A building management system includes connected equipment and a predictive diagnostics system. The connected equipment is configured to measure a plurality of monitored variables. The predictive diagnostics system includes a communications interface, a principal component analysis (PCA) modeler, and a fault predictor. The communications interface is configured to receive samples of the monitored variables from the connected equipment. The PCA modeler is configured to construct PCA models for a plurality of operating states of the connected equipment using the samples of the monitored variables. Each PCA model defines a location of one of the operating states in a multidimensional modeling space. The fault predictor is configured to determine a proximity of a new sample of the monitored variables to one or more of the operating states using the PCA models and to predict a fault occurrence based on the proximity.