Due to extraction of coal seam, in longwall coal mining, the immediate roof strata of mined panel are disturbed and sag downwards. Then, the resulting movement of the roof strata extends upwards and will cause the disturbed roof strata be fractured. Subsequently, the roof pressures will be redistributed and transferred to the neighboring solid sections where the face abutment, adjacent access tunnels and the barrier pillars are located. Hence, an accurate determination of the height of this fracture zone which is called the height of destressed zone (HDZ) is essential to maintain the desired mine extraction efficiency and to generate a safe mine working environment. The paper describes two predictive models based on artificial neural network (ANN) and statistical analysis for predicting the height of destressed zone. A suitable dataset including the panel and the roof strata geometrical characteristics along with their mechanical properties were collected from literatures which have been divided into two groups, that is, training and testing datasets. To evaluate the performance of the employed models, the coefficient of determination (R2), variance account for (VAF), mean absolute error (Ea) and mean relative error (Er) indices were calculated based on the testing data. Comparison of the above mentioned results showed that the performance of ANN model is superior to the statistical model. Finally, the results are compared with the existing methods and the in-situ measurements. The comparative results confirm a reasonable agreement that exists among the methods as well as with the in-situ measurements.
