Research On Condition Identification And Fault Diagnosis Of Agricultural Machinery

With the rapid development of science and technology,the global level of agricultural mechanization has been increasing.As a largely agricultural country,China uses 9% of the world’s arable land to feed nearly 20% of the world’s population,which is inseparable from the dividends brought by agricultural mechanization.In recent years,with the comprehensive promotion of China’s rural revitalization strategy and the continuous development of urbanization,China’s agricultural population has gradually decreased,prompting China’s agricultural mechanization process to accelerate,and the role of agricultural machinery in modern agricultural production activities has become more and more important.However,the variable load conditions and complex working environment of agricultural machinery are very likely to induce mechanical failure and lead to performance degradation,which may reduce the quality of work and affect the driving safety of agricultural machinery.Therefore,it is necessary to research the intelligent identification of the operation status of agricultural machinery and lay the theoretical foundation for the condition monitoring and fault warning of agricultural machinery.During the working process of agricultural machinery,the vibration signal generated by the movement of mechanical structure can effectively reflect the operation status and health condition of each mechanism inside the mechanical system,and the analysis of the vibration signal generated during the working of machinery can achieve the purpose of state identification.Tractor as the main power source of agricultural production is an important indicator to evaluate the degree of modern agricultural mechanization.Therefore,this paper takes the tractor as the research object and proposes a state identification method of agricultural machinery based on vibration signal analysis,which applies to the complex excitation conditions and strong background noise environment in the field,and realizes the rapid sensing and fault diagnosis of agricultural machinery state,the specific research contents are as follows:(1)For the lack of fault signal data samples in tractor condition detection,proposed a method to change the engine speed to simulate different operating states of the tractor;for the agricultural work environment is variable,the excitation conditions are complex leading to the road environment in the real vehicle experiment is difficult to quantify the problem,proposed a method to change the tire pressure to simulate different ground conditions driving state;collected the tractor work frequent several types of fault vibration signal,for the subsequent research of tractor condition identification and fault diagnosis methods to provide data support.(2)To address the problem of difficult feature extraction of machinery vibration signals in the background of farmland environmental noise,a signal processing method is proposed to obtain the modal components containing state information by using variational modal decomposition,and then quantize each modal component to reduce the signal dimension by using permutation entropy;to effectively improve the tractor state recognition accuracy and reduce the error caused by redundant signals,a method based on correlation coefficient feature selection is proposed;support vector machine and random forest are applied to the pattern recognition of agricultural machinery vibration signals in the field operation environment.vector machine and random forest were applied to the pattern recognition of agricultural machinery vibration signals in the field operation environment,and the optimal recognition accuracy reached 96.33% and 97.66%,respectively.(3)By studying the advantages and shortcomings of support vector machines in condition recognition,an optimization method based on the particle swarm algorithm is proposed,and the accuracy of the optimized model for the recognition of vibration signals at three measurement locations is increased by 1%,7.333%,and 4.333%,respectively,and the method is applied to tractor fault diagnosis,and the comprehensive fault recognition accuracy reaches 99%.Therefore,the method is feasible to apply the support vector machine recognition method of particle swarm optimization to tractor condition recognition and fault diagnosis.