Proceedings of the 24th International Conference on Flexible Automation & Intelligent Manufacturing; FAIM 2014

Paper presented at the Proceedings of the 24th International Conference on Flexible Automation & Intelligent Manufacturing, held May 20-23, 2014 in San Antonio, Texas, and organized by the Center for Advanced Manufacturing and Lean Systems, University of Texas at San Antonio; Includes bibliographical references; To cope with today market challenges and guarantee adequate competitive performances, companies have been decreasing their products life cycles, as well as increasing the number of product varieties and respective services available on their portfolio. Consequently, it has been observed an increasing in complexity in all domains, from product and process development, factory and production planning to factory operation and management. This reality implies that organizations should be able to compile and analyze, in a more agile way, the immense quantity of data generated, as well as apply the suitable tools that, based on this knowledge, will supports stakeholders to take decision envisioning future performance scenarios. Aiming to pursuing this vision was developed a proactive performance management framework, composed by a performance thinking methodology and a performance estimation engine. While the methodology developed is an extension of the Systems Dynamics approach for complex systems’ performance management, on the other hand, the performance estimation engine is an innovative IT solution responsible by capturing lagging indicators, as well as estimate future performance behaviors. As main outcome of this research work, it was demonstrated that following a systematic and formal approach, it is possible to identify the feedback loops and respective endogenous and exogenous variables responsible by hindering the systems behavior, in terms of a specific KPI. Moreover, based on this enhanced understanding about manufacturing systems behavior, it was proved to be possible to estimate with high levels of confidence not only the present but also future performance behavior. From the combination of both qualitative and quantitative approaches, it was explored an enhanced learning machine algorithm capable to specify the curve of behavior, characteristic from a specific manufacturing system, and thus estimate future behaviors based on a set of leading indicators. In order to achieve these objectives, both Neural Networks and Unscented Kalman Filter for nonlinear estimation were applied. Important results and conclusions were extracted from an application case performed within a real automotive plant, which demonstrated the feasibility of this research towards a more proactive management approac