Performance prediction of vacuum membrane distillation system for sulfuric acid solution based on multi-layer perceptron-support vector machine

To solve the problems of low accuracy and poor stability in the performance prediction of the vacuum membrane distillation (VMD) system for sulfuric acid solution, this paper first built the experimental device, multi-condition operation tests were conducted. Then a multi-layer perceptron-support vector machine (MLP-SVM) model was constructed to predict the water production flux, a total of 232 and 58 data points were selected as the training and testing sets to train and test the hybrid model. The results showed that the predicted value was basically close to the real value. During model training process, Mean square error ( MSE ), Root mean square error ( RMSE ), Mean absolute error ( MAE ), Percent bias ( PBIAS ), Nash-Sutcliffe Efficiency ( NSE ), Willmott Index of Agreement ( WI ) and R 2 were 0.0017, 0.0407, 0.0362, -0.4713%, 0.9960, 0.999 and 0.9953. During model testing process, MSE , RMSE , MAE, PBIAS, NSE, WI and R 2 were 0.0061, 0.0775, 0.0556, -0.2668%, 0.984, 0.996 and 0.9842. Due to the small number of data sample and the existence of some outliers, there was still a certain error between real and predicted values, but the vast majority of errors were within an acceptable range of <20%. Therefore, the established model can efficiently predict the VMD system performance. Model structure of MLP-SVM model. • A test device of the VMD system was designed and built. • Multi-condition operation tests were carried out with sulfuric acid solution. • Model of multi-layer perceptron-support vector machine was constructed. • Values of R 2 in the training and testing processes were 0.9953 and 0.9842. • Relative error was mostly less than 20%.