中文摘要：

采用较高频率的周期性方波信号作为载波及解调信号,通过调制与同步解调技术对应变式传感器的输出模拟信号进行变压器式隔离。将隔离后的模拟信号及传感器使用环境温度对应的数字量作为输入变量,传感器的实际负载作为输出变量,利用移动最小二乘回归（MLSR）重构传感器所受负载与使用温度及隔离信号之间的数据模型。试验结果表明,采用调制及同步解调技术的模拟信号变压器式隔离电路具有良好的温度稳定性,利用MLSR建立的传感器数据重构模型拥有比传统最小二乘回归（LSR）更高的精度,在试验条件下的温度变化范围内,采用变压器式隔离电路得到的模拟信号隔离相对误差低于±0.2%,基于MLSR的传感器数据重构模型的负载检测相对误差低于±0.07%。

英文摘要：

Used periodic square waves whose frequency was far higher than the sensor signal＇s as the carrier signal and the synchronous demodulation one, the output of strain sensor was isolated with an isolation transformer by means of modulation and demodulation. The data reconstruction module for the strain sensor was founded by moving least square regress （MLSR）. The module＇s input variables included those digital signals corresponding to the isola- ted signal and the sensor operating temperature;meanwhile the actual load for the sensor was the module＇s output variable. Experiment showed that the transformer isolation way used for analog signal isolation could get high tem- perature stability, and the reconstruction module had lower load detection error than the one based on traditional least square regress （LSR）. Within the experiment temperature, the error of the transformer isolation for the analog signal was less than ±0.2%, and the error of load detection was less than ±0.07% by using the reconstruction mod- ule based on MLSR.