Contact State Recognition Method For Shaft Hole Assembly Based On Novel Support Vector Machine

With the continuous development of robot shaft hole assembly in the industrial field,accurate and fast automatic assembly tasks have become an important subject in recent years.The contact state is the key information in the assembly task,and the intelligently identified method can automatically identify the contact state collected during the assembly process.Support vector machine is an efficient and accurate intelligent recognition method,and the novel support vector machines have a great advantage in classification performance.Therefore,the research of shaft hole assembly contact state recognition based on novel support vector machines is an important subject.This paper first analyzes the basic principles of support vector machines,and establishes algorithm models for density-dependent quantized least squares support vector machine and weighted linear loss twin support vector machine.The advantages and disadvantages of these classifiers is studied in detail,and the influence of internal parameters on its classification is analyzed.Based on this,the working principle of parameter optimization of the novel support vector machines is introduced.According to the characteristics of shaft hole assembly,an experimental platform for typical shaft hole assembly is established,the influence of different material parts on the recognition accuracy is analyzed,and the robot assembly action of shaft hole parts is designed based on this.Using RT Tool Box2 and MATLAB software to design a contact state recognition system,complete the collection and analysis of six-dimensional force data,and describe the novel support vector machine applied to contact state recognition.In order to solve the problem of various contact states of shaft hole assembly,this paper derives a multi-classification model based on density-dependent quantized least squares support vector machine.Experiments show that the multi-classification model can conduct fast and effective pattern recognition.In addition,this paper improves the whale optimization algorithm and proposes a logistic global whale optimization algorithm.This algorithm successfully determines the optimal internal parameters ofdensity-dependent quantized least squares support vector machine and weighted linear loss twin support vector machine.The experiments show that the novel support vector machines based on the logistic global whale optimization algorithm realize the rapid recognition of the contact state of the shaft hole assembly.Compared with other contact state recognition classifiers,it has obvious advantages in accuracy and calculation speed.In order to further verify the superiority of the novel support vector machine,the weighted linear loss twin support vector machine based on logistic global whale optimization algorithm is applied to complex low-voltage electrical shaft hole assembly experiments,and this method is used to diagnose the actual abnormality of the assembly process.Experimental results show that the proposed novel support vector machine can quickly and efficiently perform the abnormal diagnosis task of the low-voltage electrical appliance shaft hole assembly.