Real-time Condition Monitoring System For Industrial Process Research And Development

The work in this thesis is based on three technologies of Multivariable Statistical Process Control(MSPC), the Principal Component Analysis(PCA), the Partial Least Squares(PLS) and the Kernel Density Estimate(KDE).The work involves the following contents. Firstly, A monitoring and fault detecting system is developed with OOD to implement these three technologies as a unitary system, and to make full use of their advantages. In this thesis, the OOD method and such technologies as MFC, API, DLL, and multi-thread will be talked about. Secondly, PCA and PLS can't deal with time dependent data, and ARMA model can be used to solve this problem, but the problem of ARMA structural modeling must be solved. The work elicits a genetic algorithm that is utility to fit a ARMA structural model. The using of correlations in fitness function , the floating-point-number coding function and the two-level evolving process are talked about. Finally several examples are cited to prove the advantage of this algorithm.