Research On Fault Diagnosis Technology Of Industrial Robot Reducer Based On Reservoir Computing

Industrial robots have the advantages of programmability,anthropomorphization,and versatility,and are being widely used in various fields.Precision reducer is a key component with the highest technological content in industrial robots,and its quality directly determines the performance of the robot.But in the field of fault diagnosis research,the research on transmission systems such as precision reducers is not perfect.This subject is aimed at the transmission system failure of BRTIRUS1510 A six-degree-of-freedom industrial robots.Using attitude monitoring method,the BWT901 attitude sensor is used to complete the attitude data acquisition process,and fault diagnosis research and application are conducted from a data-driven perspective.In this research,an information acquisition method of installing an attitude sensor at the end effector of an industrial robot is proposed.Combined with machine learning theoretical algorithms,a fault recognition model for industrial robots is established,and fault diagnosis applications are developed.The main research work carried out in this topic is as follows:(1)Through analysis of the structure of industrial robots and transmission failures of key components,the common failure forms are derived.Considering the economics and feasibility of fault diagnosis,fault diagnosis methods based on reservoir computing and low-cost attitude sensors are used to carry out fault diagnosis of industrial robots.Carry out the parameter analysis of the reservoir computing model and compare it with support vector machines,sparse autoencoders and deep belief network models.The comparison results show that the reservoir computing model is feasible and effective for industrial robot fault diagnosis under different working conditions,and the reservoir computing algorithm has good stability in industrial robot fault diagnosis.(2)In industrial practical applications,the motion accuracy requirements of industrial robots are generally very high,so fault diagnosis models with higher recognition rates are vital to the positioning accuracy of industrial robots.Because the raw data from low-cost attitude sensors usually contain a lot of interference information,it will increase the difficulty of the model for industrial robot fault diagnosis and analysis.Therefore,a novel pre-classification reservoir computing method is proposed.The pre-classification strategy to reduce the distance between classes by summarizing the information labels with the same conditions can achieve the purpose of improving the accuracy of the model's fault classification.The parameter analysis of the pre-classified reservoir computing model was carried out and compared with the standard reservoir computing,support vector machine,sparse autoencoder,and deep belief network model.Experimental results show that the pre-classified reservoir computing method has the highest failure mode recognition rate among these comparison methods.(3)Although the intelligent algorithm based on the pre-classified reservoir computing is feasible for the recognition method of industrial robot failure modes,and has higher recognition accuracy than the traditional reservoir computing algorithm,it still needs time for the fault diagnosis model from the time level.Research into the problem.The analysis found that the time taken by the model from the beginning of the operation to the end is related to the dimension of the input data.If the dimensionality of the data is too high,the time that the computer spends during the calculation will increase.Therefore,a new calculation method of reduced-dimensional reservoir computing is proposed and compared with pre-classified reservoir computing,standard reservoir computing,support vector machine,sparse autoencoder,and deep belief network.Experimental results show that the proposed reduced-dimensional reservoir computing model can shorten the computing time of the diagnostic model in industrial robot fault diagnosis,thereby achieving the purpose of saving time and cost.