Abstract: Aiming at the problems of low assembly accuracy and difficult assembly quality detection of combine harvesters, a method of combine assembly quality detection based on sparrow search algorithm (SSA), optimized variational mode decomposition (VMD) and longterm and shortterm memory neural network (LSTM) is proposed. Firstly, the optimal VMD decomposition modal parameter K and penalty factor α are obtained by using SSA algorithm, then the vibration signal of the combine is decomposed into eigenmode components IMF with different central frequencies, and the joint features of each IMF are extracted to form a feature vector. Finally, the joint feature vector is used as the input of LSTM to realize the classification of different fault features. The analysis results show that the classification accuracy of SSA-VMD-joint feature extraction method is 98.1%, which is 7.1% and 6.1% higher than that of ensemble empirical mode decomposition (EEMD) and fixed parameter VMD, respectively, and which verifies the superiority of this method to the assembly quality detection of combine 

Key words: combine harvester, assembly quality inspction, joint feature extraction, sparrow search algorithm, variational mode decomposition, deep learning

摘要： 针对联合收割机装配精度不高和装配质量难以检测的问题，提出一种基于麻雀搜索算法(SSA)优化变分模态分解(VMD)和长短时记忆神经网络(LSTM)的联合收割机装配质量检测方法。该方法首先利用SSA算法自适应寻优得到最优VMD分解模态参数K和惩罚因子α，然后利用最佳参数组合［K，α］将联合收割机振动信号分解成不同中心频率的本征模态分量IMF，并对各个IMF分别进行联合特征提取组成特征向量，最后将联合特征向量作为LSTM的输入，实现不同故障特征的分类。分析结果表明，SSA-VMD-联合特征提取方法分类准确率为98.1%，分别比集合经验模态分解(EEMD)和固定参数VMD高7.1%和6.1%，验证所提方法对联合收割机装配质量检测的优越性。

关键词: 联合收割机, 装配质量检测, 联合特征提取, 麻雀搜索算法, 变分模态分解, 深度学习