A Study For Chemical Industry Process Fault Diagnosis Based On Rough Set

It is very important to ensure the chemical industry process stability and security because the product line needs high continuity. The computer monitoring control system is widely used in modern factory, chemical industry has accumulated a lot of processing and control variable data which collected from flow process. This data set contain abundant information. How to use this data into chemical process fault diagnosis becomes a hot research issue now. In this paper, Rough Sets(RS) theory has been used in the fault data about Tennessee-Eastman Process(TEP) from three aspects, which are knowledge expression system reduction,how to use fuzzy rough set aimed at continuity attribute and neural network combined with rough sets, at last, a multilayer Neural Net model based on Rough Sets for fault diagnosis system has been founded.Firstly, studied the knowledge express sysytem reduction method of RS theory. Analysed the advantage and disadvantage of the reduction method based on core. Meanwhile, because of ant colony algorithm has the characteristic of positive feedback and distributed calculation, this paper brought forward a new method about knowledge reduction based on ant colony algorithm. The simulation results shows that this approach is an effective and quick way in solving knowledge reduction.Secondly, aiming at the inherence limitation of basic rough set which cannot deal with continuous attribute, we give the Fuzzy Rough Sets model to round away from discretization of continuous values about a decision system. Effectively use the membership degree which get from fuzziness about continuous attribute value in the reduction process, raise the information utilize rate contained in knowledge expression system, and use fuzzy-rough quick reduct algorithm into TEP fault data sets, the result is better.Thirdly, the combination status about neural network and Rough Sets was comprehensively introduced. Meanwhile, according to their specialty, a strong coupling rough-set neural network was put forward. After the rules were obtained based on rough set theory, the rules were embedded in the neural network, so that a kind of strong coupling multi-layer neural network was constructed. Finally, studied the TEP process, a multi-layer neural network was used in this process of fault diagnosis based on the fuzzy-rough theory. This type of neural network mostly exert the advantage about both rough sets and neural network, simplified the structure of neural network, raised the diagnosis speed and got a good diagnosis result.