Description: Accurately predicting Remaining Useful Life (RUL) provides significant benefits—it increases safety and reduces financial and labor resource requirements. Relying on just one methodology for RUL prediction is unsuitable because certain methods of prediction perform better for certain use cases and conditions. Approaches must be combined to maximize accuracy. Encoding these hybrid methods is challenging because their models are complex, change frequently, and represent a wide range of devices, components, and systems. The algorithms associated with these models also require deep mathematical understanding. We propose using probabilistic programming (PP) to integrate physical models and data-driven methods into a probabilistic model that can predict RUL under the Programming Useful Life Prediction (PULP) project. We will use Charles River's Figaro™ probabilistic programming language (PPL) to fuse physical models of critical fault modes and data-driven methods in a hybrid approach to accurately predict the RUL of critical flight systems. Figaro is an ideal solution because it eases construction of Probabilistic Relational Models (PRMs). PRMs can represent a wide range of complex, constantly changing domains that involve uncertainty and require flexibility. Figaro also contains a vast library of reasoning algorithms that can be applied to models, so users do not need deep mathematical expertise.

Benefits: The mature research developed under PULP will provide NASA with the ability to accurately predict Remaining Useful Life (RUL). PULP will provide analysts the ability to easily integrate model-based and method-based approaches in their predictions, taking full advantage of available data and leveraging the strengths of both approaches. Primary NASA commercial candidates include many projects led by the Prognostics Center of Excellence in support of potential NASA systems health management work. Other applications cited by NASA's Discovery and Systems Health (DaSH) may also become candidates as this technology matures.

We expect the full-scope PULP framework to have immediate and tangible benefits for a number of military prognostic and health-monitoring applications. PULP technologies will enable more accurate Remaining Useful Life (RUL) prediction during maintenance operations by using a hybrid approach to RUL prediction and increasing the efficiency of analysts performing RUL computations. We envision a number of commercial applications, particularly in the aerospace, aviation, and transport industries. We also plan to incorporate new PULP technology into our Figaro probabilistic modeling and analysis framework, which will enable us to use the tool to provide consulting services based on Figaro to customers within NASA, the DoD, other Government agencies, and commercial markets. Technology developed under the PULP effort will enhance Figaro's dynamic reasoning and integration capabilities.