Research On Unknown Mode Detection In Fault Diagnosis Of Air-conditioning

Fault detection and diagnosis for HVAC system in intelligent building is the guarantee ofbuilding safety, comfort and energy-saving. Optimization of fault detection and diagnosismethod is able to improve the air conditioning efficiency, improve air quality, and reducemaintenance costs. A new idea of clustering based on gaussian mixture model in faultdetection and diagnosis is put forward on the basis of study of mature detection technologyabout domestic and foreign air conditioning system, and combination with the feature thatsame unit that has similar characteristics.A new idea that using Gaussian mixture model(GMM) to identify unknown modes onthe basis of comparision about different clustering methods and analysis about characteristicsof air conditioning is proposed. And the advantage of clustering by using the gaussian mixturemodel is introduced. Upon this method, the structure and process of the temperature andhumidity units are discussed. According to the law of conservation of energy and material,models are built.Then parameters to represent attributes are selected, and estimated throughkalman filter algorithm.These characteristic parameters are as the basis of gaussian mixturemodel. EM algorithm is the core of Gaussian mixture model, which is composed of two steps,namely Step.E and Step.M. Algorithm is realized by the iterations of Step.E and Step.M, untilthe convergence of mixture model parameters. Then the optimal estimated result is obtained.After the gaussian mixture model is built, distribution of data points and identification ofmodes is according to the maximum a posteriori probability criterion.Using simulated air conditioning system platform, the gaussian mixture model ofclustering method is verificated. By the MATLAB program, the corresponding algorithm isdesigned. Testing effect based on gaussian mixture model for air conditioning unit can begotten from simulation experiments.