Reliability Prediction Based on Dynamic Data Collected with Modern Technology

该奖项的研究目标是开发一个通用框架，该框架可以纳入大规模动态数据，以获得更准确的可靠性预测。智能芯片、传感器和无线网络等现代技术已经改变了数据收集过程。越来越多的产品安装了自动数据收集设备（adcd），它可以动态记录现场单个单元的系统性能、使用情况和环境信息，并/或在所有者许可的情况下将这些信息传输到数据中心。这些产品范围从喷气发动机、风力涡轮机、电力变压器和CAT扫描仪，到汽车、复印机和智能手机。这项研究将首先开发将动态数据纳入预测的一般模型。然后将开发量化统计不确定性的方法和使用动态数据的优势。将进行敏感性分析以评估模型的不确定性。计算效率高的算法和能够处理大规模数据集的免费软件也将被开发出来。开发的方法将通过工业和政府合作伙伴提供的数据集进行验证。如果成功，该研究将为即将到来的现场可靠性数据生成提供急需的新范式。在不久的将来，随着adcd成本的进一步降低，越来越多的产品将配备adcd。这项研究将应用于制造业、可再生能源和医疗保健等各个重要领域，因为可靠性信息对制造商提高其产品的竞争地位至关重要，对成本分析、资本支出和风险控制也很重要。自由软件的发展将使所开发的方法得到广泛传播成为可能。来自弱势群体和女性的研究生和本科生将参与这项研究。研究与教学的结合将向学生展示现代可靠性数据分析的概念和技术。

The research objective of this award is to develop a general framework that can incorporate large-scale dynamic data to obtain more accurate reliability predictions. Modern technology, such as smart-chips, sensors and wireless networks, has changed data collection processes. There are more and more products installed with automatic data-collecting devices (ADCDs) which can dynamically record system performance, usage and environmental information for individual units in the field and/or transmit this information to data centers with owner's permissions. These products range from jet engines, wind turbines, power transformers and CAT scanners, to automobiles, copier machines and smart-phones. This research will first develop general models for incorporating dynamic data into predictions. Then methods will be developed for quantifying statistical uncertainties and advantage of using dynamic data. Sensitivity analysis will be conducted to assess model uncertainties. Computationally efficient algorithms and free software that is capable of processing large-scale datasets will also be developed. The developed methods will be validated with datasets from industrial and government partners.If successful, this research will provide a much-needed new paradigm for the arriving generation of field reliability data. In the near future when the cost of ADCDs further decreases, more and more products will be equipped with ADCDs. This research will have applications in various important areas such as manufacturing, renewable energy, and health care, because reliability information is critical for manufacturers to improve the competitive position of their products, and is also important for cost analysis, capital expenditures, and risk controls. The development of free software will make it possible for the developed methods to be widely disseminated. Graduate and undergraduate students from under-represented groups and women will be involved in this research. The integration of research with teaching will present students with modern reliability data analysis concepts and techniques.