Bearings vibration in gas turbines is considered as an injurious event, which results in incidents such as emergency shutdown or damages in turbine blades and imposes expensive costs to the system. Thus, measuring and analyzing of vibration rate in gas turbines is very important and knowing about its operational conditions and prediction of this phenomenon can help a lot in reducing vibration, avoiding damage to the blades and eventually financial savings. In this paper, we are modelling the vibration rate of a real double shaft 25 MW gas turbine, located in Iran, by making use of a hybrid intelligent model based on multi-layer perceptron neural network and cuckoo optimization algorithm; so, the model in this paper is abbreviated as MLP-COA. It should be noted that this work is an absolutely novel work and the idea is implemented in a real turbine for first time. We have used a real dataset with 161 samples which are collected during a year from a gas turbine in a gas pressure booster station. Furthermore, to obtain the effect of each input parameter on the vibration rate, we have applied sensitivity analysis using the cosine amplitude technique. Evaluation of predicted vibration rates was performed and prove satisfactory efficiency of this model than other predictive models such as radial basis function and multi-layer perceptron. The model can also be used for prediction of online vibration rate without any constraint in selection of data points in training phase.

No abstract

No abstract