least squares method can accurately estimate various types of censored samples. MLE performance degradation due to the likelihood function with probability distribution function of censored data, based on the tectonic events and the truncated probability density function, the formation of the probability density function of censored data. Instead of the probability distribution function of traditional MLE with truncated probability density function the data, with the complete sample of the probability density function of the joint likelihood function. This function is composed of all likelihood probability density function, so the improved MLE conforms to the MLE definition. A large number of simulation experiments show that this improved method can well evaluate truncated small sample reliability estimation is more accurate than the traditional MLE. Especially when the censored data in a sample of a relatively high proportion of cases, the advantages of the improved method is more obvious. From the evaluation of elevator actuator For instance and aileron actuator reliability can be seen, compared with the traditional method, the improved method to assess the reliability of tendency of conservative and more accurate. Usually in terms of civil aircraft reliability, are assumed in the whole life cycle of risk have the same rules and fault obey the same distribution, so the use of reliability model unity in the whole life cycle. In fact, many civil aircraft equipment failure risk are complex, often with the running time has different characteristics and has multiple peaks, such as bathtub curve, therefore tend to have larger errors in the single model fitting equipment reliability method in the whole life cycle. If the the whole life cycle is divided into different sections according to the characteristics of risk changes, then fitting can more accurately approximate the reliability change. So this article takes the risk as a piecewise function, Two kinds of reliability distribution segment is proposed: a class of risk function is continuous; another kind of risk function can be discontinuous. In order to estimate the parameters of sectional reliability model, based on clustering of life samples, this paper discusses the reliability distribution of piecewise regression estimation and maximum likelihood estimation. Finally, because usually the information criterion is not suitable for piecewise distribution model, this paper improves the information criterion to evaluate the reliability of the segment of the models. After a great number of experiments and simulations verify the reliability of the segmentation model and its estimation method is more accurate than the traditional single model, especially with stage and multiple peaks in the risk. Sectional risk model in evaluating engine bleed air system and the front flap proximity sensor reliability in good performance, the sectional reliability model has practicability and good accuracy in accurate assessment. Estimation of reliability after reasonable selection of maintenance strategies, optimizing the maintenance decision-making is a key factor for aircraft maintenance safeguard. At present, although the function test is the most commonly used maintenance strategy for aircraft maintenance, application function tests encountered two major problems: under certain conditions the inspection accuracy is not high enough; the lack of potential defects and the actual sample the time delay function. Most current models do not consider maintenance inspection project degradation process of maintenance strategy, many good theoretical checking strategy in practice is not very accurate, can not accurately find fault. According to the constraint of the degradation process inspection interval, the time delay function test model, the establishment of a constraint the inspection fee rate of unequal interval optimization model, and put forward the measure of the accuracy of examination indicators. This assumption in the failure rate before and after the occurrence of potential defects is obvious The same, then proposed a piecewise distribution and time series clustering method based on KS test to estimate the time of occurrence of potential defects. Then, from the life of samples extracted samples to estimate the delay time, further delay time distribution. Finally, because life is the sum of the delay time and the potential defects in time, so can the potential defects of life distribution is estimated from the time after the estimated. Finally the application of easy maintenance decision model of this paper to speed control system for civil aircraft has achieved good results. Examples and simulation results showed that the unequal interval checking strategy to be more economic than the check interval strategy; compared with the check model without constraints the constraint checking model, while the rate of cost value may not be optimal, but the determination of inspection strategy is more efficient and accurate and practical. In addition, in the civil Aircraft maintenance field, single objective model is sometimes monotonous, there is no optimal value, can not find the optimal maintenance strategy. Therefore, considering the multiple targets, a multi-objective programming model for maintenance strategy, we can calculate the optimal strategy at the same time close to multiple targets. And put forward the method to solve the multi-objective optimization the problem of simplex optimization a new center of symmetry. Experiments show that this model can make maintenance more efficient and more accurate decisions. The maintenance task optimization not only for the organization and management of maintenance activities but also reduce the maintenance cost. But to achieve the optimum combination of maintenance tasks in a variety of constraints, not only relates to numerical calculation and a NP hard problem. This paper discusses combinatorial optimization using genetic algorithm and clustering model two methods to realize maintenance tasks. Firstly, this paper proposes a suitable for solving combinatorial optimization question The genetic algorithm -- genetic algorithm based on cluster traversal, to replace the timing block replacement policy for example, realize the optimization of combined maintenance tasks. When the only known maintenance task interval and the floating range, the improved fuzzy C means clustering model to achieve the optimal combination of maintenance tasks. The improved FCM model eliminate the clustering center representative variables, reduces the dimension. In the experiment, the new algorithm and the model have achieved good results.

【学位授予单位】：南京航空航天大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：V267
，

本文编号：1430987