Study On The Non-linear Prediction Theory Of Old Goaf Residual Subsidence And Its Application

Building on the old goaf, an effective measure to reclaim subsidence land induced by mining, will be widely applied for treating the subsidence land ascribing for these advantages, such as increasing available construction ground, resuming the scenery of old goaf surface, improving the environment of the mining area, and so on. On the basis of collecting and analyzing the field data of surface subsidence and using non-linear theories, the law of spatio-temporal deformation, prediction methods and parameters system were studied systemically and the evaluation model of foundation stability of old goaf buildings was constructed, which would provide the indispensable technology support for construction utilization above the old goaf. The main results are as follows:(1) Integrated prediction model of surface dynamic subsidence was proposed in this dissertation, and combining with probability-integral method, another prediction model was presented to calculate the movement and deformation of arbitrary point in mining area in any direction and at any time. The relationship was analyzed between geological mining conditions and the duration time and value of residual subsidence. The results show that the duration and the value of residual subsidence increase with increasing the thickness of rock strata and decreasing the advance speed of the working face.(2) Dynamic parameters system of probability-integral method was established. Statistical analysis shows that residual subsidence coefficient and the main influential angle tangent decrease with increasing the mining degree, and the time parameter decreases with increasing the ratio of rock strata and mining thickness and the movement process of subsidence coefficient and the main influential angle tangent become slower.(3) A novel method named multi-scale kernel partial least-squares regression (multi-scale KPLS) was proposed to identify parameters of the probability-integral method in mining area lacking of filed data, at the same time, amended probability-integral method was put forward in this dissertation, which effectively solved the problems of quick convergence on the edge of subsidence and horizontal movement curves.(4) Multiple scales analysis on time variational character of old goaf residual subsidence was carried out. It has been found that residual subsidence of the old goaf presents obvious periodicity, and each cycle shows certain characteristics; in the same goaf, there are equal or approximately equal notable periodicity and duration time in different observation points.(5) Aiming at the character of random fluctuation of residual subsidence sequence of old goaf, three prediction models of it were presented in this dissertation, including wavelet support vector machines (WT-SVM), grey Markov prediction model (GM-Markov) and discrete grey model of 2 order 2 variables (DGM(2,2)), which was applicable to predict the length of observation sequence l conforming to the requirements of l≥20, 10≤l<20 and 6≤l<10 respectively. Meanwhile, the corresponding prediction step should limit within l/5, 4 and 2 respectively.(6) Height prediction of water fractured zone based on fuzzy SVM was brought forward. According to the prediction value, stability of the old goaf could be concluded by the criterion, namely, the old goaf is steady if the influential depth does not reach the water fractured zone. Moreover, stability evaluation of old goaf based on fuzzy extension model was proposed to estimate whether the old goaf was suitable for buildings or not. A case study was given in this dissertation and the results indicated that the two methods mentioned above and residual subsidence prediction based on amended probability-integral method could be adopted to evaluate the old goaf stability effectively.There are 66 figures, 47 tables and 202 references in this dissertation.