Prediction of APS in-flight particle temperature and velocity using a novel extended Holt's damped trend model with small training data

The nonlinear relationship between the input process parameters and in-flight particle characteristics of the atmospheric plasma spray (APS) is of paramount importance for coating properties design and quality. Among these information, the in-flight particle characteristics is the most difficult to assess since they are not available during coating production. It is thus highly desirable to be able to predict the in-flight particle characteristics. Recently, machine learning has proven to be able to model complex nonlinear interactions. Predictive models of the in-flight particle characteristics proposed previously are mostly Artificial Neural Network based, which demand large volumes of training data. This work presents a novel hybrid model to predict the in-flight particle characteristics during an APS process considering torch electrodes ageing. The novel model only requires little training data. It extends the Holt’s damped trend model to consider both the characteristics of the present electrode pair as well as the overall trends observed from historical data and the inputs. Experiments were performed to record simultaneously the input process parameters, the in-flight powder particle characteristics and the electrodes usage time for multiple electrode pairs at conditions typical for TBC production. Model performance for both one step and multiple steps predictions will be evaluated.