Data Driven Prognostics

This proposal is to describe a unique and promising solution for data-driven prognostics called the hidden-Markov-model (HMM) based prognostic (HBP), and to study the feasibility of using the above solution to build products for military and commercialmarkets. Data-driven prognostics have been studied for many years; however, the performances of existing systems do not meet the requirements of military and commercial applications because of inherent limitations in the approaches. We note that theprognostic is a dynamic-pattern recognition problem, but most existing approaches only use techniques and models developed years ago for steady-pattern recognition, which is not adequate in terms of accuracy and lacks true and full replication of operatingmachines. From our current and previous research, we know that speech signals have characteristics similar to the signals collected from operating machines or equipment. The HMM techniques developed in automatic speech recognition (ASR) are indeed fordynamic-pattern recognition and have provided a solution for ASR successfully; therefore, the proposed HMM approach has the potential to solve the problem and to develop products for data-driven prognostics for commercial and defense applications. We willdevelop prognostic products including software and hardware through this research and development. The products will be introduced to DoD first, and then provided to commercial aircraft and engine companies in the private sector market, such as Boeing, GE,and other system manufacturers. The size of the potential market is quite large since virtually every aircraft engine or weapon system needs a prognostic system to predict the failure, fault, or errors, and to ensure the safety of complex aircraft andweapon systems. We expect to sell significant number systems during the first year after finishing Phase II, and also expect a significant increase per year following the first year since we will have new products, such as chips for prognostics, at thattime.