Abnormal condition detection in a cement rotary kiln with system identification methods

In this paper, we use system identification methods for abnormal condition detection in a cement rotary kiln. After selecting proper inputs and output, an input–output model is identified for the plant’s normal conditions. A novel approach is used in order to estimate the delays of the input channels of the kiln before identification part. This method eases the identification since with determining the input channels delays, the dimension of search space in the identification part reduces. Afterward, to identify the kiln’s model, Locally Linear Neuro-Fuzzy (LLNF) model is used. This model is trained by LOcally LInear MOdel Tree (LOLIMOT) algorithm which is an incremental tree-structure algorithm. Finally, with the model for normal condition of the kiln, the incident of abnormalities in output are detected based on the length of duration and magnitude of difference between the real output and model’s output. We distinguished three abnormal conditions in the kiln, two of which are known as common abnormal conditions and another one which was not characteristically known for cement experts either.