A gas–solid flow pattern identification algorithm based on cross-rod electrostatic sensor array

Abstract The accurate identification of gas–solid two-phase flow patterns is an important but challenging subject for pneumatic conveying. In this study, the sensitivity deficiencies of a single electrode were analysed via finite element analysis and a more sensitive cross-rod electrostatic sensor array structure was designed to measure the flow pattern signals. The experiment used Geldart D particles to verify the feasibility of the designed sensor array. Three types of feature vectors were extracted: the mean value, variance, and energy ratio. To identify the flow pattern accurately, the sine–cosine algorithm (SCA) is exploited to optimise the smoothing factor critical for a probabilistic neural network (PNN), namely SCA-PNN. The identification results show that the identification accuracy of the proposed algorithm outperforms the traditional PNN, the back propagation neural network (BPNN) and the support vector machine (SVM).