变形数据处理中,由于建模过程中存在各种近似,所建立的平差模型不可避免地存在模型误差。论述模型误差对参数估值的影响,提出利用半参数补偿最小二乘方法对模型误差进行求解,讨论正则矩阵R和平滑参数α的选取对变形预测结果的影响,通过实例将半参数模型与时间序列模型、灰色模型、小波灰色模型的结果进行比较。结果表明,半参数方法能有效地处理变形平差系统中存在的模型误差,具有较强的预测能力。
In deformation data processing,due to various approximations existing in the process of building model,it is inevitable that the model errors exist in the adjustment system. We discusses some theoretical problems affecting parameters estimation, proposes that model error can be solved with semiparametric penalized least squares and discusses the choice of regular matrix R and smoothing parameter α affecting the forecasted results. Finally,an adjustment problem in dam safety monitoring is used to test the method and the semiparametric model, autoregressive model,grey model and grey wavelet analysis method are compared with each other. The results demostrate that the model errors can be compensated correctly by semiparemetric method.