A Study On On-line Recognition Technology Of Geotechnical Types Of Intelligent Excavator

With the rapid development and application of artificial intelligence technology and the huge market demand of excavator in the field of engineering and construction,intelligent excavator is receiving extensive attention.When the intelligent excavator performs autonomous excavation,it needs to identify the rock and soil types of excavation in real time,so as to adjust the excavation track in time and avoid dangerous working conditions due to excessive resistance.In addition,accurate estimation of excavation geotechnical types can further optimize the excavation strategy and improve the excavation efficiency.Aiming at the problems that the existing geotechnical classification methods are difficult to realize on-line application,this paper studies the way of detecting the bucket rod vibration acceleration signal for the first time,and designs a geotechnical classification method based on empirical mode decomposition and BP neural network off-line training combined with on-line recognition.The specific research contents are as follows:Firstly,the modeling and analysis of machine-ground integration are carried out.The composition and influencing factors of the resistance in the general excavation process are analyzed,and the excavation resistance is calculated theoretically.The multi-rigid body sub model and medium soil sub model of the working device are established by using AMESim software,which constitutes the dynamic simulation model of machine ground integration.The synchronization time series of cylinder displacement is set,and the PID control parameters are adjusted.The accuracy of the simulation model is verified by the tracking accuracy of cylinder displacement.The key characteristic parameters of three geotechnical types of loam,clay and rock are set,and the simulated excavation test and dynamic characteristic analysis are carried out.The results show that the acceleration change of the three geotechnical types of simulated excavation is significantly different,and the signal attenuation is small in the stick.So an acceleration sensor can be installed at the stick for effective signal acquisition.Then a pattern recognition algorithm of geotechnical types based on EMDBP is designed.EMD is used to decompose the double channel signal of stick acceleration and extract the signal characteristics of effective IMF,which is used as the input of BP neural network pattern recognition,and the connection weight is obtained through adaptive training.Based on different signal extraction features such as mean,peak to peak and energy factor,the neu ral network for geotechnical type recognition is trained and tested.The results show that the algorithm can effectively identify different geotechnical types,and peak to peak is the best signal extraction feature.Based on the same data set and peak to peak signal extraction features,EMD-PSO-BP is compared with several other off-line training algorithms.The results show that EMD-PSO-BP algorithm has the highest recognition accuracy and the best optimization effect.Finally,the on-line identification system of geotechnical types of intelligent excavator is built and field verification test is carried out.The acceleration sensor is selected,and the embedded signal acquisition and analysis module is designed.Relying on the test prototype of intelligent excavator,the field test platform of intelligent excavator geotechnical type online identification system is built.Based on the vibration acceleration signal data set generated by the excavation of three different geotechnical types of loam,clay and rock,the EMD-PSO-BP geotechnical type recognition algorithm is used for learning and training to obtain the parameters of neural network.The online verification test of risk aversion is carried out,and the field results verify the effectiveness and reliability of the developed system and algorithm.