Maintenance Modeling and Control of Multi-unit Systems Subject to Condition Monitoring

这项研究将集中于多单元系统的维修模型和最优维修策略的发展，特别是在实际系统中广泛出现的结构，如串联系统和由N个独立运行单元组成的系统。串联系统的例子包括复杂的设备和机械。N单元系统的例子包括能源系统，如不同地点的风能或太阳能电池板发电场，特定地理区域的石油钻井平台，由几台机器或生产线组成的生产系统，以及卡车、巴士、火车或飞机车队等车辆。在许多情况下，单元本身可以被认为是串联系统。以往的研究主要集中在单机系统的维护模型和最优策略的开发上。已发表的关于两单元和多单元系统的研究成果非常有限。由于拟议研究中考虑的系统的维护设置成本很高，可能还包括差旅成本和系统停机时间成本，因此各单元之间存在经济上的相互依赖，新制定的利用机会主义维护的维护政策应该会带来实质性的好处。这项研究的其他挑战包括与这些系统相关的其他约束，如能源或生产需求的满足，计划的定期维护的调度，以及为满足运输需求而为机组组制定的CBM政策，适当的车队规模和配备不同推进系统的车辆数量的选择，以及随后的更换决策。这项拟议的研究为博士和硕士提供了几个极具挑战性的主题。预计将有学生和顶级期刊发表文章。将应用的理论和方法包括基于隐马尔可夫和半马尔可夫模型的煤层气建模和控制、贝叶斯动态规划、半马尔可夫决策过程理论、最优停止、更新和马尔可夫更新理论、再生和半再生过程理论、比例风险建模、EM算法及其推广、相型建模、组和机会维护建模、最近发展起来的近似随机动态规划方法和指标策略、向量时间序列建模和数据降维方法、振动数据去噪方法和仿真。成果将发表在一线期刊上，并在主要国际会议上展示，它们将为该学科的进一步发展做出重大贡献，将带来很高的经济效益，在许多行业将具有高度的应用性和可取性。

The proposed research will focus on the development of maintenance models and optimal maintenance policies for multi-unit systems with the special interest in widely appearing structures of real systems such as series systems and the systems consisting of N independently operating units. Examples of series systems include complex equipment and machinery. Examples of N-unit systems include energy systems such as wind or solar panel farms at different locations, oil drilling platforms in a particular geographical area, production systems consisting of several machines or production lines, and fleets of vehicles such as trucks, buses, trains or airplane fleets. In many situations the units themselves can be considered as series systems. Previous research has focused mainly on the development of maintenance models and optimal policies for single unit systems. The published research results for two-unit and multi-unit systems are very limited. Due to the high maintenance setup cost for systems considered in the proposed research which may include also the travel cost and the system downtime cost, there is an economic inter-dependence among the units, and the newly developed maintenance policies which will utilize opportunistic maintenance should bring substantial benefits. Other challenges in this research include the additional constraints associated with these systems such as energy or production demand satisfaction, scheduling of planned, periodic maintenance combined with the developed CBM policies for groups of the units to meet the transportation requirements, selection of the proper size of the fleets and the number of vehicles with different propulsion systems, and the subsequent replacement decisions. The proposed research provides several excellent challenging topics for Ph.D. and M.A.Sc. students and top journal publications are expected. The theories and methodologies which will be applied include CBM modeling and control using hidden Markov and semi-Markov modeling, Bayesian dynamic programming, semi-Markov decision process theory, optimal stopping, renewal and Markov renewal theory, theory of regenerative and semi-regenerative processes, proportional hazards modeling, EM algorithm and its extensions, phase-type modeling, group and opportunistic maintenance modeling, recently developed approximate stochastic dynamic programming methods and index policies, vector time series modeling and data dimensionality reduction methods, vibration data de-noising methods, and simulation. The results will be published in the front line journals and presented at major international conferences, they will contribute significantly to the further development of the discipline, will bring high economic benefits, and will be highly applicable and desirable in many industries.