Study On Dynamic Mining Subsidence Prediction Based On Robust Adaptive Collocation

The paper discusses the ill-conditioned problem of parameter inversion in probability-integral method and the application of adaptive collocation to mining subsidence prediction, based on the material of mining subsidence on a mining area and MATLAB software. The main works and contributions are summarized as follows:1. The basic concepts of probability-integral method are expounded, including static prediction and dynamic prediction. The least squares process of parameter estimation is derived. The parameter modification for the conditions of various mining extents is discussed.2. The geodetic ill-conditioned problem is analyzed, which exists in parameter inversion of probability-integral method. The ill-conditioned problem could be weakened by use of ridge estimation, SVD and a new singular value modification scheme. The relationship between the improvement of ill-conditioned problem and initial values of parameters is also analyzed. The improvement often increases the independence on the initial values. Resolution is derived, in order to quantitatively analyzing the independence.3. Ground surface subsidence can be divided into system part and stochastic part. The system part is described through the function of probability-integral method while the stochastic part is processed as signal with prior expectation and variance. In order to balance the covariance matrices of signals and observations, a Helmert type estimator of the variance components is derived, in which the corresponding adaptive factor is constructed by the ratio of the variance components of signals and observations. In order to avoid the influences of gross error, robust estimation is used, building the model of robust adaptive collocation. The actual numerical example shows that the accuracy of robust adaptive collocation is evidently improved.