Development of Tribosensor System Using Neural Networks

本课题的目的是开发一种具有手指感的软智能触觉传感器系统。该传感器首先通过有限元分析进行设计，然后在铝管周围铺设海绵橡胶片和PVDF薄膜传感材料。为了保护聚偏氟乙烯薄膜，提高灵敏度，引入了一种玻璃纸胶带和一块纱布来冒充指纹。针对表面粗糙度信号的复杂性，提出了十点平均高度法、对称化点阵法和快速傅立叶变换积分法三种数据处理方法对表面粗糙度进行定性和定量评价。给出了在超细纸、报纸和卫生纸上的实验结果，发现该触觉传感器对这些纸张的识别是有效的。在此基础上，研制了两指柔性微型夹持器，用于通过主动传感来识别微小柔软物体的特性，并用于物体的力调节。夹持器由两个柔性手指组成，由压电双晶片驱动。在一根手指的表面上粘贴了半导体应变计。通过施加测试阶跃电压到压电致动器来驱动夹持器夹住物体时，用应变计测量对称和反对称模式的固有频率。然后通过测量的固有频率来评估对象的质量和刚度。最后，根据被控对象先前测量到的信息，利用PID和模糊控制算法设计了抓取被控对象的力控制器。理论和实验结果均表明，本文提出的识别方法和柔性微小物体力觉调节器的设计方法简单、实用有效。

The purpose of this project is to development a soft intelligent tactile sensor system having the sense of human's finger. The sensor is designed first on FEM analysis and then constructed by layring the sponge rubber sheet and the sensory meterial of PVDF film around an aluminum pipe. For protection of PVDF film and improvement of the sensitivity, a cellophane tape and a gauze are introduced pretended as fingerprint. Since the signals are very complicated, three data processing methods, such as, ten points average height, symmetrized dot pattern method and FFT integration method, are proposed to evaluate the surface roughness both qualitatively and quantitatively. The experimental results obtained on superfine paper, newspaper and toilet paper are presented and it is found that the proposed tactile sensor is efficient for recognition of these papers.Further, two-fingered flexible miniature gripper was developed for identification of the properties of minute soft objects by active sensing and for force regulating pinch of the objects. The gripper is consisting of two flexible fingers actuated by piezoelectric bimorph strips. A semi-conductor stain gauge is bonded on the surface of one finger. The natural frequencies of symmetric and antisymmetric modes were measured by the strain gauge when the gripper is driven to pinch an object by applying a testing step voltage to the piezoelectric actuator. The mass and stiffness of the object are then evaluated by the measured natural frequencies. Finally a force regulator for grasping the object is designed by the PID and Fuzzy control algorithms based the information measured previously from the object. Both theoretical and experimental results show that the proposed identification method and the design of the force regulator for soft minute objects are simple and efficient in practical use.