基于机器学习的非接触式水位计校准算法研究

Research on calibration algorithm of non-contact water level gauge based on machine learning

由于环境等因素影响,用于测量水位的超声波、雷达式等非接触式水位计存在一定的测量误差,更换高精度设备会增加成本,为解决测量误差,提出基于机器学习的水位计校准算法。该算法融合测量误差产生的环境因素,采用迭代Boosting学习算法,构建Adaboost的单层决策树模型,采用误差辗转递送的强学习算法对测量误差进行校准。算法仿真结果显示,校准结果可以很好地拟合标准设备测量值,校准算法不仅克服更换设备带来的成本,还将传统的仪器校准迁移至后端软件层面,为解决非接触式水位计校准提供新手段。

Due to environmental and other factors, non-contact water level gauges such as ultrasonic and radar types used to measure water levels have certain measurement errors. The replacement of high-precision equipment will bring extra costs. In order to solve the measurement error, this paper proposes a water gauge measurement calibration algorithm based on machine learning. The algorithm integrates the environmental factors generated by the measurement error. The algorithm adopts iterative Boosting learning algorithm, constructs single-layer decision tree model of AdaBoost, and uses the strong learning of error rolling to transmit. Simulation results show that the calibration results can fit the measured values of standard equipment well. The algorithm not only improves the efficiency of calibration, but also migrates the traditional equipment calibration to the back-end software level, which provides a new means for the calibration of a large number of non-contact water level meters in the industry.