Breakout Prediction Model Based On PSO Optimization Of BP Neural Network

With the development of continuous casting technology, the high efficient continuouscasting technology has become a main research direction in the field of continuous castingtechnology. The high casting speed is the core of high efficient continuous casting.However, with the increasing of the casting speed, the risk of breakout is getting higher.Once the breakout accident happening, it will bring about serious consequences to theenterprise. Employing the real-time and effective breakout prediction system is the mosteffective means to solve the problem. Breakout prediction model is the core of thebreakout prediction system. Therefore, it is of important theoretical and practicalsignificance to study the prediction model.Firstly, in the thesis, the formation mechanism of sticker breakout is discussed andthe preventive measures are put forward according to the inductive factors, then theprediction principle and the spread of fracture are analyzed according to the formationprocess of sticker breakout.Secondly, the BP neural network algorithm is studied, which is the most commonlyused in neural network breakout prediction model. Aiming at the defects of BP algorithm,PSO (Particle Swarm Optimization, PSO) algorithm is introduced into the training processof it, and the PSO algorithm is improved. Then the breakout prediction model is set upbased on PSO optimization of BP neural network, which include time series networkmodel and space network model. The time series network model is designed to identifytemperature changes of single thermocouple on time sequence. If the time series networkmodel alarms, the space network model will identify whether the change pattern exists inthe surrounding thermocouples, then the breakout alarm information is generated bylogical judgment.Finally, based on the Windows platform, using the programming software MicrosoftVisual C++6.0, the continuous casting breakout prediction simulation system is developed,and BP neural network based on PSO optimization algorithm is used. Then the system istested in laboratory, results show that the continuous casting breakout prediction simulation system can reduce the false alarm and avoid the missing alarm effectively, andthe breakout prediction model based on PSO optimization of BP neural network can beused in the breakout prediction system.