Breakout accident

Breakout accident is great danger of casting production in the production of the accident. Breakout accident, not only affects the continuous casting, serious damage to the mold, roller, increasing the amount of continuous casting equipment repair and maintenance costs, resulting in huge economic losses, but also a huge security risk. It is estimated that a "normal" breakout accident resulting directly or indirectly, the economic loss will reach $ 200,000.

To reduce the breakout occurs, it has been developing breakout prediction system. By sensors mounted on the mold, it has been able to use automated control system to effectively reduce the breakout rate.

The first breakout prediction are: (1) estimates mold water temperature; (2) detection of mold and the friction between the slab; (3) buried in the local heat flux sensor monitoring; (4) in the copper mold embedded copper thermocouples for thermal monitoring. This method is now widely used, proven to be more effective in detecting and forecasting bond breakout technology. Based on temperature detected for breakout prediction of the theory is: As the breakout is often accompanied by copper mold temperature, mounted on copper by monitoring the temperature of the thermocouple on the changes made according to certain methods to determine the breakout.

Overview of the breakout prediction based on temperature prediction methods, there are three main technical approaches. They are the logic method, neural network methods and multivariate statistical methods.

Most of the new steel mills in China in recent years on the casting machine used in breakout prediction system is based on logic approach produced steel pipe,like seamless steel tube and piling pipe for construction. Implementation of the forecasting method does not require a lot of raw data, especially for the new casting machine. But the logic to set the parameters and conditions of the type of steel, technology and the environment have a great relationship, the model has some limitations.

Neural network breakout prediction method using neural network technology to the scene of breakout data for learning, coupled with the process conditions necessary for a breakout to identify the characteristics of the thermocouple temperature curve, and then make a breakout prediction . Complete neural network system must be able to significantly increase the breakout prediction accuracy, but the neural networks that exist in the breakout prediction of its limitations. Complete neural network must be a large number of training samples, the sample will not lead to a comprehensive network of dysfunction. Second, continuous casting machines working conditions are very complex, many factors, access to sample data on serious interference, in order to be completely ruled out with a sample of the noise is almost impossible. Casting in some special circumstances, such as open water, for the final package and pouring so neural network system to become unstable; for the new continuous casting machine, neural network is not feasible, must be combined with other methods.

Multivariate statistical process control (MSPC) method is based on principal component analysis (PCA) and partial least squares (PLS) method for continuous monitoring of process control methods. In the continuous casting process, especially in the breakout prediction process, MSPC method very few applications.