Research On Monitoring Technology Of Beer Brewing Process Based On Support Vector Machine

The process of modern industrial production is generally a nonlinear, timevarying, strong coupling and complex system. It is difficult to establish a precise process monitoring and analyzing model. With the development of sensor and computer technology, the control system can record large amount of data, which lays the foundation for the establishment of the statistical analysis model of production process monitoring. This paper takes the typical batch process object----beer fermentation as the object, studying batch process fault diagnosis and monitoring model. The main contents and work of this research are as follows:With standard support vector machine(SVM) based least squares support vector machine fault diagnosis, for least squares support vector machine(SVM) robust sparsity and the lack of problems, the weighting method and pruning method respectively to the least squares support vector machine is improved, makes the diagnosis model with the standard support vector machine similar robust and sparse of, to enhance the ability of anti-jamming, diagnosis speed is improved, and through simulation experiments verify the effectiveness of the above method.For support vector machine diagnosis model parameter selection lack of theoretical guidance of this problem, the particle swarm optimization algorithm, the model of penalty factor and kernel parameter for optimization, which make the choice of model parameters directly, quickly, and diagnosis model generalization ability is enhanced.For fermentation process of beer production in different operation period of running state, and puts forward a phase weighted least squares support vector machine fault diagnosis model based on the algorithm of different periods of the operation models are established respectively. To avoid the use of a single model can not very good performance of original data information, lead to the lack of someimportant information loss, to improve the accuracy of fault diagnosis, can be timely and accurate diagnosis of the fault type. The effectiveness of the method is verified by off-line modeling and on-line diagnosis.