| Modern complex industrial process systems have the characters of high nonlinearity, high dimension, signal with loss attributes, continuous and discrete variables coexisting, etc. Most of the classical real-time status monitoring methods are based on mathematical models, that are only suitable for the systems with known precision models. But in most complex systems, it is hard to get the model, and the engineers can gain some rules for status monitoring and fault diagnosis by long time practices. So we can take some machine learning and artificial intelligent methods to simulate human's inference from experience.Rough set theory is a new mathematical tool to deal with uncertain knowledge. Its main property is that the decision rules are determined by knowledge reducing and dependence analyzing on given data without any extra prior information. Rough set theory is a rule based knowledge system and does not require a mathematical description of the process.A method for complex industrial process status monitoring was presented in this thesis by combing the characters of rough set theory and complex industrial process. Firstly, a hybrid decision table discretization mehod was presented based on the classical greedy algorithm and the attribute distribution, and the characters of the complex industrial process were considered. Secondly, a sequential attribution reduction algorithm and incremental algorithm was presented by combing the idea of sequence and discernibility matrix. Finally, a real-time status monitoring method was given using the above algorithms. By considering the intrinsic characteristic of the complex industrial system, the method can avoid too many rules and low accuracy compared to the calssical rough set theory, which can avoid complex logical calculating and improve the updating efficiency of attribute reduction. Theory analysis and example simulation results show that the mehtod is efficient and feasible. |
PDF Full Text Request