Study On Data-processing Algorithm For Dynamic Weighing System

The dynamic weighing system is often used in the industrial productionoperation site, not only the environment interference; its own running speed will beactually affecting the weight accuracy. Therefore, only to improve the hardware istime consuming and energy consuming without a good achievement. We canimprove the performance of the system through the software algorithm, not only lowcosting, flexible, and easily popularized, but also with great significance.Based on the dynamic weighing system through independent research, wediscuss the algorithms for data-processing of dynamic weighing system. Because theinterferences of the operating environment are complicated, the original experimentaldata sample is rough, so, there will be two steps of the data-processing which aredata-filter-preprocessing and the forecasting of the weight forecasting model.Through data-filter-preprocessing, the interference can be almost filtered. The modelby the filter-preprocessed data could be more accuracy.Two common filter will be used in data-filter-preprocessing which are low passfiltering and Kalman filtering, and this two filtering will be used todata-filter-preprocess the sample data, then we will get preprocessing results. Get abetter data-filter-preprocessing method for dynamic weighing system.There are two different methods will be used to build the forecasting model ofthe dynamic weighing system, which are PCR (principal component regression) andLS-SVM (LS-support vector machine).The forecasting model of the dynamic weighing system will be build in theMATLAB environment to realize the forecasting, and then the mean square error andR2statistics of two models will be compared to get the better model. On one hand,when building the PCR-forecasting model, firstly, the PCA will be picked up andused in the regression. The lifted data will be used to test the model to make sure itfits the require，we get the best forecasting model whose R2is0.5784, and RMSE is103.2851; on the other hand, build the forecasting model through SVM, the samedata will be used, we get the best forecasting model using SVM whose R2is0.9676, and RMSE is84.5110.this all proves that SVM, the non-liner algorithm fits thedynamic weighing system better than others.