The invention relates to the technical field of dry
quenching, in particular to a dry
quenching abnormal material level judgment method which comprises the following steps: data loading and preprocessing: setting an abnormal
detection threshold: calculating an abnormal upper
limit value and an abnormal
lower limit value based on statistical distribution of a
coke loading rate; key parameters are initialized, wherein a
backtracking step number threshold value and a material level abnormity judgment threshold value are set; and
anomaly detection and classification: judging whether the
coke loading rate exceeds an abnormal upper
limit value or is lower than an abnormal
lower limit value point by point from the latest time point to the front, and identifying an abnormal sequence, a material suspension phenomenon and a material collapse phenomenon according to a judgment result. The method has the advantages that the key parameters of the
coke loading rate and the material level change rate are integrated, and the dynamic abnormal threshold value is set based on statistical distribution, so that false alarms caused by
data noise of a single sensor and instantaneous interference are effectively reduced; and in combination with a reverse anomaly
backtracking mechanism, an abnormal mode (such as material suspension and material collapse) with
time sequence characteristics can be identified, the missing report is reduced, and the real anomaly capture capability is improved.