Research On Error Source And Control Of Body In White Based On OCMM Data

In the process of automobile production,the welding error is important to the quality of the body,so it is very important to control the welding error.Fixture adjustment is the most commonly used way of controlling the welding error,but there hasn't been any effective method right now,and the adjustment process is time-consuming and inefficient.To solve the problem,improved support vector machine(SVM)model is proposed to establish the relationship between online measurement data and fixture adjustment.A method based on improved SVM to adjust the fixture is proposed.And according to the actual project need,a set of white body error source and control software system is developed.The work of this article is as follows:The characteristics of the different error sources of the white body is analyzed based on online measurement data.The error of parts and stamping parts is group-related,human and environmental factors have the characteristics of randomness and strong relationship with human.Tooling equipment lead to error with average jump,unilateral fixed up and irregular wave fluctuations in three kinds of characteristic.And finally three kinds of special error traceability model is established: random mutation,manual operation,group-related parts.Improved support vector machine(SVM)model of fixture adjustment is established.And how to choose kernel function is analyzed based on historical fixture adjustment quantity and the measuring point data.To optimize the performance of RBF kernel function,first of all coding the parameters in the binary,deciding the termination function,setting the initial population,the total evolution algebra and mutation,crossover probability,and running the algorithm until the result is up to the stop.Finally combined with the actual amount fixture adjustment model of the white body,and the neural network and support vector machine(SVM)model of the grid search method are compared.The results show that the improved support vector machine(SVM)'s mean square error is lower than other models.Based on the practical engineering case,this paper puts forward the method of adjusting fixture based on the improved support vector machine algorithm.Firstly establishing the location determination criterion.It has three steps: firstly determine the error source type,then building the relationship between points and production process.Finally,fixture adjustment probability map is established according to the history of the feature point diagnosis.Then data preparation and pre-process is established for historical data and OCMM data: firstly removing gross error,and then denoising,filling the vacancy and finally calculating the mean deviation to build the feature vector.Then sample training and selection of model: firstly using correlation analysis to extract three feature points associated with strong fixture adjustment quantity and establishing the historical samples set,then normalizing transformation and distribution,and finally giving out the quantity of fixture adjustment.According to the algorithm and method above,a white body error traceability control software integrated with on-line measurement system is developed.It consists function of basic data preprocessing,error source,fixture adjustment quantity prediction.Data input is the data of BIW's feature points.Finally the software is used in the diagnoses of BIW's circumference column above on the high side and the back door hinge installation fluctuated in X direction.The results verify the effectiveness of the algorithm and engineering value.The results show that the improved support vector machine model can improve the quality and efficiency of the white body welding error control.This method can be applied to not only build body inspection data and tooling adjustment.It also can be applied to aircraft,ship,3D printing,and other manufacturing areas.