Collaborative Research: Improving the Safety of Complex Engineered Systems by Identifying Failure Paths Early in the Design Process

复杂的工程系统，如在能源生产、航空和机器人工业中发现的系统，依赖于多种技术、科学和工程学科的整合。确保这些系统安全运行的传统做法是使用高度可靠的组件，并严格测试潜在的设计方案。在这个过程中，安全特性通常只是在制造之前进行验证，而不是在早期阶段指导设计中发挥有意义的作用。这种时间和成本密集的测试和失败后重新设计的方法限制了美国先进制造能力的竞争力。该奖项支持开发安全驱动的早期设计新范例的基础研究，并验证相应的系统设计理论。这项研究的直接结果是一组指标，可以让设计师在设计周期的早期识别和比较不同系统替代方案的安全性。传统上，系统工程师只使用启发式可靠性度量来解决安全性问题，或者在设计阶段结束时使用蒙特卡罗分析或概率风险分析。最近在设计决策制定和形式化系统建模方面的研究支持将系统分析和测试向前推进到设计阶段。基于这些最近的趋势，该奖项侧重于在早期设计中考虑安全性。具体而言，本研究旨在通过将统计聚类方法应用于数千种复杂故障场景，对系统故障进行分类和理解，从而了解导致安全运行的系统行为。通过使用危险本体和自动生成潜在故障路径的搜索算法，可以对潜在故障行为进行探索。度量标准将被识别，以帮助系统工程师识别危险系统行为的风险。此外，这些指标将使设计人员能够根据安全性能在早期设计阶段评估和比较设计架构和技术替代方案。从这项研究中得出的方法和基于安全的评估将通过一个新的复杂系统设计试验台进行验证，该试验台由一个机器人漫游者合作团队组成。

Complex engineered systems, such as those found in energy production, aviation and robotic industries rely on the integration of multiple technical, scientific and engineering disciplines. The traditional practice to ensure the safe operation of these systems is to use highly reliable components and to rigorously test potential design alternatives. Under this process, safety characteristics are often validated just prior to manufacturing instead of having a meaningful role in directing design from the earliest stages. This time and cost intensive approach of testing and redesigning after failures limits the competitiveness of the United States' advanced manufacturing capabilities. This award supports fundamental research for developing a new paradigm for safety-driven early-phase design, and for validating the corresponding system design theories. The immediate outcome of this research is a set of metrics that can allow designers to identify and compare the safety of different system alternatives early in the design cycle. Systems engineers have traditionally addressed safety only using heuristic reliability metrics or, towards the end of the design phase, using Monte Carlo analysis or Probabilistic Risk Analysis. Recent research in design decision making and formal systems modeling supports moving system analysis and testing forward in the design phase. Building on these recent trends, this award focuses on considering safety in early design. Specifically, this research aims to develop an understanding of system behaviors that lead to safe operation, by applying statistical clustering methods to thousands of complex failure scenarios to classify and understand system faults. The exploration of potential failure behaviors will be enabled through the use of hazard ontologies and search algorithms that automatically generate potential failure paths. Metrics will be identified to help system engineers identify the risk of hazardous system behaviors. Further, these metrics will enable designers to evaluate and compare design architectures and technological alternatives in the early design stage based on safety properties. The methods and safety-based assessments that are derived from this research will be validated using a novel complex system design testbed, consisting of a cooperative team of robotic rovers.