A Deep Learning Method to Realize The Load Sensing Control

Traditional load sensing system is based on the electromechanical feedback control, which is of inferior robustness, susceptible to oscillation and difficult to be precisely adjusted, because of the innate delay of the sensor. To solve this issue, this work proposes a deep learning method. We constructed a simplified system. Then we extracted the data sets from the sensor installed on the blender, using least square method to select the best data set as the principal of the future automatic running, whose procedures are scatter diagram generating, least squares curve generating and optimal data group selecting. Running according to the data set selected, the system can run autonomously, thus supposed to avoid the errors aroused by the oscillations and delays. This work paves a promising way to ensure the precise working of the load sensing system.