Data-driven condition-based maintenance models

制造、服务和流程导向的公司可能会将总运营成本的25%到50%用于关键设备的维护。这些例子包括矿业公司、钢铁生产商、航空公司、纸浆厂、石油生产商以及许多其他公司。关键设备在需要时无法使用可能会给公司造成严重的经济损失。基于状态的维护（CBM）是组织内采用的关键维护策略之一，其目标是在决定更换或维修设备之前，在经济上合理地延长设备的使用寿命。****一旦确定CBM适用于给定的维护情况，做出最佳维护决策取决于对设备状况的理解，使用适当的故障和决策模型，选择合适的优化方法，以及强大的实施工具。多伦多大学的维修优化和可靠性工程中心（C-MORE）在所有这些领域都取得了重要的研究成果。C-MORE现在寻求在几个方向上扩展和加强这些成就。****我们将在可靠性数据的统计解释方面推进最先进的技术，主要是为了处理在长寿命资产历史的早期部分缺少信息的问题。我们将研究退化模型，因为它们适合剩余寿命最优决策计算框架，以及针对剩余寿命估计的进一步进展。我们还将利用专家知识启发和贝叶斯统计扩展我们以前在CBM方面的有限数据的工作。

Manufacturing, service, and process-oriented companies can spend 25% to 50% of their total cost of operations on maintenance of critical equipment. Examples of these include mining companies, steel producers, airlines, pulp mills, oil producers, as well as many others. The unavailability of a critical piece of equipment when it is needed can result in severe financial loss to the company. Condition-based maintenance (CBM) is one of the key maintenance tactics employed within organizations with the goal of obtaining as much life out of the equipment as is economically justifiable before making the decision to replace or repair it.****Once CBM is identified as appropriate for a given maintenance situation, making optimal maintenance decisions depends on an understanding of the equipment's condition, the use of appropriate failure and decision models, the selection of a suitable optimization methodology, and a powerful implementation tool. The Centre for Maintenance Optimization and Reliability Engineering (C-MORE) at the University of Toronto has generated significant research outputs in all of these areas. C-MORE now seeks to extend and enhance these achievements in several directions.****We will advance the state of the art in the statistical interpretation of reliability data, mainly to deal with problems with missing information in the early parts of long-lived asset histories. We will investigate degradation models as they fit into frameworks for remaining life optimal decision calculations, along with further advances specifically for remaining life estimation. We will also expand our previous work on CBM with limited data using expert knowledge elicitation and Bayesian statistics.