基于DSP和自组织竞争神经网络的数粒机控制系统  

The Control System of Counting-machine Based on DSP and Self-organizing Competitive Neural Network

作　　者：蔡锦达[1] 唐静[1] 齐建虹[1] 李祥伟[1] 

机构地区：[1]上海理工大学机械工程学院,上海200093

出　　处：《控制工程》2017年第3期487-494,共8页Control Engineering of China

基　　金：2012国家重大科学仪器设备开发专项资助项目(No.2011YQ150040);上海市研究生创新基金项目(JWCXSL1302)

摘　　要：为满足数粒机高速高精度的控制要求,设计了以TMS 320F2812型DSP为核心的控制系统,结合线阵CCD、ARM9微型控制器、触摸屏及配套硬件对数粒机的生产流程进行控制。主要讨论了数粒机的工作原理、硬件系统的设计和HMI应用程序设计。其中,药粒CCD检测计数系统和检测通道优先级的确定是数粒机控制的难点,也是保证数粒精度的基础。采用经粒子群算法优化的自组织竞争型神经网络确定检测通道的优先级,可根据数粒环境的变化不断更新通道优先级,实现了数粒机控制系统的智能化。实际生产证明:该控制系统适用于检测各种形状尺寸的药粒、计数准确可靠,受粉尘影响小,速度快且运行稳定,应用成效明显。In order to meet the control requirements of high speed and precision for counting-machines, the control system with TMS320F2812 DSP as the core is designed, combined with linear array CCD, ARM9 microcontroller, touch screen and supporting hardware to control the production process of counting-machines. The working principle, hardware system design of the counting-machine and design of the HMI application program are mainly discussed. Among them, the pill CCD detector counting system and determination of the channel priority are the control difficulties of the counting-machine, also the basis to ensure the accuracy of the counting-machine. The self-organizing competitive neural network that has been optimized by PSO （particle swarm optimization） algorithm is used to determine the detection channel priority and the channel priority is updated constantly according to the counting environment change, to achieve intelligent control of the counting-machine. The actual production proves that the control system is suitable for the detection of pills with a variety of shapes and sizes, the counting is accurate and reliable, less affected by dust, the system works fast, stable, and achieves obvious application results.

关 键 词：数粒机 DSP CCD 自组织竞争神经网络 粒子群优化算法 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]