Modeling of Neural Network on Moisture Test for Grain Based on Dielectric Loss Factor

摘要：

This paper gives hardware structure of moisture test system for soybean centering on TMS320F2812, and introduces the multiple sensing mechanism of online moisture test for grain and improved BP neural network. In this paper, dielectric loss factor is measured by the method of orthogonal separation on phase-sensitive detection. Though measuring three parameters of dielectric loss factor, power fluctuations and temperature changes, using BP neural network to construct multi-input single-output model, applying gradient descent method with forgetting factor for parameters adjustment of BP neural network, utilizing the nonlinear mapping ability and learning generalization ability of the BP neural network, and using high precision samples for the training of BP neural network, the mathematic model of moisture test system for soybean based on BP neural network was established finally. The sample testing experiment shows that the measurement accuracy obtains a great enhancement comprehensively considering the effect on testing output with the aim sensor characteristic and non-aim parameter. This model overcomes single sensor detection and the method of single curve fitting. Using multi-sensor detection and data processing with neural network, the built model has high measurement precision and good reproducibility.

关键词： dielectric loss factor multi-sensor BP neural network soybean moisture online test

作者: Jishun Jiang Hua Ji

作者单位: Department of Electrical and Electronic Engineering, Shandong University ofTeclmology, Zibo, China Department of Electrical and Electronic Engineering, Shandong University of Teclmology, Zibo, China

会议类型: 国际会议

会议名称: 2009 International Workshop on Information Security and Application(2009 信息安全与应用国际研讨会)

会议地点: 青岛

会议语种:英文

页码: 236-239

在线出版日期: 2009-11-21（万方平台首次上网日期，不代表论文的发表时间）