Acoustic Signal-Based Method for Recognizing Fluid Flow States in Distillation Columns

Real-time monitoring of the running states of processing equipment is essential for the stability and safety of unit operations in chemical engineering. In this study, an acoustic signal based method is proposed for recognizing fluid flow states in distillation columns. This method could accurately identify various fluid flow states in distillation columns, including dispersion, flooding, and weeping states. First, acoustic signals corresponding to various fluid flow states were recorded in various columns. Then, the characteristic parameters of these acoustic signals, including the short-time-averaged energy (STAE), linear predictive cepstral coefficients, and Mel frequency cepstral coefficients (MFCCs), were extracted and fused from the time, frequency, and cepstrum domains. Finally, a multi-classification support vector machine (MSVM) model was coupled with the characteristic parameters to recognize the fluid flow states in distillation columns. The MSVM based on the STAE + MFCC characteristics yielded the highest identification accuracy (97.25%).