Cartas adaptativas de controle : desenvolvimento de metodologia para implementação em processos de manufatura.

Abstract

Statistical control charts have been widely and successfully used for monitoring the performance of industrial processes. Several modifications on the classic Shewhart charts have been proposed in the literature; they aim at providing better tools for monitoring processes with special characteristics. One such modification is the recently developed adaptive control charts. Control charts are said to be adaptive when one or more of its parameters (sample size, sampling frequency, and control limits coefficient) are allowed to vary during process monitoring. Determination of the best set of values for the chart parameters may be based on statistical issues, economic issues or on a combination of both. Economic-statistical models consider, in addition to statistical performance indices, costs associated with the control scheme represented by the chart. Designing of charts based on such models may be a complex task, which requires a structured design methodology. This thesis presents a methodology proposal for the design of adaptive control charts to monitor processes that are currently under the supervision of traditional control charts. The proposed methodology, comprised of five main steps, was devised based on knowledge gathered in a case study from the automotive parts industry. A detailed account of the meaning and determination of input parameters to an economic-statistical model used in the design methodology is also given. A comparison of operational costs under a traditional and adaptive control scheme is presented using case study data. Results obtained favor the use of adaptive control charts for process monitoring, particularly when small deviations from the mean are relevant in the quality characteristic under analysis. Although developed from the results of a case study, the methodology steps described may be easily adapted to suit a wide range of applications.