Development of an adaptive relevance vector machine approach for slope stability inference

摘要: Uncertainty is commonly encountered in such problems as the stability inference of slopes earth science and geotechnical engineering. This uncertainty can be approached by artificial intelligence techniques experts systems. paper presents adaptive relevance vector machine (ARVM) for soil slopes. Based on failure mechanisms due to data availability, here realized according three categories slope parameters: (1) geomaterial parameters, (2) geometry (3) pore pressure coefficient R u . A database dozens cases collected reasonably execute inference. Then, ARVM introduced approach problem. Some are used train model so that an optimized obtained. other then test ability model. Four models obtained different numbers compared show possible effects dataset size. Also, sensitivity parameters investigated. The results width hyper-parameter has apparent performance ARVMs, kernel type well size result optimal values. Meanwhile, generalized regression neural network support machines (SVM) accuracy function. suggest ARVMs have satisfactory generalization perform better than simple SVM applied networks.