The aim of this work was to investigate the application of an acoustic monitoring system for production scale circulating fluidized bed processes. The relationship between acoustic signals and the coke amount deposited on the catalysts was investigated on both lab scale and production scale reactors. By using Power spectrum density analysis, the relationship between the coke amount and the frequency shift of the characteristic peak of acoustic signals was found to be linear. In addition, a model to detect the coke amount on catalyst was established based on partial least-squares regression. Comparing the model results with experimental data, the amount of deposited coke is determined with a correlation coefficient 0.947 and a root mean squares error of cross validation 0.948%, which means that this model is accurate and capable of monitoring the coke amount on catalyst. Furthermore, the results of the preliminary industrial tests are considered to be promising. This work provides valuable insights into the online monitoring of coke amounts in circulating fluidized beds based on an AE technique.