Research On Classification Of Non-ferrous Metals Based On Electromagnetic Induction

The development of modern methods of metal detection and sorting as well as industrial equipment is one of the hot research issues and important challenges in metal recycling.According to the existing technology,steel can be sorted out relatively easily in the recycling process,but it is very difficult to sort out non-ferrous metals.Therefore,it is of great significance to establish effective-detection theoretical models and develop efficient automated equipment suitable for industrial environment for sorting and recycling of non-ferrous metals.This paper focuses on the electromagnetic properties of non-ferrous metals and the classification of non-ferrous metals,using eddy current sensors based on the principle of electromagnetic induction.The main research contents of this thesis are as follows.(1)For the metallic sample motion model,a non-ferrous metal classification method combining eddy current sensor and photoelectric sensor is proposed.The method mainly uses the photoelectric sensor to find the tilting angle of non-ferrous metals;meanwhile,the characteristic phase curve is found based on the extraction of local features of the mutual inductance trajectory;finally,a phase compensation algorithm is proposed to overcome the problem of phase overlap of different tilted non-ferrous metals.The experimental results show that this classification method can achieve 95% classification rate within 9 degrees of tilting angle.(2)For the eddy current sensor motion model,a non-ferrous metal classification method based on global feature extraction is proposed.Compared with the model of metallic sample motion model,in which multi-valued mutual inductance trajectories are obtained,this method has better symmetry and can help obtain single-valued mutual inductance trajectories on the complex plane;then this paper proposes a circle fitting method to extract the global features of different mutual inductance trajectories;finally the longitudinal coordinates of the center of the fitted circle are used to construct the classification criteria.The experimental results show that the proposed classification technique has high classification accuracy and can achieve 96.7% classification rate within 14 degrees.(3)A multi-frequency sloping-invariance is proposed for the eddy current sensor motion model,and the classification method based on this invariance only requires the characteristic quantity at non-tilting cases to construct the classification criteria.The method firstly finds that the pseudo-linearity exists in the mutual inductance curve of non-tilted non-ferrous metals in the specific liftoff range by theoretical analysis;secondly,it is found that this pseudo-linearity also exists in the tilted metal samples at 20 k Hz by simulation and experiment;then the intercept is obtained by fitting the pseudo-linear curve by the least squares method as the multi-frequency sloping-invariant;finally,the intercept is tested at 40 k Hz-100 k Hz and the sloping-invariance is maintained.The experimental results show that the classification accuracy based on this multi-frequency sloping-invariance is 100%,100%,96% and 100% at 40 k Hz-100 k Hz where the step size is 20 k Hz(20k Hz data as the validation set and 40 k Hz-100 k Hz data as the test set).