Research On AMT Shifting Strategy Of Intelligent Tractor Based On Working Condition Recognition

As the main force in agricultural power machinery,the intelligent control and precise operation of tractors play a decisive role in agricultural mechanized production.With the concept of intelligent farm operation,the automatic shift tractor should judge the engine output and tool operation status according to the high-performance sensor data in the field operation environment with complex working conditions,and adjust the shifting law in the face of the traction power demand of different operating tasks,so as to improve the tractor’s power,fuel economy and driving experience,simplify the operation of the tractor driver,and avoid the phenomenon of poor operation results due to the driver’s lack of driving experience.Realizing automatic working condition recognition of tractors and selecting shifting strategies that are suitable for the current job needs can free the driver from heavy shifting tasks and focus on the operation of agricultural machinery and tools,improving efficiency and accuracy in agricultural production processes.In this paper,the FJ1004 tractor equipped with a variety of sensors and data communication systems is used as the research object,and the identification of tractor working conditions and the optimization and correction of shifting strategy under multiple working conditions are related to research,and the specific research content and conclusions are as follows:(1)The tractor operation characteristics of tractor units under three typical working conditions are analyzed,and the tractor operation load test based on engine running state parameters,gearbox output shaft speed and average fuel consumption is designed,and the operation of time axis reconstruction,filter denoising and curve fitting is carried out on the test data,so as to realize the extraction of characteristic parameters under multiple working conditions of tractor,and establish the image dataset of tractor working conditions.(2)The vehicle speed,working load,fuel consumption and other information data in the field operation process are converted into corresponding working condition images,and the Res Net34 convolutional neural network model is used to train the working condition image dataset for working condition recognition classification tasks,combined with the idea of transfer learning,shorten the training cycle,avoid overfitting problems,and test through the working condition image test set,the overall accuracy reaches more than 99%,and the tractor working condition recognition function is realized.A soil specific resistance prediction scheme for ploughing operation using BP neural network model is proposed and verified.(3)Based on the software Cruise,the vehicle model was built according to the power transmission system parameters of the FJ1004 intelligent tractor,and the traction subsystem and PTO operation load submodule were built.Relying on the test of agricultural machinery appraisal station,the dynamic and economic index values of tractors were collected,and compared with the simulation data,the maximum speed error was 0.76%,and the fuel consumption error of ploughing and rotary tillage was 1.2% and 1.3%,respectively.The overall error is within the effective range,which verifies the effectiveness of the tractor operating condition simulation model.(4)Establish a Cruise-Isight joint simulation and optimization platform to develop economical and comprehensive gear shifting strategies suitable for different operating conditions.The simulation results show that the fuel consumption is reduced by 9.21% under the transportation condition and 1.28% under the light load operation condition that ensures the power demand.The shift strategy is modified based on the slip rate under heavy load conditions,and a fuzzy controller is used to correct the shift speed in the event of sudden changes in soil resistance,ensuring the stability of the tractor during driving.