The Research Of On-line Observation For The Unmeasured Inputs Of The Chemical System Basing On The Neural Net

To the chemical systems such as FCCU and CFBB that are non-linear, parameter distributing, variable numerous, coupling close, lag prodigious, the conformation of non-linear postural observation station do not have the uniform methods. These systems' rigidity is usually formidable, so it is quite difficult to explain the contrary system directly from the mathematic model. The methods of data distinguished such as the neural net have the favorable imminent capability, so it is an effective approach to apply the neural net to construct the contrary system to realize the on-line observation of these systems.This article firstly makes the FCCU as the study object, based on the simplified FCCU reaction-regeneration mathematic model, the simulation system of FCCU reactor-regenerator is developed on XD-APC. And use this simulation to get the industrial data of the FCCU unmeasured inputs, basing the mechanism model the experimenter build the on-line observation for the unmeasured inputs of the FCCU system and use the BP neural-net to observe the unmeasured inputs.Meanwhile, this article extends such means to the CFBB. With the simulation system of CFB boiler developed on the XD-APC, this article aims at the CFB boiler and is based on the CFBB mathematic model from which we get the industrial data of the CFBB unmeasured inputs. Making use of these data and some other industrial data, we observe the unmeasured inputs (volatile concentration and carbon concentration) with the help of BP and RBF neural-net.The research result indicates that using the neural net to construct the contrary system, the on-line observation of the unmeasured inputs of correlative chemicalprocess is available, while the effect of the observation favorable.