This research proposes a novel framework for autonomous root cause fault analysis, in a complex process with continuous learning. The potential root cause candidates are selected according a data mining process with 2 algorithms fully automated: Random Committee (RC) and Logistic Model Trees (LMT); they are competing for the best result. To determine the performance and application, it has been developed in a real case study, with the root cause analysis based on 65,000 inverters, 10,273,928 millions of data structured from February 2019 to February 2020, and their failures analysis; the results provide high accuracy, with a performance of 99.21% for the root cause analysis; it has been validated in a real solar photo-voltaic plant.