Research On Decoupling Algorithms Of Six-Axis Force Sensor Based On Neuron Theory

Six-axis force sensor is important accessory of industrial robot, which can test full force information within three dimensional space, ensuring robots have sensory ability of force and exchange information with outside environment safely and efficiently. But as a consequence of sensor design theory, manufacturing, mount technology, coupling happens among channels in six-axis force sensor, which limits application of sensor in high accurate occasions. So it is imperative to study coupling while optimizing structure of six-axis force sensor.Focused on problems of complicated modeling process, large calculated amount, halfway decoupling, difficulty of achieving project in six-axis force sensor, this thesis intends to explore modeling of six-axis force sensor and decoupling of six channels information. Comprehensively using many kinds of theory of sensor technology, artificial neural network, process neural network, information acquisition, etc, this thesis deeply analyzes key problems about calibration experiment, static decoupling, dynamic decoupling and so on. Mainly based on decoupling methods of artificial BP neural network to study static decoupling of six-axis force sensor and based on decoupling methods of process neural network to study dynamic decoupling of six-axis force sensor, this thesis takes advantage of unlimited approaching of artificial neural network and discuss using intelligent calculation of ant algorithm, particle swarm algorithm, etc to optimize neural network calculation and apply to decoupling of six-axis force sensor to increase effect of decoupling. Main content are as follows,1. I came up with decoupling methods of six-axis force sensor based on BP neural network calculation and revealed mapping relation between input and output of sensor to realize decoupling among signals of six channels and discussed influence of neural number in different hidden layer on decoupling effect to certify feasibility and effectiveness of decoupling methods.2. Focused on these problems showing up in the process of network training, I came up with improved algorithm to optimize BP neural network based on ant algorithm and particle swarm algorithm to make sure BP neural network can get effectiveness and high accuracy in decoupling of six-axis force sensor.3. I proposed dynamic decoupling study of six-axis force sensor based on process neural network. Through establishing model of dynamic process neural network of six-axis force sensor, introducing a suitable standard system of orthogonal function to spread binomial of network input component, discussing feasibility of using binomial coefficient to express mapping relation between input and output of six-axis force sensor and training neural network, I certify that process neural network is feasible in dynamic decoupling of six-axis force sensor.Research achievement above realized effective decoupling of six channels signals in six-axis force sensor, increased decoupling accuracy effectively and provided a new way to decouple multidimensional force sensor.