Modeling and Prediction of Welded Joints Lifetimes by GMAW Process Using Maximum Entropy Regression Model

Abstract

Accelerated life testing is a technique that is widely used to get timely reliability information on materials, components, and systems. The regression classic models related to accelerated testing have been developed during the last years. Commonly, these models are used to make inference and reliability analysis about systems due to their characteristics. However, the Generalized Maximum Entropy (GME) model is a powerful method for modeling complex welding engineering processes. GME offers the advantage of fast calibration and it is possible to make accurate predictions about the fatigue of welded joints. The prediction rates for classic models are compared with MEM using different functions. This method achieves better overall performance than robust regression in measures such as R2. Fatigue testing data of welded joints by Gas Metal Arc Welding (GMAW) process are used, the MEM showed best results. In order to predict lifetimes of welded joints and these could be used to establish the warranty time.