|
Abstract A kind of random weight particle swarm optimization(RandWPSO)-least squares support vector machine (LSSVM) prediction model of limit displacement of large underground cavern is established and applied to limit displacement prediction and analysis of large surge shaft of Nuozhadu hydropower station. The reliability of the model is validated by a case study. Through comparing with the results of numerical calculation，the surrounding rock stability of engineering area is evaluated comprehensively. The results show that the maximum relative error of limit displacement obtained by RandWPSO-LSSVM prediction model to the real value is 6.72%. The prediction error is quite less；and the prediction effect is well and meets the engineering requirements. The maximum displacement and tensile stress obtained by numerical calculation under condition of design support parameters are 19.45 mm and 0.54 MPa，respectively，mainly located at the five-tunnel area sidewall. The maximum tensile stress is far less than the tensile strength of little weathering rock mass；and the plastic zone distribution is less. The maximum displacement obtained by numerical calculation is less than the predication value obtained by RandWPSO-LSSVM prediction model. The surrounding rock of engineering area is stable. The results have great significance for estimating the stability of large underground cavern and will provide important guidance for making construction decisions reasonably.
|
|
Received: 09 April 2012
|
|
|
|
2012, Vol. 31 
