The Research On Mechanism And Compensation Method Of The Dynamic Weighing Error Of Electronic Belt Weighers

The electronic belt weigher can automaticly scale the cumulative weight of bulk materials without interrupt.It is widely used in the field of industrial production process and trade settlement.However,due to the influence of belt creep and other factors,the electronic belt weigher is difficult to keep its ostensive measurement performance for a long time(within the verification cycle).This paper studies the dynamic weighing error mechanism of the electronic belt weigher,seeks the available empirical relationship by using supervised learning methods such as neural network,and proposes weighing error compensation method of electronic belt weigher,in order to reduce the weighing error effectively.The main research work is as follows:1.The functional structure and weighing principle of the electronic belt weigher are analysed,the quantitative relationship of cross influence among various weighing error sources(influencing factors)of the electronic belt weigher are searched.The double lever type electronic belt weigher being taked as a typical example,its mechanical property is analysed.Its weighing model,tension model,formula of weighing gravity error,and formula of speed error are deduced.The conclusion that the change of belt sag has a decisive impact on the dynamic weighing accuracy of the electronic belt weigher are found out through simulation.2.The experiment device for dynamic weighing error compensation with the function of monitoring belt sag is constructed based on the test system for type evaluation of the electronic belt weigher.Multigroup tests are collected as experiment data support for the subsequent error compensation and the training of neural network model for compensating.The collected signal data are removed outliers and reduced noise through Kalman filter based on gain adjustment.The weighing gravity error,the speed error and the dynamic weighing error of the typical electronic belt weigher are compensated based on the change of belt sag.The absolute values of the relative error of dynamic weighing after compensating are all less than 0.3% and more than 6% lower than that before compensating.3.According to the characteristics of input-output relationship of the electronic belt weigher,a new idea that establishs BP neural network model and process neural network model suitable for the electronic belt weigher is proposed.Different trainning algorithms of error reduction are explored and tried.The validity and applicability of neural network model are verified and improved in the laboratory and in the field.The dynamic weighing error of the electronic belt weigher is corrected based on the process neural network model,then the absolute values of the relative error of the dynamic weighing after compensating are all less than 0.1%.4.The method that uses FPGA as hardware to realize the calculating process of process neural network method for compensating the dynamic weighing error of the electronic belt weigher is researched.The nonlinear activation function of the neural network is approximated through piecewise linear approximation.The logic design and packaging of each neuron unit and module are carried out.The top-level design of the process neural network is instantiated.The new realizing method of process neural network based on FPGA for compensating the dynamic weighing error of the electronic belt weigher is verified.The absolute values of the relative error between the verification result and the real value are all no more than 0.2%.Developing small portable device of error compensation is actively explored.