With the continuous improvement in agricultural machinery technology in China, the mechanical level of tractors in the south has been rising year by year [
1]. However, at the same time, the rural labor force has decreased, which has put forward new requirements and challenges for agricultural machinery technology to meet. Unmanned and intelligent agricultural machinery has become an inevitable trend. Through the intelligent transformation of tractors [
2], agricultural machinery can better meet the development needs of Chinese agriculture. In terms of tractor condition recognition, there are few relevant studies, and most of them are based on the use of tractor spatial position information [
3,
4] data to study tractor conditions. The 13th Five Year Plan for the development of intelligent manufacturing in China proposes strengthening innovation around key common technologies; focusing on perception, control, decision making, execution and other functions of intelligent manufacturing systems; and researching and developing corresponding intelligent manufacturing core support software to provide technical support for the intelligence of the production equipment and processes. Perception, as a prerequisite for intelligent implementation, is the foundation of control, decision making, and execution. Machine perception is the use of machines or computers to simulate, extend, and expand human perception or cognitive abilities. Technical forms of this perception include machine vision, machine hearing, machine touch, etc. The use of signal processing and condition recognition technology to monitor the conditions of tractors can provide references for improving the conditions of agricultural machinery, provide safety guarantees for agricultural production activities, and provide technical support for the construction of intelligent and unmanned modern agriculture. Li Jingyao et al. proposed the use of tractor operating condition recognition to diagnose the direction of tractor mechanical faults, while Deng Tao et al. used operating condition recognition to assess the adaptive energy management direction of hybrid vehicles. Turson Maimaiti et al. proposed using density clustering, combined with agricultural machinery operation status characteristics, to cluster tractor spatial position information in order to recognize tractor working conditions. Wang Pei et al. [
5] proposed a method for identifying the operating state of typical tractors based on data mining and spatial data analysis methods, using tractor spatial location information data. However, the above studies indirectly determined the operating state of tractors from a spatial perspective by analyzing the characteristics of their spatial trajectories, which had certain limitations. In terms of studying the conditions of tractors, in addition to monitoring spatial location information and using other indirect methods, it is also possible to collect a variety of parameter data from the tractor itself for use in condition recognition. Deeply mining the status information of the tractor can more directly and clearly identify the conditions of the tractor, thereby further satisfying the fine management of tractors [
6] and their unmanned and intelligent transformation, and helping agriculture to achieve electrification, intelligence, networking, and digitalization as soon as possible, thus comprehensively promoting rural revitalization [
7]. The acceleration and angular velocity of the lifting arm in the three-point suspension system of a tractor will vary with the different conditions of the tractor, which will affect the detection of the angle of the lifting arm in the three-point suspension system. The angle of the lifting arm in the three-point suspension system [
8] is linearly related to the depth of tillage. In order to determine this difference more accurately and improve the accuracy of agricultural machinery-related tillage depth detection, it is necessary to judge the conditions of the tractor.
Pattern recognition using neural networks [
9] is a recognition technology based on neural networks. It achieves the automatic classification and recognition of various patterns through learning and the simulation of large amounts of data. Pattern recognition using neural networks has been widely applied in various fields, providing more efficient and accurate solutions for people. The use of neural networks for tractor condition recognition essentially relates to the application of pattern recognition in practical engineering. Tian Yi et al. [
10] established a driving condition recognition method based on fuzzy neural networks, which identified the driving conditions of the main roads in Guangzhou and Shanghai. The mainstream classification methods are basically distinguished according to the methods used for condition recognition. Neural network, fuzzy control [
11], clustering analysis [
12], and other methods provide a theoretical basis for the development of condition recognition.
This article is based on research into neural networks in order to improve tractor condition recognition. Based on the analysis of the measured data, it is necessary to improve the relative values of the parameters of the tractor body and the three-point suspension system in order to identify the tractor condition and comprehensively analyze the motion parameters of the tractor body and the three-point suspension system. Generally speaking, traditional control systems are based on mathematical models. However, in some special cases, the mathematical models of control systems and control objectives do not exist or are difficult to obtain, which causes many inconveniences in efforts to solve problems. In recent years, intelligent control has developed rapidly and has been widely applied in automation, electronics, and other industries. Neural networks are among the best methods of intelligent control. In theory, the application of neural networks mainly refers to two aspects: one is the perception of the surrounding environment through various sensors, and the other involves taking the next step according to the control strategy. For tractors equipped with a three-point suspension system, it is necessary to consider using the results of condition recognition to determine the magnitude of the angle difference between the three-point suspension system and the tractor body under various conditions. Therefore, it is necessary to conduct research and innovate theoretically. Because of this, condition recognition and neural network also need to cooperate and influence each other.
In summary, among the mainstream algorithms for multi-parameter pattern recognition in complex environments, neural networks have wide scope for application, strong applicability, and high accuracy. Therefore, this article selects LVQ neural networks and CNNs with good performance in tractor condition recognition in order to verify the feasibility of the algorithm.