基于PSO-BP神经网络的既有桩基极限承载力预测研究

打开文本图片集

Abstract：Aimingatthe dificulty in evaluating theultimate bearing capacityofexisting pilefoundations forreuse，this paperanalyzes the ultimate bearing capacityof existing pile foundations throughon-site static load tests and numerical simulations involving 13existing pile foundations.Basedon230datasamples，this study predicts theultimate bearing capacityof single pilesusing backpropagation （BP）and particle swarmoptimization-BP（PSO-BP） neural networks，and then evaluates the prediction results using three metrics： coefficient of determination （R2） ，root mean square error （MAE）， and rootmean absolute error（RMSE）.The results demonstrate thattheultimate bearing capacitiesofall9destructively testedpilesis2to3times thereusedesignvalueforpilefoundation，indicatingarelativelysuficientsafetymargin. Furthermore，therebound rateof the4non-destructivelytested pilesexceeds 80% ，andnumerical simulationsconfirmthat theultimate bearingcapacityof each single pile is greaterthan the reusedesign value，verifying the feasibilityof their reutilization.Acomparativeanalysisof the prediction models reveals thatthecoefficientof determinationofthe PSO-BP model increases by 196% compared to the traditional BP model，while the MAE and root mean square error RMSE decrease by 66% and 62% ，respectively.The prediction errorsare mostlycontrolled within 1±2000kN .Theresearch findings providea scientific basis for the eficient evaluation of bearing capacity and reuse of existing pile foundations.

Keywords：Existingpile foundation; ultimate bearing capacity;bearingcapacity prediction;PSO-BPneuralnetwork

随着我国高速公路网络的不断扩展，大量既有高速公路迎来了改造升级。（剩余9964字）