动态称重系统的建模及神经网络辨识

Model establishment and BP neural networks identification for dynamic weighing system

为兼顾动态称重系统的快速性和精度,提出用参数辨识方法来进行处理。由于动态称重系统为时变非线性系统,因此采用分段线性化的方法推导出系统的动态数学模型,并将问题转化为参数辨识问题。在辨识算法上,采用BP神经网络不断在线辨识系统参数,从而使其工作在最佳参数下。结果表明:本试验提出方法是可行的,达到了试验提出的技术要求,测量相对误差小于2.5‰FS,不确定度小于10g,系统在全量程范围内的准确度不低于2‰。在提高称重速度的同时,也保证了系统的测量准确性。

The parameter identification is employed to solve the speed and precision simultaneously in dynamic weighing system. The dynamic weighing system which is non - linearity time variable system is equivalent to two orders system, therefore the segmented lincarity method is used to analyze and establish the dynamic mathematic model. According to the characters of the obtained dynamic： model, the parameter identification method can be used for solving this prnblem, therefore the parameter identification based on BP neural networks is employed to identify the system parameter on -line and the optimal parameter can be obtained. A series of experiments are carried out to verify the feasibility of the method adopted, and the experiment results showod that the parameter identification based on BP neural networks is capable of aehieving high accuracy t for dynamic weighing system, the relative error of measurement less than 2.5‰ Fs, uncertainty of measurement less than 10 g, and the accuracy in the full scale less than 2‰ are obtained, therefore, not only the weighing speed but also precision are achieved; Moreover the method adopted has the referential value for the practicality development of the similar weighing system to a great extent.