Research On Fault Diagnosis Of Rotary Wiped Film Evaporator

Distillation chemical industry is a pillar industry of our country, it makes a great contribution to the economic development of the country every year. It also can affect many aspects like energy, materials, etc. As a large complex automation processes, Distillation process is also one of the extremely dangerous industry. In the production process of the distillation, It may cause a huge disaster if some minor fault is not solved in time. Production equipment becomes more perfect and more complex with the development of technology, improvement of the level of automation and the highly integrated distillation processes. Maintenance personnel is difficult to find the point of failure quickly when the equipment is faulty. Resulting in extended repair time, the entire production process stalled and the equipment will have a negative impact at the same time. In order to improve system security and reliability, the use of fault diagnosis technology is extremely necessary.Bayesian Network is a research hotspot troubleshooting directions.This paper studies the fault diagnosis system of rotary wiped film evaporator, considering the complexity and the danger of the evaporation process,According to the entire process of evaporation, Design the Diagnostic Bayesian Networks(DBN) to match it. Considering The evaporation device parameters as nodes on Bayesian Network. According to the causal relationship between variables nodes, we can Determine the network structure of the Bayesian network. Train the Bayesian network through fault sample to get the conditional probability of all the child nodes.The trained network can accurately determine the cause of the vast majority of failures Symptom.In the process of establishing the diagnosis Bayesian network, The Conditional Probabilities Table(CPT) of Each node in the network obtained from the training samples data We need to give a priori probability for the nodes with no parent variable, Taking into account the Impact on the final diagnosis prior probability by Bayesian network. In order to improve the accuracy of prior probability, according to The operating state of rotary wiped film evaporator Forecasting the evaporator environmental parameters by using BP neural network prediction. Achieve the goals to improve the accuracy priors and improve the accuracy of the diagnosis of the whole Diagnosis Bayesian Network. In order to meet experts in the field of thinking and reduce the instability caused by subjective reasons when giving the priori probability, giving the priori probability values of Diagnosis Bayesian network by way of fuzzy reasoning. The simulation results of the model, The Diagnosis Bayesian Network before optimization made error judgments on those similar symptom cause of the closer points of failure,The overall prediction error rate of Bayesian networks is around 10% before optimization. But the Diagnostic Bayesian Networks can quickly reduce the error rate after optimization, Overall prediction error rate can be reduced to about 5%, Achieve the goals to enhance the diagnosis accuracy of fault by Diagnosis Bayesian Network.