Research on Destroyed Floor Depth 1 ---the example of the 51302 working face in the Liangzhuang coal

: Based on the analysis on factors affecting floor depth, mining depth, dip angle of coal seam, mining thickness, dip length of working face, damage variable of floor strata and whether 18 there are faults and crushed zones in the working face are defined as main influence factors. The 19 classical Down Three Zone Theory, Down Four Zone Theory based on damage mechanics, 20 formula written in the national standard, BP (Back Propagation) neural network algorithm 21 considering various influencing factors and Numerical Simulation Method were applied for 22 predicting floor depth of the 51302 working face. A double-side seal borehole water injection device was used to determine the actual floor depth of the 51302 working face. By comparing the 1 prediction results with the actual measurement results, the prediction results obtained by 2 considering rock damage are closer to reality. 3

device was used to determine the actual floor depth of the 51302 working face. By comparing the 1 prediction results with the actual measurement results, the prediction results obtained by 2 considering rock damage are closer to reality. In fact, after going through the long-term geologic process and 19 tectonization, there are many fracture planes in the interior of coal seam floor strata. 20 With coal seam mining, mine pressure and in situ stress have further influence on 21 floor strata, and floor strata is damaged more seriously, which is especially evident in 22 angle is 12°. The basal water influencing No. 13      The Down Three Zone Theory (Li, 1999) is put forward from the field practice. 11 Based on elastic mechanics and assuming that rock is continuous, perfectly elastic and

59.88ln
is the maximum concentration coefficient of mine pressure,  is 13 the average unit weight of overlying strata, and H is the mining depth.   Calculating destroyed floor depth according to BP neural network computing 13 Most of factors influencing destroyed floor depth are together with noise which 14 affects the prediction of destroyed floor depth directly. Based on BP neural network 15 computing, the establishment of destroyed floor depth prediction model can reduce 16 effectively reduce or eliminate the influence of these noises.

17
Acquisition of samples 18 Actual field data of 26 mined working faces are were chosen to be the training 19 samples and testing samples of BP neural network computing (Table 1). Samples

20
1~22 are set as learning samples to conduct training to the network, and samples 21 23~26 as checking samples to test the network performance. 22 The program is written in Matlab software and was run many times. Then, the 1 three-layer neural network structure is chosen as the model structure and its errors 2 meet the evaluation criteria (Fig. 5). The aims and demands can be obtained by 4000 3 training (Fig. 6).    2 According to the mechanical parameters (Table 3)   The process and results analysis of numerical simulation 1 The simulation process is conducted step by step (Fig. 8). The simulation mining  It is known from the simulated results that:  Setting of observation points 14 According to roadway conditions of working face (Fig. 10), the observation  Table 4.  Table 4 Drilling construction elements

11
(1) It is shown ( Fig. 11(a)) from the observation results before mining of No.1 12 borehole that there is obvious water leakage 3 m from the hole depth and the water 13 leakage rate is 1.2 L/min. This leaking occurs because the shallow surrounding rock 14 near the aperture is damaged and the broken rock zone is formed when the borehole is 15 constructed. When the hole depth is 21~33 m, the water leakage value increases 16 obviously up to 1.6 L/min. According to the analysis of water leakage value and floor 17 strata, it may be caused by partial original fracture development or roadway 1 construction. When the hole depth is 33 m, the water leakage value approaches to 0 2 L/min, which shows that the deep strata is basically in a continuously complete state 3 and the fractures do not develop. 4 (2) Because the No.1 borehole is complete, it is set as post-mining observation 5 pore to be tested. It is shown (Fig. 11(b)) from the data that after mining, the water

16
(1) By comparing the calculated and the measured results of the complete type and damaged type simulated by Down Three Zone Theory, Down Four Zone Theory, 1 Regulation, BP neural network and RFPA software, it is found that the calculated 2 value markedly differs from measured value a lot without considering the damage of