SBIR Phase I: A Fleet-based Fault Detection System Based on Clustering Approach

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project will be to improve efficiency and reduce operating costs across a number of target industries. An innovative cluster-based approach is proposed for analyzing the performance of groups of similar machines and predicting maintenance requirements. Adopting this approach will broaden the scope of existing equipment monitoring methods, increasing the potential for capturing critical issues before they occur, thereby improving safety and uptime while also reducing maintenance costs. Increased competitiveness - especially in manufacturing - through increased uptime and lower maintenance costs is critical to the viability of the nation's industries. The broad applicability of this approach - which can be implemented in a wide range of industries that utilize groups of similar machines or devices - increases its potential for making real technological and economic impact. This Small Business Innovation Research (SBIR) Phase I project aims to advance data-driven fault detection technology by leveraging data streams obtained from a collection of similar machines to develop analytic models for predicting performance and maintenance requirements. An advantage of the proposed technology is that it is applicable to newly commissioned or recently repaired equipment, for which historical data may be limited or non-existent. A cluster-based approach is utilized to aggregate data from similar machines operating under similar conditions, allowing the development of accurate, localized health assessment models that will enable machine operators to benchmark machine performance and optimally prioritize maintenance activities. This project will primarily cover fault detection and maintenance, but the proposed approach can evolve to other condition based monitoring tasks such as performance prediction and fault diagnosis. A technology prototype will be demonstrated with fleets of discrete equipment, such as industrial robots used in automotive manufacturing. The approach can also be applied to equipment containing a multiplicity of similar components.