In this talk we will present a framework for analyses of building energy models including uncertainty and sensitivity analysis, optimization, calibration, and failure mode effect analysis. The methodology begins with efficient uniformly ergodic numerical sampling and regression analysis based on machine learning to derive an analytic representation of the full energy model (e.g. EnergyPlus, TRNSYS, etc). Once these steps are taken, and an analytical representation of the dynamics is obtained, multiple avenues for analysis are opened that were previously impeded by the mathematical properties of full energy models (discontinuities, computation time, etc.) including sensitivity analysis (1000’s of parameters), model reduction, derivative based optimization and model calibration, as well as Failure Mode Effect Analysis. The results from test cases on energy models of actual real world buildings will be discussed at length. In this talk, a short discussion of Modelica activities at UCSB, as well as spectral methods for building data will also be presented time permitting.