STTR Phase II: Probabilistic and Explainable Deep Learning for the Intuitive Predictive Maintenance of Industrial and Agricultural Equipment

小企业技术转让（STTR）第二阶段更广泛的影响/商业潜力是改善基于时间表的维护计划，以确保工业和农业设备每周7天，每天24小时运行。与这种高生产率设备相关的停机时间可能导致重大的收入损失，研究表明，制造商平均每年要处理800小时的停机时间。该技术旨在有效减少或消除这种停机时间，为制造商创造价值。该项目提出了一种新的深度学习方法来预测旋转工业设备中的轴承故障，并使维护团队能够自信地围绕正在退化的设备计划最佳维护活动。该解决方案还旨在经济高效地使用一种专利方法，通过固定和移动的电池供电的无线传感器组合来监控工业物料搬运系统。该小型企业技术转让（STTR）第二阶段项目提出了一种新的深度学习方法，用于机械自动化。许多现有的深度学习方法专注于给定一组训练数据的最可能的失败场景。并行设备可能无法显示该训练集中涵盖的行为，从而导致低置信度预测。所提出的方法不仅可以预测机器部件的剩余使用寿命，而且还试图通过模型的集合和预测的时间融合来量化预测的不确定性。因此，可以从基于风险的角度做出维护决策，从而消除源自低置信度预测的不必要的维护。此外，许多现有的深度学习方法也缺乏向人类用户直观解释其预测的能力。在预测不佳会造成严重后果的关键应用中，维护人员必须理解并信任人工智能预测性维护合作伙伴。所提出的解决方案通过突出显示和动画显示对预测贡献最大的原始数据信号的片段来为模型的预测产生直观的视觉解释。这项技术可以让训练有素的人员通过融合数据驱动的见解与他们现有的领域专业知识，快速做出最佳的维护决策。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this Small Business Technology Transfer (STTR) Phase II is to improve schedule-based maintenance programs to ensure that industrial and farming equipment can function 24 hours a day, 7 days a week. The downtime associated with such high productivity equipment can result in significant lost revenue, and research shows that the average manufacturer deals with 800 hours of downtime per year. The proposed technology seeks to effectively reduce or eliminate this downtime, creating value for manufacturers. This project proposes a novel deep learning approach to predicting bearing failure in rotating industrial equipment and enable maintenance teams to confidently plan optimal maintenance activities around equipment that is in the process of degrading. This solution also aims to cost-effectively use a patented methodology to monitor industrial material handling systems with a combination of stationary and mobile battery-powered wireless sensors.This Small Business Technology Transfer (STTR) Phase II project proposes a novel deep learning approach to machinery prognostics. Many existing deep learning approaches focus on the most likely failure scenarios given a set of training data. Monitored equipment may not exbibit behavior covered in that training set, leading to low-confidence predictions. The proposed approach may not only predict the remaining useful life of a machine component, but also seeks to quantify the uncertainty of a prediction through an ensemble of models and a temporal fusion of predictions. As a result, maintenance decisions may be made from a risk-based perspective, eliminating unnecessary maintenance stemming from low-confidence predictions. Additionally, many existing deep learning approaches also lack the ability to intuitively explain their predictions to human users. In critical applications where poor predictions have serious consequences, maintenance personnel must understand and trust an artificially intelligent predictive maintenance partner. The proposed solution produces an intuitive visual explanation for the model’s prediction by highlighting and animating the segments of a raw data signal that are contributing most significantly to the prediction. This technology may allow trained personnel to quickly make optimal maintenance decisions by fusing data-driven insights with their existing domain expertise.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.