Hybrid ANN–PLS approach to scroll compressor thermodynamic performance prediction

• Hybrid ANN–PLS is utilized to predict the thermodynamic performance of scroll compressor.
• ANN–PLS model is determined with 5 hidden neurons and 7 latent variables.
• ANN–PLS model demonstrates better performance than ANN and PLS working separately.
• The values of MRE and RMSE are in the range of 0.34–1.96% and 0.9703–0.9999, respectively.

In this paper, a scroll compressor thermodynamic performance prediction was carried out by applying a hybrid ANN–PLS model. Firstly, an experimental platform with second-refrigeration calorimeter was set up and steady-state scroll compressor data sets were collected from experiments. Then totally 148 data sets were introduced to train and verify the validity of the ANN–PLS model for predicting the scroll compressor parameters such as volumetric efficiency, refrigerant mass flow rate, discharge temperature and power consumption. The ANN–PLS model was determined with 5 hidden neurons and 7 latent variables through the training process. Ultimately, the ANN–PLS model showed better performance than the ANN model and the PLS model working separately. ANN–PLS predictions agree well with the experimental values with mean relative errors (MREs) in the range of 0.34–1.96%, correlation coefficients (R2) in the range of 0.9703–0.9999 and very low root mean square errors (RMSEs).