介绍了目前大坝安全监测数据处理工作中几种常用粗差识别技术的优缺点及适用范围，并针对这些常规方法中所存在的不足，采用时空判别技术和基于稳健性处理方法的粗差识别技术，对粗差数据进行判别。其中时空判别技术充分利用了观测序列本身的时空基本信息，将观测值与历史的或相邻的观测数据相比较来判别粗差；基于稳健估计算法的监控模型判别法克服了经典的最小二乘法所存在的抗粗差干扰性差这一缺点，在最小二乘回归的基础上逐步按残差大小加稳健化权，迭代求得模型参数的稳健估计，这一估计值最接近于无粗差影响时的正常估值。隔河岩大坝安全监测数据的实例分析表明，这些方法具有较强的粗差识别能力。

 The authors introduce a couple of prevailing gross error identification techniques adopted in the process of current dam safety monitoring data treatment and analyze their respective advantages & disadvantages as well as their application ranges. In combination with the shortcomings of those techniques, the gross error data were identified by using space-time identification technology and robustness-based technique. Space-time identification technology utilizes fully the basic space and time information of the data series and compare data with historical data or adjacent observation data to identify the gross error. The identification approach of robustness-based monitoring model overcomes the disadvantage of poor gross error interference immunity of traditional least square method and obtains robust estimation through iterative calculations with gradual weighted robustness. The estimation figures proved to be the most closely to the normal data without interfering by gross errors. The case analysis of safety monitoring data of Geheyan Dam verifies that the above approaches are capable of identifying gross errors outstandingly.