| Soft-sensor technique can be used to measure those very important variables which are difficult to be measured due to technical and economical reasons at present time. The rare-earth separation process by countercurrent extraction is characterized of nonlinearity, time-variant properties, and severe lag. In monitoring and controlling the rare-earth separation process by countercurrent extraction, on-line information of component content is accessible. In order to acquire the real time data and contribute to real time closed loop control, this thesis applied soft-sensor technique based on radial basis function neural network according to engineering application,made the best of the capability of network's nonlinear approach and learning and got a good result. The primary content of this thesis are as follows:1. The engineering designing of soft-sensor technique is researched.2. It depicts the prccess of rare-earth separation, then analysises the factors which influence the component content. The soft-sensor model based on RBF neural network is implemented to realize online predict of component content in rare earth separation process through mechanics analysis of soft-sensor.3. Three methods to build soft-sensor model of Radial Basis Function neural network (K-means clustering algorithm, mixed Genetic Algorithm, mixed hierarchical Genetic Algorithm) is investigated and verified in simulation.In the condition of setting hidden layer's node numbers, K-means clustering algorithm and mixed Genetic Algorithm can satisfy online prediction of component content. Mixed hierarchical Genetic Algorithm can optimize the structure and parameter of RBF neural network at one time, can adjust training aims according to the error requirement and is of high efficient, is fit for online prediction of component content.Results shows that it is of important to online prediction of component content in rare earth separation process by countercurrent extraction.
|
PDF Full Text Request