| With the continuous improvement of the level of manufacturing industry,industrial robots have been widely used in the manufacturing industry because of their advantages of reliable work and high work efficiency.Harmonic reducer has the advantages of large transmission ratio and high transmission accuracy,which is widely used in the end joint of industrial robot.As the core component of industrial robot,the high precision operation of harmonic reducer ensures the precise action of industrial robot,and its health state directly determines the operation accuracy of industrial robot.Therefore,it is necessary to assess the health state of harmonic reducer.The mechanical arm of industrial robot usually makes periodic reciprocating motion.At the same time,the frequent start,acceleration and deceleration in the reciprocating motion lead to the speed transient of the harmonic reducer,which makes the obtained signal shows typical transient characteristics,which brings great challenges to the operation state characterization and health state assessment of the harmonic reducer.Based on this,the paper takes the health state assessment of industrial robot harmonic reducer as the goal,and carries out related research around the operation state information characterization,multi-sensor information fusion and other aspects.The main contents are as follows:(1)Aiming at the problems that it is difficult to describe the industrial robot harmonic reducer running state and assess its health state due to its cyclic motion and speed transients,the health state assessment method of harmonic reducer based on the integer-period data and convolutional neural network(CNN)is proposed.First of all,the integer-period data sample of harmonic reducer is used to describe its running state information accurately.Secondly,continuous wavelet transform(CWT)is used to decompose the integer-period data to fully show the transient characteristics of the harmonic reducer in the operation period.Finally,the convolutional neural network is used to fully learn the transient characteristics of the harmonic reducer so as to realize the health state assessment of the harmonic reducer.(2)As the working conditions of industrial robot harmonic reducer are cyclic,it is difficult to depict the whole operation state of harmonic reducer only by relying on a single sensor,which will lead to high uncertainty of health state assessment results.A method for assessing the health state of harmonic reducer based on the deep fusion of multi-sensor information was proposed.The image fusion based on wavelet transform is used to fuse the time-frequency information of multiple sensors to describe the operation state of harmonic reducer.Convolutional neural network is used to automatically learn the depth features of fused images to realize health state assessment.(3)Because the current health state assessment of industrial robot harmonic reducer mostly takes vibration signal as the carrier,additional testing system is needed,which increases the difficulty and cost of data acquisition.At the same time,its accuracy and effectiveness are affected by the location of sensor installation.Based on this,the health state assessment method of harmonic reducer based on voltage signal depth feature learning is proposed.The motor voltage signal of industrial robot is used to characterize the health state of harmonic reducer.Continuous wavelet transform is used to transform voltage signals into time-frequency graphs.Finally,convolutional neural network is used to self-learn time-frequency information of voltage signals to realize health state assessment.At the end of the paper,the work of this paper is summarized and the next research is prospected. |
