Monitoramento da condição de unidades motogeradoras em usinas termelétricas usando regras de Nelson

Abstract

The physical structure of Thermoelectric Power Plants is made up of several components, among which the internal combustion engines (MCI) stand out. Fundamentally, in this type of engine, as the name suggests, energy is produced by burning the fuel inside. As this generation of thermoelectric energy remains constant during the plant’s operating period, these engines are constantly subject to stress, due to different factors such as number of starts, average load, load variation and ambient temperature. However, most of the time, these existing monitoring systems only indicate whether the monitored parameters are non-compliant after a failure has occurred – not allowing an early analysis of the machine’s operating conditions. Knowing this, in this work, a methodology for early analysis of failures in the internal combustion engine, based on the Statistical Control Process and Nelson’s Rules, is proposed in an unprecedented way to analyze, in a predictive manner, the operating conditions of the machine based on in historical data from the UTE supervisory system. In order to give a practical appeal to the analyzes carried out in this work, the operational data used are the real records of the year 2019, of the internal combustion engine model 18V46 (from the manufacturer Wärtsilä), which constitute the thermoelectric plant in question. The analysis mechanism is validated through two case studies, the first being a comparison between the degree of severity of the operational condition of generating unit 5 and the shutdown event of this same generating unit on 08/01/19 and the second case study statistical notes on the degree of severity of generating unit 5, when it returns from post-shutdown maintenance. It is demonstrated that the results obtained with the proposed methodology adequately corresponded to the entries in the shift record, namely the degradation of certain subsystems of the motor generating unit 5 until its stop in August 2019 - which makes it a promising tool both for making early decisions about plant maintenance and for helping machine operators check whether machines that have returned from maintenance are in good condition.